SLEEP

, de Zambotti, Massimiliano;Barresi,, Giacinto;Colrain, Ian, M;Baker, Fiona, C

doi: 10.1093/sleep/zsaa178pmid: 33009913

In the digital health era, the accessibility and advancements in virtual reality (VR) technologies (e.g. high-resolution head mounted displays, high frame rate, wide stereoscopic binocular field of view, and low latency), have turned the spotlight on VR as an emerging revolutionary diagnostic and therapeutic approach for several disorders, including phobias, depression, post-traumatic stress disorder (PTSD), and acute and chronic pain [1, 2]. VR could be considered a computer-generated simulation of a physical world, acting through immersion (level of sensory fidelity provided by a VR system) and presence (the user’s subjective and phenomenological response to a VR system). These elements lead to a sense of “being there” in the VR place (place illusion) and to the belief that events in VR are really happening (plausibility illusion) [3]. VR systems provide the capability for designing and controlling simulated interactive environments, which can be used to modulate one’s emotional state and physiology [4]. For example, in healthy undergraduates, Riva et al. [5] showed that even short exposure (about 3 min) of participants to anxious, relaxing, and neutral VR environments led to pre-to-post-immersion changes in participants’ mood. Relaxing environments elicited greater perceived quietness and happiness and reduced anger, sadness, anxiety, and negative affect than neutral VR; opposite changes were detected after anxious environments, with the sense of presence being implicated in the participants’ emotional experience. While the majority of VR systems are designed for entertainment (e.g. gaming), the use of VR is growing within the health space. Initial studies of VR have shown promise in advancing the understanding of mental health disorders and to elicit specific therapeutic effects (e.g. VR exposure-based therapy for treating specific phobias, anxiety disorders, and PTSD; VR-based imagery rehearsal for nightmares; and VR as a distracting tool for acute and chronic pain) [1, 2]. For example, McNamara et al. [6] found that 4 weeks of use (twice each week) of an immersive VR intervention (ReScript) involving the manipulation of threatening images (to make them less threatening and improve mental control over the image) led to pre-to-post-intervention changes in perceived anxiety, nightmare distress, and psychosocial impairment due to nightmares, in 19 adult men and women reporting frequent nightmares. While still in its infancy in the health field and requiring research and validation, we argue that VR has the potential to be useful as a clinical tool to promote sleep, facilitating the falling asleep process in those individuals with difficulties initiating sleep (e.g. sleep-onset insomnia). VR could work by distracting and immersing the user at bedtime in simulated relaxing realities designed to induce and control specific changes in mood and physiology that lead toward sleep-facilitating psychophysiological levels. Falling asleep is a complex process characterized by several psychophysiological changes that are still not fully understood in the context of healthy sleep or insomnia. Indications of cortical and autonomic physiologic upregulation and cognitive and affective hyperarousal around sleep onset are frequently observed in individuals with insomnia [7]. For example, studies have reported elevated cognitive activity (worries, intrusive thoughts, rumination, and elevated anxiety) [8], pre-sleep cortical (elevated high-frequency EEG activity), metabolic (elevated brain and whole-body metabolism), and autonomic (high heart rate, blood pressure, and cardiac sympathetic tone) activity [9] in people with insomnia compared with good sleeper controls. However, not all studies have replicated these findings, with some instead reporting overwhelming interindividual differences in physiological measures of hyperarousal [10]. Interventional studies suggest that modifying the bedtime state can alter subsequent sleep quality: Acute pre-sleep psychophysiological stress delays sleep onset and worsens sleep quality both in healthy sleepers and insomnia sufferers [11]. On the other hand, bedtime relaxation protocols appear to facilitate sleep onset and improve sleep [12, 13]. For example, one study found that when people with insomnia were instructed to imagine interesting and engaging, but also pleasant and relaxing situations (imagery distraction), sleep-onset latency was shortened and the discomfort associated with pre-sleep unwanted thoughts was reduced [8]. Emerging mind–body practices (e.g. mindfulness) may also reduce sleep initiation problems by targeting pre-sleep arousal [14]. Within this framework, VR has the potential to serve as a distractor for moving an individual’s mind away from worry, rumination, negative cognition, and anxiety. It is also possible that a relaxing sleep-suitable VR experience could induce a relaxed state, and thus facilitate sleep initiation. In a pilot study of 16 women suffering from insomnia symptoms (most with difficulty falling asleep), we showed that being immersed in a relaxing environment (provided via HD visual/audio immersive technology) while performing slow diaphragmatic breathing across the sleep-onset period, reduced bedtime physiological arousal (a reduction in heart rate of about 5 bpm and increases in total heart rate variability) [13]. The intervention also resulted in fewer nocturnal awakenings, less sleep fragmentation, and lower heart rate during sleep compared with no intervention [13]. While this implementation suggests that VR could be applied to help individuals fall asleep, either directly or via enhancing the relaxing effect of paced breathing, the underlying mechanisms and extent to which VR may directly contribute to these effects are unclear, and further research is required. Across the falling asleep period, an individuals’ biobehavioral dynamics, level and content of consciousness, sensory thresholds, and processing change. We argue that several strategies used to facilitate sleep relate to some form of distraction and the attempt to immerse oneself in an “alternative reality” in the context of relaxation. Reading a book, visual imagery, and listening to relaxing environmental sound all involve some partial movement into an alternate reality. The careful application of VR-related technology has the potential to achieve an even greater movement away from an environment (primary reality) to which poor sleep may have become a conditioned response. One conceptualization of how this might work is to view different technologies as enabling a person to move along a continuum from full awareness of the actual physical reality through increasing levels of immersion and presence to a complete acceptance of a different “virtual” reality, hitherto only available in dreams or perhaps hallucinations (self-generated VR). With the development of appropriate sleep-promoting interfaces, technologies on this continuum could involve augmented reality, augmented virtuality or a full computer-generated VR [15] (Figure 1). Whether it will ever be possible to completely remove our anchor to the physical world when experiencing a simulated VR environment while falling asleep, remains an open question, however, these different forms of virtual technology open up the possibility of different applications in clinical sleep medicine and research to investigate and modulate an individual’s experience and the psychophysiology of the falling asleep process. Figure 1. Open in new tabDownload slide Schematic representation of the experience of falling asleep in subjective (primary, “individual’s bedroom”) and external technology-generated augmented reality, augmented virtuality, and virtual reality. The immersion in technology-generated realities usually involves immersive visual and immersive auditory experiences, or even tactile and olfactory experiences, and thus it could be suitable to accommodate the biobehavioral dynamic of the falling asleep process, in which a person usually transitions from eyes-open to eyes-closed states. A person could begin relaxing by receiving the full virtual reality sensory experience, and switch to a different sensory experience when closing their eyes (e.g. immersive audio only), until they fall asleep. Also, bedtime light exposure can be controlled, and blue light-filtering technology easily integrated, reducing the impact of blue wavelengths on melatonin secretion. Figure 1. Open in new tabDownload slide Schematic representation of the experience of falling asleep in subjective (primary, “individual’s bedroom”) and external technology-generated augmented reality, augmented virtuality, and virtual reality. The immersion in technology-generated realities usually involves immersive visual and immersive auditory experiences, or even tactile and olfactory experiences, and thus it could be suitable to accommodate the biobehavioral dynamic of the falling asleep process, in which a person usually transitions from eyes-open to eyes-closed states. A person could begin relaxing by receiving the full virtual reality sensory experience, and switch to a different sensory experience when closing their eyes (e.g. immersive audio only), until they fall asleep. Also, bedtime light exposure can be controlled, and blue light-filtering technology easily integrated, reducing the impact of blue wavelengths on melatonin secretion. Current work has only scratched the surface in using VR as a potential sleep facilitating tool. Implications of falling asleep in virtual realities require further discussion, and comparisons are needed with the gold-standard treatment of insomnia (cognitive behavioral treatment for insomnia [CBTi]) and with other sleep-promoting bedtime mind–body relaxation techniques (e.g. deep breathing, progressive muscle relaxation, visualization, and guided imagery techniques) to determine mechanisms (e.g. determining whether autonomic, cognitive, and emotional pathways are involved) and efficacy of VR-based interventions. In a direct comparison, CBTi was shown to lead to larger improvements in subjective and objective sleep outcomes compared with progressive muscle relaxation training [16]. A recent exhaustive metanalysis of randomized controlled trials concluded that full CBTi as well as its individual components, including relaxation techniques have large effect sizes (Hedges g > 0.56) for the treatment of insomnia, with relaxation alone working particularly well for shortening sleep-onset latency [17]. Authors recommend that more high-quality research is needed to compare different treatment components with one another [17], a recommendation that should also be applied to any emerging VR-based sleep interventions. In particular, future work should carefully examine efficacy of individual components of any VR-accentuated relaxation procedures to determine whether the VR component does indeed add value. A possibility that should also be explored is that the immersive experience of VR could improve the engagement of an individual with other techniques, thus enhancing their efficacy. Further work is also needed from a technology standpoint if VR is to be applied as a sleep-facilitating tool accessible to patients in different communities and regardless of socioeconomic status. The current state of VR technologies poses several challenges for adoption, including its high cost, which limits accessibility and scalability, and questions around the form factor of VR headsets, which are not yet suitable for sleep. If these barriers can be addressed, further investigations into VR applications in the sleep field (e.g. insomnia treatment, pre- or mid-sleep VR exposure-based therapy for PTSD) could open new lines of research and future possibilities for the treatment of sleep disorders. Funding This study was supported by the National Heart, Lung and Blood Institute (NHLBI) grant R01 HL139652 (M.dZ.). The content is solely the responsibility of the authors and does not necessarily represent the official views the National Institutes of Health. Conflict of interest statement. The authors declared no conflict of interest related to the current work. M.dZ., F.C.B., and I.M.C. have received research funding unrelated to this work from Ebb Therapeutics Inc., Fitbit Inc., International Flavors & Fragrances Inc., and Noctrix Health, Inc. I.M.C. is a member of the board of Forest Devices. References 1. Freeman D , et al. Virtual reality in the assessment, understanding, and treatment of mental health disorders . Psychol Med. 2017 ; 47 ( 14 ): 2393 – 2400 . Google Scholar Crossref Search ADS PubMed WorldCat 2. Rus-Calafell M , et al. Virtual reality in the assessment and treatment of psychosis: a systematic review of its utility, acceptability and effectiveness . Psychol Med. 2018 ; 48 ( 3 ): 362 – 391 . Google Scholar Crossref Search ADS PubMed WorldCat 3. Pillai JS , et al. Achieving presence through evoked reality . Front Psychol. 2013 ; 4 : 86 . Google Scholar Crossref Search ADS PubMed WorldCat 4. Diemer J , et al. The impact of perception and presence on emotional reactions: a review of research in virtual reality . Front Psychol. 2015 ; 6 : 26 . Google Scholar Crossref Search ADS PubMed WorldCat 5. Riva G , et al. Affective interactions using virtual reality: the link between presence and emotions . Cyberpsychol Behav. 2007 ; 10 ( 1 ): 45 – 56 . Google Scholar Crossref Search ADS PubMed WorldCat 6. McNamara P , et al. Virtual reality-enabled treatment of nightmares . Dreaming . 2018 ; 28 ( 3 ): 205 – 224 . Google Scholar Crossref Search ADS WorldCat 7. Levenson JC , et al. The pathophysiology of insomnia . Chest. 2015 ; 147 ( 4 ): 1179 – 1192 . Google Scholar Crossref Search ADS PubMed WorldCat 8. Harvey AG , et al. The management of unwanted pre-sleep thoughts in insomnia: distraction with imagery versus general distraction . Behav Res Ther. 2002 ; 40 ( 3 ): 267 – 277 . Google Scholar Crossref Search ADS PubMed WorldCat 9. Bonnet MH , et al. Hyperarousal and insomnia: state of the science . Sleep Med Rev. 2010 ; 14 ( 1 ): 9 – 15 . Google Scholar Crossref Search ADS PubMed WorldCat 10. Varkevisser M , et al. Physiologic indexes in chronic insomnia during a constant routine: evidence for general hyperarousal? Sleep. 2005 ; 28 ( 12 ): 1588 – 1596 . Google Scholar PubMed OpenURL Placeholder Text WorldCat 11. de Zambotti M , et al. Acute stress alters autonomic modulation during sleep in women approaching menopause . Psychoneuroendocrinology. 2016 ; 66 : 1 – 10 . Google Scholar Crossref Search ADS PubMed WorldCat 12. De Niet GJ , et al. Review of systematic reviews about the efficacy of non-pharmacological interventions to improve sleep quality in insomnia . Int J Evid Based Healthc . 2009 ; 7 ( 4 ): 233 – 242 . Google Scholar Crossref Search ADS PubMed WorldCat 13. de Zambotti M , et al. Reducing bedtime physiological arousal levels using immersive audio-visual respiratory bio-feedback: a pilot study in women with insomnia symptoms . J Behav Med. 2019 ; 42 ( 5 ): 973 – 983 . Google Scholar Crossref Search ADS PubMed WorldCat 14. Blake M , et al. The SENSE study: post intervention effects of a randomized controlled trial of a cognitive-behavioral and mindfulness-based group sleep improvement intervention among at-risk adolescents . J Consult Clin Psychol. 2016 ; 84 ( 12 ): 1039 – 1051 . Google Scholar Crossref Search ADS PubMed WorldCat 15. Milgram P , et al. Augmented reality: a class of displays on the reality-virtuality continuum. In proceedings from the SPIE 2351, Telemanipulator and Telepresence Technologies, Photonics for Industrial Applications, December 21, 1995; 1994; Boston, MA . https://doi.org/10.1117/12.197321 16. Edinger JD , et al. Cognitive behavioral therapy for treatment of chronic primary insomnia: a randomized controlled trial . JAMA. 2001 ; 285 ( 14 ): 1856 – 1864 . Google Scholar Crossref Search ADS PubMed WorldCat 17. van Straten A , et al. Cognitive and behavioral therapies in the treatment of insomnia: a meta-analysis . Sleep Med Rev. 2018 ; 38 : 3 – 16 . Google Scholar Crossref Search ADS PubMed WorldCat © Sleep Research Society 2020. Published by Oxford University Press on behalf of the Sleep Research Society. All rights reserved. For permissions, please e-mail [email protected]. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)

Harskamp-van Ginkel, Margreet W; Ierodiakonou, Despo; Margetaki, Katerina; Vafeiadi, Marina; Karachaliou, Marianna; Kogevinas, Manolis; Vrijkotte, Tanja G M; Chatzi, Leda

doi: 10.1093/sleep/zsaa110pmid: 32496519

Abstract Study Objectives The objective of this study was to evaluate the association between gestational sleep deprivation and childhood adiposity and cardiometabolic profile. Methods Data were used from two population-based birth cohorts (Rhea study and Amsterdam Born Children and their Development study). A total of 3,608 pregnant women and their children were followed up until the age of 11 years. Gestational sleep deprivation was defined as 6 or fewer hours of sleep per day, reported by questionnaire. The primary outcomes included repeated measures of body mass index (BMI), waist circumference, body fat, serum lipids, systolic and diastolic blood pressure (DBP) levels in childhood. We performed a pooled analysis with adjusted linear mixed effect and Cox proportional hazards models. We tested for mediation by birthweight, gestational age, and gestational diabetes. Results Gestational sleep deprivation was associated with higher BMI (beta; 95% CI: 0.7; 0.4, 1.0 kg/m2) and waist circumference (beta; 95% CI: 0.9; 0.1, 1.6 cm) in childhood, and increased risk for overweight or obesity (HR; 95% CI: 1.4; 1.1, 2.0). Gestational sleep deprivation was also associated with higher offspring DBP (beta; 95% CI: 1.6; 0.5, 2.7 mmHg). The observed associations were modified by sex (all p-values for interaction < 0.05); and were more pronounced in girls. Gestational diabetes and shorter gestational age partly mediated the seen associations. Conclusions This is the first study showing that gestational sleep deprivation may increase offspring’s adiposity and blood pressure, while exploring possible mechanisms. Attention to glucose metabolism and preterm birth might be extra warranted in mothers with gestational sleep deprivation. gestational sleep deprivation, DOHaD, child, BMI, obesity, blood pressure, adiposity Statement of Significance A suboptimal intrauterine environment is now a recognized risk factor to overweight/obesity and higher blood pressure during later life during later life. The vast majority of pregnant women experience significant sleep disruption. However, whether gestational sleep deprivation affects offspring adiposity and blood pressure in childhood remains unclear. This is the first study showing that gestational sleep deprivation may increase offspring’s adiposity and blood pressure. By exploring possible mechanisms with formal mediation analysis, we recognize that attention to glucose metabolism and preterm birth might be extra warranted in mothers with gestational sleep deprivation. Besides sleep duration, future studies should also investigate the role of sleep quality during pregnancy. Introduction A suboptimal intrauterine environment is now a recognized risk factor to overweight/obesity and higher blood pressure during later life [1, 2]. Pregnancy is a period when lifestyle interventions are encouraged, and parents are aware of their choices. Current interventions are mainly focused on maternal physical activity and/or a healthful diet, and appear effective in decreasing gestational weight gain and diabetes, with some evidence for positive maternal and child outcomes [3–6]. During pregnancy, the great majority of women experience significant sleep disorders including increased rates of inadequate sleep [7, 8]. Sleep disorders in pregnancy have been associated with increased gestational weight gain and pregnancy complications such as hypertension, pre-eclampsia, and gestational diabetes mellitus [9, 10], as well as with adverse perinatal outcomes including intrauterine growth restriction, low birthweight and preterm birth [11–16], longer labor, more pain during labor, and cesarean sections [12, 16]. Some of these factors, such as gestational diabetes, pre-term delivery, and birthweight [11–16], have also been associated with child’s risk of overweight/obesity and cardiometabolic status [17–21], suggesting a plausible link between the two. Yet there is a lack of human studies linking gestational sleep disruption with child’s cardiometabolic health or exploring potential mediating pathways. Current evidence to support the hypothesis that sleep disorders during pregnancy has long-term cardiometabolic effects on offspring comes solely from mice studies. Sex dimorphism has been found in a mice study on metabolic dysfunction due to late gestational sleep fragmentation; male offspring had higher food intake, body weight, visceral fat mass, and insulin resistance and lower adiponectin levels, but not female offspring. Dyslipidemia was apparent in both male and female offspring after gestational sleep deprivation [22]. Two other mice studies found that gestational sleep deprivation increases blood pressure in offspring via alterations in cardiovascular autonomic regulation and renal morphofunctional changes [23, 24]. The effects of gestational sleep deprivation were similar between male and female mice, but in females, the effects were bigger in mice that underwent an ovariectomy and lacked female hormones. In epidemiologic studies, poorer sleep in children has been associated with metabolic risk, adiposity, and altered lipid profile [25–30], and these effects in children have been more prominent in girls compared with boys [25, 31, 32]. As far as we know, there is no published human-based research on the role of sleep during pregnancy on childhood obesity and metabolic health. Our aim was to evaluate the association between gestational sleep deprivation and childhood adiposity and cardiometabolic profile in a pooled analysis of mother–child pairs from two European birth cohorts, with attention to possible interaction by sex and plausible factors mediating these associations. Methods Study population This study utilized data from two European birth cohorts, the Greek “Rhea” birth cohort [33] (n = 1,363) and the Dutch Amsterdam Born Children and their Development (ABCD) study [34] (n=12,379). Both studies are population-based birth cohorts that started during pregnancy. Children from the Rhea cohort were examined at ages 4 (n = 879) and 6 (n = 606) years, while children from the ABCD study were examined at ages 5 (n = 3,260), 10 (n = 2,162), and 11 years (n = 935). Gestational sleep deprivation Information on sleeping habits of the participating mothers of the Rhea cohort was collected through a computer-assisted interview in the third trimester of pregnancy (median (25th–75th) gestational week: 32 (31–35) week) [13]. Sleep duration was obtained by the following close-ended question: “During the past month, how many hours did you sleep per day?” The mother reported sleep duration as 5 or fewer hours, 6–7 h, 8–9 h, and 10 or more hours [13]. Sleep deprivation was defined as five or fewer hours of sleep. Information on gestational sleep duration was available in 685 children with available outcome data at age 4 years and in 436 children with data available at age 6 years. Pregnant women in the ABCD-study received a written questionnaire (median: [25th–75th] gestational week: 16 [14–18] week) and were asked an open-ended question: “How many hours did you sleep or rest lying down per day (of 24 h) on average in the past week.” Sleep deprivation was defined as 6 or fewer hours of sleep, compared with 5 for Rhea, in order to account for the extra daytime resting hours that were reported. Information on gestational sleep duration during pregnancy was available in 3,191, 2,112, and 917 children with available outcome data at age 5, 10, and 11 years old, respectively. Gestational sleep deprivation was used as a binary variable to assess the associations of extremely short gestational sleep with the outcomes of interest instead of sleep duration differences in hours. The cutoff was set at 5 hours of sleep for Rhea and at 6 hours for ABCD due to differences in the sleep questionnaires administered in the two cohorts. We decided on this as extremely short sleep is generally considered as unhealthy, whereas sleep duration needs may vary from person to person and differ across cultures. However, as sensitivity analysis we also used two additional cutoffs at 5 and 7 h of sleep in both cohorts. Child outcome measurements Details of child anthropometry, blood pressure, and serum lipids outcomes are given in the online supplement. In summary, children’s weight and height, waist circumferences, percentage of body fat, diastolic (DBP) and systolic (SBP) blood pressure, and lipid profile were measured in the two cohorts at health clinic visits and/or planned follow-up study assessments. For both cohorts we defined overweight using the same procedure. First, we calculated BMI (weight/height2) [35] and then categorized children into normal, overweight, or obese according to the cutoff points for sex and age proposed by the International Obesity Task Force (IOTF) definitions [36]. As a sensitivity analysis we also used age and gender specific z-scores for the outcomes BMI and blood pressure. Serum lipids included: fasting plasma, total cholesterol, triglycerides, high-density lipoprotein cholesterol (HDL-C), and low-density lipoprotein cholesterol (LDL-C). Statistical analysis We conducted descriptive analysis using standard univariate statistic procedures (chi-square, t-test). We compared mother–child pairs with normal sleep duration or gestational sleep deprivation for baseline characteristics between cohorts. Additionally, we compared the mother–child pairs with follow-up and complete data on covariates (participants) to mothers that participated in the study during pregnancy but missed information on one or more covariates on follow-up after pregnancy (non-participants) per cohort. Our main analysis was a pooled analysis of the Rhea and ABCD cohorts. For continuous outcomes, we used linear mixed models and for overweight/obesity, we used Cox proportional hazards models. Linear mixed models included random effects for cohort and child and a random slope for child age. Mixed models also included an interaction term between the exposure and child age at examination. Child age at examination in the interaction term was used categorically (4, 5, 6, 10, and 11 years in the models for BMI; 4, 5, 6, and 11 years in the models for all other outcomes). The overall effect of the exposure was evaluated using the marginal effects and the difference between the two groups was tested using Wald’s test. The associations are reported in terms of beta coefficients and their corresponding 95% CIs. In Cox proportional hazards models, shared frailties for cohort were introduced in order to account for the shared risk within each cohort and hazard ratios (HR) and 95% CIs were reported. Birth was considered as the time of study entry and age at study visit as the time scale in our analysis. The exact age at the visit during which the child first became overweight/obese was used as the time of event. Children who did not become overweight/obese during follow-up were censored at the end of study follow-up or when lost to follow-up. The proportional hazards assumption was tested using both graphical inspection methods and Schoenfeld residuals. We constructed a directed acyclic graph (DAG) based on previous knowledge and selected the set of confounders using DAGitty version 3.0 (Supplementary Figure S1) [37]. The confounders included in all models were maternal age at conception, parity (nulliparous and multiparous), maternal smoking early in pregnancy (yes/no), pre-pregnancy BMI (normal weight/overweight/obese), maternal education (low/middle/high), and maternal origin (country of cohort/other). Child sex and age at assessment (years) were also included. Models with blood pressure as an outcome were further adjusted for child height and BMI and models with lipids as an outcome were further adjusted for child BMI. In order to evaluate potential effect modification by sex, we included a multiplicative exposure-sex interaction term in each model. As a sensitivity analysis, we performed a random effects meta-analysis. For this, we obtained cohort specific estimates using mixed effects models with child random effects and a random slope for child age for continuous outcomes and Cox proportional hazards models for binary outcomes. Consequently, we combined these estimates using random effects meta-analysis, in order to check the consistency with the pooled analysis and for quantifying heterogeneity among included studies with chi-square test from Cochran’s Q and I2 statistics. We tested if there was significant mediation by three plausible mediators (gestational diabetes; gestational age; and birthweight) on the association between gestational sleep deprivation and childhood BMI and other outcomes (Figure 1). We made two separate mediator models with structural equation modeling (SEM); the first one with the continuous mediators gestational age and birthweight as parallel mediators; the second one with gestational diabetes as a single binary mediator. The second model included only mother–child pairs from the ABCD study, as gestational sleep deprivation was measured in early pregnancy, before the diagnosis of gestational diabetes would be made. The a-path of a mediator reports the association between gestational sleep deprivation and the mediator and the b-path reports the association between the mediator and offspring BMI at age four to five years. The indirect effect is a product of the a- and b-path. A 95 percentile bootstrap CI was calculated based on 1,000 bootstrap resamples for the indirect effect (ab), in order to test for significance. The total indirect effect is a sum of both indirect effects in a parallel model. The total direct effect (c′-path) refers to the association between gestational sleep deprivation and offspring BMI, corrected for the b-path. The total effect (c-path) is the association between gestational sleep deprivation and offspring BMI. The following confounders were added to the simple adjustment model: child sex, age at assessment (years). Considering the small numbers in each group, we did not perform a mediation analysis with full adjustment for all confounders. Figure 1. Open in new tabDownload slide Mediation models. Theoretical model for mediation: the a-path reports the change in gestational age or birthweight (continuous mediators) or the odds ratio for gestational diabetes (binary mediator) if the mother had gestational sleep deprivation; and the b-path reports the increase in offspring BMI at age 4–5 years associated with a one point increase of each mediator. The c′ path reports the direct effect of gestational sleep deprivation on offspring BMI. Figure 1. Open in new tabDownload slide Mediation models. Theoretical model for mediation: the a-path reports the change in gestational age or birthweight (continuous mediators) or the odds ratio for gestational diabetes (binary mediator) if the mother had gestational sleep deprivation; and the b-path reports the increase in offspring BMI at age 4–5 years associated with a one point increase of each mediator. The c′ path reports the direct effect of gestational sleep deprivation on offspring BMI. All analyses were conducted using Stata version 13 and 15 and significance level for all two-sided tests was set at the 5% level. We used capture program for mediation analysis. Results Participant characteristics In the present analyses complete data on exposure, outcome, and covariates were available in a total of 661 and 453 Rhea mother–child pairs at ages 4 and 6 years, respectively and in a total of 2,947, 1,957, and 874 ABCD mother–child pairs at ages 5, 10, and 11 years, respectively. Table 1 shows maternal and infant characteristics. In total, 144 (4.0%) mothers were sleep deprived during the index pregnancy (5.6% in Rhea and 3.6% in ABCD). Cardiometabolic characteristics of the children are presented in Supplementary Table S1. In the Rhea-cohort, 21.4% of the children was overweight at age 6 years and 11.0% was obese; whereas in the ABCD-cohort 7.1% of the children was overweight at age 5 years and 1.5% was obese. In the ABCD, sociodemographic characteristics of the mothers with gestational sleep deprivation varied significantly from mothers with adequate gestational sleep; they had higher rates of gestational diabetes (6.5% vs. 1.5%); and children were born at lower gestational age (39.1 vs. 39.5 weeks) and with a lower birthweight (3,364 vs. 3,477 g). Besides parity, we did not see these differences in the Rhea cohort (Supplementary Table S2). Nonresponse analysis revealed that participants were of higher education and lower BMI-pregnancy in both cohorts compared to lost to follow-up mother–child pairs (Supplementary Table S3). Table 1. Maternal and infant characteristics . Overall, N = 3,608 . Rhea, N = 661 . ABCD, N = 2,947 . . . No. (%) or mean ± SD . No. (%) or mean ± SD . No. (%) or mean ± SD . P-value* . Maternal characteristics  Maternal age at conception (years) 31.6 (4.6) 29.8 (4.9) 32.0 (4.5) <0.001  Maternal education <0.001  Low 181 (5.0) 103 (15.6) 78 (2.6)  Middle 805 (22.3) 337 (51.0) 468 (15.9)  High 2,622 (72.7) 221 (33.4) 2,401 (81.5)  Maternal origin-non native 717 (19.9) 31 (4.7) 686 (23.3) <0.001  Parity-Nulliparous 1,965 (54.5) 305 (46.1) 1,660 (56.3) <0.001  Smoking in early pregnancy-yes 487 (13.5) 223 (33.7) 264 (9.0) <0.001  Pre-pregnancy BMI (kg/m2) 23.3 (4.1) 24.7 (4.8) 22.9 (3.8) <0.001  Underweight and normal weight (BMI < 25 kg/m2) 2,764 (76.6) 426 (64.4) 2,338 (79.3) <0.001  Overweight (BMI 25–30 kg/m2) 606 (16.8) 147 (22.2) 459 (15.6)  Obese (BMI ≥ 30 kg/m2) 238 (6.6) 88 (13.3) 150 (5.1) Gestational characteristics  Gestational diabetes 110 (3.1) 60 (9.2) 50 (1.7) <0.001  Gestational age at delivery (weeks) 39.2 (1.8) 38.2 (1.5) 39.5 (1.7) <0.001  Cesarean section 739 (20.5) 330 (49.9) 409 (13.9) <0.001  Gestational sleep deprivation 144 (4.0) 37 (5.6) 107 (3.6) 0.020 Infant characteristics  Female 1,785 (49.5) 309 (46.7) 1,476 (50.1) 0.121  Birth weight (g) 3,424.7 (542.2) 3,208.8 (448.6) 3,473.1 (549.6) <0.001  Ever breastfed 2,929 (82.0) 558 (86.9) 2,371 (80.9) <0.001 . Overall, N = 3,608 . Rhea, N = 661 . ABCD, N = 2,947 . . . No. (%) or mean ± SD . No. (%) or mean ± SD . No. (%) or mean ± SD . P-value* . Maternal characteristics  Maternal age at conception (years) 31.6 (4.6) 29.8 (4.9) 32.0 (4.5) <0.001  Maternal education <0.001  Low 181 (5.0) 103 (15.6) 78 (2.6)  Middle 805 (22.3) 337 (51.0) 468 (15.9)  High 2,622 (72.7) 221 (33.4) 2,401 (81.5)  Maternal origin-non native 717 (19.9) 31 (4.7) 686 (23.3) <0.001  Parity-Nulliparous 1,965 (54.5) 305 (46.1) 1,660 (56.3) <0.001  Smoking in early pregnancy-yes 487 (13.5) 223 (33.7) 264 (9.0) <0.001  Pre-pregnancy BMI (kg/m2) 23.3 (4.1) 24.7 (4.8) 22.9 (3.8) <0.001  Underweight and normal weight (BMI < 25 kg/m2) 2,764 (76.6) 426 (64.4) 2,338 (79.3) <0.001  Overweight (BMI 25–30 kg/m2) 606 (16.8) 147 (22.2) 459 (15.6)  Obese (BMI ≥ 30 kg/m2) 238 (6.6) 88 (13.3) 150 (5.1) Gestational characteristics  Gestational diabetes 110 (3.1) 60 (9.2) 50 (1.7) <0.001  Gestational age at delivery (weeks) 39.2 (1.8) 38.2 (1.5) 39.5 (1.7) <0.001  Cesarean section 739 (20.5) 330 (49.9) 409 (13.9) <0.001  Gestational sleep deprivation 144 (4.0) 37 (5.6) 107 (3.6) 0.020 Infant characteristics  Female 1,785 (49.5) 309 (46.7) 1,476 (50.1) 0.121  Birth weight (g) 3,424.7 (542.2) 3,208.8 (448.6) 3,473.1 (549.6) <0.001  Ever breastfed 2,929 (82.0) 558 (86.9) 2,371 (80.9) <0.001 *Univariate analysis with chi-square or t-test. BMI, body mass index. Open in new tab Table 1. Maternal and infant characteristics . Overall, N = 3,608 . Rhea, N = 661 . ABCD, N = 2,947 . . . No. (%) or mean ± SD . No. (%) or mean ± SD . No. (%) or mean ± SD . P-value* . Maternal characteristics  Maternal age at conception (years) 31.6 (4.6) 29.8 (4.9) 32.0 (4.5) <0.001  Maternal education <0.001  Low 181 (5.0) 103 (15.6) 78 (2.6)  Middle 805 (22.3) 337 (51.0) 468 (15.9)  High 2,622 (72.7) 221 (33.4) 2,401 (81.5)  Maternal origin-non native 717 (19.9) 31 (4.7) 686 (23.3) <0.001  Parity-Nulliparous 1,965 (54.5) 305 (46.1) 1,660 (56.3) <0.001  Smoking in early pregnancy-yes 487 (13.5) 223 (33.7) 264 (9.0) <0.001  Pre-pregnancy BMI (kg/m2) 23.3 (4.1) 24.7 (4.8) 22.9 (3.8) <0.001  Underweight and normal weight (BMI < 25 kg/m2) 2,764 (76.6) 426 (64.4) 2,338 (79.3) <0.001  Overweight (BMI 25–30 kg/m2) 606 (16.8) 147 (22.2) 459 (15.6)  Obese (BMI ≥ 30 kg/m2) 238 (6.6) 88 (13.3) 150 (5.1) Gestational characteristics  Gestational diabetes 110 (3.1) 60 (9.2) 50 (1.7) <0.001  Gestational age at delivery (weeks) 39.2 (1.8) 38.2 (1.5) 39.5 (1.7) <0.001  Cesarean section 739 (20.5) 330 (49.9) 409 (13.9) <0.001  Gestational sleep deprivation 144 (4.0) 37 (5.6) 107 (3.6) 0.020 Infant characteristics  Female 1,785 (49.5) 309 (46.7) 1,476 (50.1) 0.121  Birth weight (g) 3,424.7 (542.2) 3,208.8 (448.6) 3,473.1 (549.6) <0.001  Ever breastfed 2,929 (82.0) 558 (86.9) 2,371 (80.9) <0.001 . Overall, N = 3,608 . Rhea, N = 661 . ABCD, N = 2,947 . . . No. (%) or mean ± SD . No. (%) or mean ± SD . No. (%) or mean ± SD . P-value* . Maternal characteristics  Maternal age at conception (years) 31.6 (4.6) 29.8 (4.9) 32.0 (4.5) <0.001  Maternal education <0.001  Low 181 (5.0) 103 (15.6) 78 (2.6)  Middle 805 (22.3) 337 (51.0) 468 (15.9)  High 2,622 (72.7) 221 (33.4) 2,401 (81.5)  Maternal origin-non native 717 (19.9) 31 (4.7) 686 (23.3) <0.001  Parity-Nulliparous 1,965 (54.5) 305 (46.1) 1,660 (56.3) <0.001  Smoking in early pregnancy-yes 487 (13.5) 223 (33.7) 264 (9.0) <0.001  Pre-pregnancy BMI (kg/m2) 23.3 (4.1) 24.7 (4.8) 22.9 (3.8) <0.001  Underweight and normal weight (BMI < 25 kg/m2) 2,764 (76.6) 426 (64.4) 2,338 (79.3) <0.001  Overweight (BMI 25–30 kg/m2) 606 (16.8) 147 (22.2) 459 (15.6)  Obese (BMI ≥ 30 kg/m2) 238 (6.6) 88 (13.3) 150 (5.1) Gestational characteristics  Gestational diabetes 110 (3.1) 60 (9.2) 50 (1.7) <0.001  Gestational age at delivery (weeks) 39.2 (1.8) 38.2 (1.5) 39.5 (1.7) <0.001  Cesarean section 739 (20.5) 330 (49.9) 409 (13.9) <0.001  Gestational sleep deprivation 144 (4.0) 37 (5.6) 107 (3.6) 0.020 Infant characteristics  Female 1,785 (49.5) 309 (46.7) 1,476 (50.1) 0.121  Birth weight (g) 3,424.7 (542.2) 3,208.8 (448.6) 3,473.1 (549.6) <0.001  Ever breastfed 2,929 (82.0) 558 (86.9) 2,371 (80.9) <0.001 *Univariate analysis with chi-square or t-test. BMI, body mass index. Open in new tab Gestational sleep deprivation and childhood cardiometabolic health Table 2 shows the association of gestational sleep deprivation with child BMI, waist circumference, body fat, blood pressure, and the risk of overweight/obesity after adjusting for covariates. Gestational sleep deprivation was associated with higher child BMI (beta 0.7 kg/m2 [95% CI: 0.4, 1.0]), waist circumference (beta 0.9 cm [95% CI: 0.1, 1.6]) and DBP (beta 1.6 mmHg [95% CI: 0.5, 2.7]) but not with per cent body fat (beta 0.7% [95% CI: −0.3, 1.7]). Children born to mothers with sleep deprivation in pregnancy had 40% increased risk of overweight and obesity (HR 1.4 [95% CI:1.1, 2.0]). There were no significant associations with child lipid profile (Supplementary Table S4). Table 2. A longitudinal pooled analysis of associations between gestational sleep deprivation and adiposity and blood pressure during childhood adjusting for potential confounders and testing for sex interaction . Rhea and ABCD (n = 3,608) . . . Overall . Boys . Girls . . Outcomes . N . Estimate (95% CI) . P-value . N . Estimate (95% CI) . P-value . N . Estimate (95% CI) . P-value . P-interaction with sex . BMI (kg/m2)§,|| 3,607 0.7 (0.4, 1.0) <0.001 1,823 0.5 (0.0, 1.0) 0.050 1,784 0.9 (0.4, 1.3) <0.001 0.046 Overweight or obese‡,* 3,607 1.4 (1.1, 2.0) 0.019 1,823 0.9 (0.5, 1.5) 0.663 1,784 2.2 (1.5, 3.3) <0.001 0.004 Waist circ. (cm)*,|| 3,601 0.9 (0.1, 1.6) 0.031 1,818 0.4 (−0.8, 1.6) 0.498 1,783 1.3 (0.2, 2.3) 0.018 0.167 Body fat (%)*,|| 3,590 0.7 (−0.3, 1.7) 0.164 1,816 0.7 (−0.7, 2.1) 0.332 1,774 0.7 (−0.7,2.1) 0.319 0.957 SBP (mmHg)* ,†,|| 3,491 0.5 (−0.8, 1.8) 0.416 1,758 −0.4 (−2.2, 1.5) 0.687 1,733 1.8 (−0.1, 3.6) 0.063 0.135 DBP (mmHg)* ,†,|| 3,485 1.6 (0.5, 2.7) 0.006 1,753 0.3 (−1.3, 1.9) 0.726 1,732 2.8 (1.2, 4.3) 0.001 0.045 . Rhea and ABCD (n = 3,608) . . . Overall . Boys . Girls . . Outcomes . N . Estimate (95% CI) . P-value . N . Estimate (95% CI) . P-value . N . Estimate (95% CI) . P-value . P-interaction with sex . BMI (kg/m2)§,|| 3,607 0.7 (0.4, 1.0) <0.001 1,823 0.5 (0.0, 1.0) 0.050 1,784 0.9 (0.4, 1.3) <0.001 0.046 Overweight or obese‡,* 3,607 1.4 (1.1, 2.0) 0.019 1,823 0.9 (0.5, 1.5) 0.663 1,784 2.2 (1.5, 3.3) <0.001 0.004 Waist circ. (cm)*,|| 3,601 0.9 (0.1, 1.6) 0.031 1,818 0.4 (−0.8, 1.6) 0.498 1,783 1.3 (0.2, 2.3) 0.018 0.167 Body fat (%)*,|| 3,590 0.7 (−0.3, 1.7) 0.164 1,816 0.7 (−0.7, 2.1) 0.332 1,774 0.7 (−0.7,2.1) 0.319 0.957 SBP (mmHg)* ,†,|| 3,491 0.5 (−0.8, 1.8) 0.416 1,758 −0.4 (−2.2, 1.5) 0.687 1,733 1.8 (−0.1, 3.6) 0.063 0.135 DBP (mmHg)* ,†,|| 3,485 1.6 (0.5, 2.7) 0.006 1,753 0.3 (−1.3, 1.9) 0.726 1,732 2.8 (1.2, 4.3) 0.001 0.045 Gestational sleep deprivation was defined as at least 5 and 6 h for Rhea and ABCD cohort, respectively. All models are adjusted for child sex, age at assessment (years), parity (nulliparous and multiparous), maternal smoking early in pregnancy (yes/no) maternal age at conception, pre-pregnancy BMI (normal weight/overweight/obese), maternal origin (country of cohort/other), and maternal education (low/middle/high). Bold-faced text indicated significant associations (p-value < 0.05). *Point for sex and age that was proposed by the IOTF. †Additionally adjusted for child height and BMI at assessment. ‡Hazard ratios and 95% CIs obtained by Cox proportional hazard models with shared cohort frailties. §Defined with use of the BMI cutoff point for sex and age that was proposed by the IOTF. ||Beta coefficient and 95% CIs as marginal effect estimates obtained by mixed effects models with cohort and child random effect and age interaction. BMI, body mass index; BP, blood pressure. Open in new tab Table 2. A longitudinal pooled analysis of associations between gestational sleep deprivation and adiposity and blood pressure during childhood adjusting for potential confounders and testing for sex interaction . Rhea and ABCD (n = 3,608) . . . Overall . Boys . Girls . . Outcomes . N . Estimate (95% CI) . P-value . N . Estimate (95% CI) . P-value . N . Estimate (95% CI) . P-value . P-interaction with sex . BMI (kg/m2)§,|| 3,607 0.7 (0.4, 1.0) <0.001 1,823 0.5 (0.0, 1.0) 0.050 1,784 0.9 (0.4, 1.3) <0.001 0.046 Overweight or obese‡,* 3,607 1.4 (1.1, 2.0) 0.019 1,823 0.9 (0.5, 1.5) 0.663 1,784 2.2 (1.5, 3.3) <0.001 0.004 Waist circ. (cm)*,|| 3,601 0.9 (0.1, 1.6) 0.031 1,818 0.4 (−0.8, 1.6) 0.498 1,783 1.3 (0.2, 2.3) 0.018 0.167 Body fat (%)*,|| 3,590 0.7 (−0.3, 1.7) 0.164 1,816 0.7 (−0.7, 2.1) 0.332 1,774 0.7 (−0.7,2.1) 0.319 0.957 SBP (mmHg)* ,†,|| 3,491 0.5 (−0.8, 1.8) 0.416 1,758 −0.4 (−2.2, 1.5) 0.687 1,733 1.8 (−0.1, 3.6) 0.063 0.135 DBP (mmHg)* ,†,|| 3,485 1.6 (0.5, 2.7) 0.006 1,753 0.3 (−1.3, 1.9) 0.726 1,732 2.8 (1.2, 4.3) 0.001 0.045 . Rhea and ABCD (n = 3,608) . . . Overall . Boys . Girls . . Outcomes . N . Estimate (95% CI) . P-value . N . Estimate (95% CI) . P-value . N . Estimate (95% CI) . P-value . P-interaction with sex . BMI (kg/m2)§,|| 3,607 0.7 (0.4, 1.0) <0.001 1,823 0.5 (0.0, 1.0) 0.050 1,784 0.9 (0.4, 1.3) <0.001 0.046 Overweight or obese‡,* 3,607 1.4 (1.1, 2.0) 0.019 1,823 0.9 (0.5, 1.5) 0.663 1,784 2.2 (1.5, 3.3) <0.001 0.004 Waist circ. (cm)*,|| 3,601 0.9 (0.1, 1.6) 0.031 1,818 0.4 (−0.8, 1.6) 0.498 1,783 1.3 (0.2, 2.3) 0.018 0.167 Body fat (%)*,|| 3,590 0.7 (−0.3, 1.7) 0.164 1,816 0.7 (−0.7, 2.1) 0.332 1,774 0.7 (−0.7,2.1) 0.319 0.957 SBP (mmHg)* ,†,|| 3,491 0.5 (−0.8, 1.8) 0.416 1,758 −0.4 (−2.2, 1.5) 0.687 1,733 1.8 (−0.1, 3.6) 0.063 0.135 DBP (mmHg)* ,†,|| 3,485 1.6 (0.5, 2.7) 0.006 1,753 0.3 (−1.3, 1.9) 0.726 1,732 2.8 (1.2, 4.3) 0.001 0.045 Gestational sleep deprivation was defined as at least 5 and 6 h for Rhea and ABCD cohort, respectively. All models are adjusted for child sex, age at assessment (years), parity (nulliparous and multiparous), maternal smoking early in pregnancy (yes/no) maternal age at conception, pre-pregnancy BMI (normal weight/overweight/obese), maternal origin (country of cohort/other), and maternal education (low/middle/high). Bold-faced text indicated significant associations (p-value < 0.05). *Point for sex and age that was proposed by the IOTF. †Additionally adjusted for child height and BMI at assessment. ‡Hazard ratios and 95% CIs obtained by Cox proportional hazard models with shared cohort frailties. §Defined with use of the BMI cutoff point for sex and age that was proposed by the IOTF. ||Beta coefficient and 95% CIs as marginal effect estimates obtained by mixed effects models with cohort and child random effect and age interaction. BMI, body mass index; BP, blood pressure. Open in new tab There was significant effect modification by sex on the observed associations (p-values for interaction < 0.05; Table 2). When stratified by sex, short sleep duration in pregnancy was significantly associated with higher DBP, BMI and risk for overweight/obesity in girls only, whereas these associations in boys were smaller and not significant. The adverse associations of short maternal sleep with child’s waist circumference and SBP was also stronger in girls compared to boys, however the interactions did not reach statistical significance (Table 2). Sensitivity analysis When using age and sex specific z values for BMI and blood pressure, we also found BMI and DBP to be associated with gestational sleep deprivation in girls (Supplementary Table S5). The second sensitivity analysis showed us that using the same cutoff of ≤5 h of sleep/day for gestational sleep deprivation in both cohorts made the associations stronger and still significant, even with a prevalence of gestational sleep deprivation of 2%. When we used ≤7 h as a cutoff in both cohorts, the prevalence of gestational sleep deprivation was 19% and associations remained significant for overweight/obesity and blood pressure in girls (Supplementary Table S6). The random effects meta-analysis of the cohort specific estimates from the mixed models confirmed the girl-specific associations of short maternal sleep during pregnancy with BMI, waist circumferences, and blood pressure (Figure 2 and Supplementary Table S7). The associations were stronger and only significant in the ABCD-cohort, compared to the Rhea-cohort. There was significant interaction by age for the association with BMI, waist circumference, total cholesterol, and LDL as the effects of gestational sleep deprivation became stronger with age (Supplementary Table S8). The I2 statistic for BMI was suggestive for heterogeneity of the effect in the two studies (I2 = 71.6, p-value = 0.061) but the stratification according to child sex, revealed evidence for heterogeneity among boys (I2 = 71.6, p-value = 0.115) and not among girls (I2 = 0.0%, p-value = 0.323). No heterogeneity was observed for the other outcomes (I2 = 0.0%, p-values < 0.1; Figure 2). Figure 2. Open in new tabDownload slide A random effects meta-analysis of adjusted associations between gestational sleep deprivation and adiposity and blood pressure in childhood. Gestational sleep deprivation was defined as at least 5 and 6 h for Rhea (n = 661) and ABCD cohort (n = 2,947), respectively. Cohort specific estimates were obtained by mixed effects models with child random effects and a random slope for child age. Cohort-specific estimates were adjusted for child sex, age at assessment (years), parity (nulliparous and multiparous), maternal smoking early in pregnancy (yes/no) maternal age at conception, pre-pregnancy BMI (normal weight/overweight/obese), maternal origin (country of cohort/other), and maternal education (low/middle/high). Models for blood pressure were additionally adjusted for child height and BMI at assessment. Figure 2. Open in new tabDownload slide A random effects meta-analysis of adjusted associations between gestational sleep deprivation and adiposity and blood pressure in childhood. Gestational sleep deprivation was defined as at least 5 and 6 h for Rhea (n = 661) and ABCD cohort (n = 2,947), respectively. Cohort specific estimates were obtained by mixed effects models with child random effects and a random slope for child age. Cohort-specific estimates were adjusted for child sex, age at assessment (years), parity (nulliparous and multiparous), maternal smoking early in pregnancy (yes/no) maternal age at conception, pre-pregnancy BMI (normal weight/overweight/obese), maternal origin (country of cohort/other), and maternal education (low/middle/high). Models for blood pressure were additionally adjusted for child height and BMI at assessment. Mediation by gestational diabetes, gestational age, and birthweight Table 3 presents results for the mediation analysis on BMI. The total direct effect (c′ path) was 0.5, meaning that children of mothers with gestational sleep deprivation had a 0.5 kg/m2 higher mean childhood BMI. Gestational diabetes was a significant mediator in the association between gestational sleep deprivation and offspring BMI. Mothers with gestational sleep deprivation during early pregnancy had 4.5 times higher odds of gestational diabetes (a-path), and gestational diabetes was associated with a mean increase of 1.1 kg/m2 in offspring BMI. The confidence interval of the indirect effect was wide, due to small numbers. Gestational age was also a significant mediator in the association between gestational sleep deprivation and offspring BMI, leading to on average a 0.06 point higher BMI. We found that children of mothers with gestational sleep deprivation were born with half a week shorter gestational age (a-path), and that a shorter gestational age was associated with a higher offspring BMI (b-path). Both indirect effects were found significant as the bootstrap confidence interval of the indirect effects did not contain zero, even though the numbers for gestational diabetes were small resulting in a wide confidence interval. Table 3. Mediation by gestational age, birthweight, and gestational diabetes in the association between gestational sleep deprivation and offspring BMI at ages 4–5 years . . Sleep → Mediator (a-path) measure of association . Mediator → BMI (b-path) measure of association . Mediation (a × b) . Single mediator model with binary mediator (ABCD n = 2,947) Gestational diabetes at delivery (yes/no) OR 4.51 β 1.10 5.25 (1.35, 21.28) Total direct effect (c′ path) 0.52 (0.24, 0.80) Multiple paralell mediator model with continuous mediators (ABCD and Rhea, n = 3,607) Gestational age (weeks) β −0.48 β −0.12 0.06 (0.02, 0.10) Birthweight (g) β −77 β 0.001 −0.04 (−0.09, 0.01) Total indirect effect 0.02 (−0.02, 0.057) Total direct effect (c′ path) 0.49 (0.23, 0.75) . . Sleep → Mediator (a-path) measure of association . Mediator → BMI (b-path) measure of association . Mediation (a × b) . Single mediator model with binary mediator (ABCD n = 2,947) Gestational diabetes at delivery (yes/no) OR 4.51 β 1.10 5.25 (1.35, 21.28) Total direct effect (c′ path) 0.52 (0.24, 0.80) Multiple paralell mediator model with continuous mediators (ABCD and Rhea, n = 3,607) Gestational age (weeks) β −0.48 β −0.12 0.06 (0.02, 0.10) Birthweight (g) β −77 β 0.001 −0.04 (−0.09, 0.01) Total indirect effect 0.02 (−0.02, 0.057) Total direct effect (c′ path) 0.49 (0.23, 0.75) Gestational sleep deprivation was defined as at least 5 and 6 h for Rhea and ABCD cohort, respectively. Mediation model based on SEM, adjusted for child sex and age at assessment (years). Bold-faced text indicates significant associations (p-value < 0.05). OR, odds ratio. Open in new tab Table 3. Mediation by gestational age, birthweight, and gestational diabetes in the association between gestational sleep deprivation and offspring BMI at ages 4–5 years . . Sleep → Mediator (a-path) measure of association . Mediator → BMI (b-path) measure of association . Mediation (a × b) . Single mediator model with binary mediator (ABCD n = 2,947) Gestational diabetes at delivery (yes/no) OR 4.51 β 1.10 5.25 (1.35, 21.28) Total direct effect (c′ path) 0.52 (0.24, 0.80) Multiple paralell mediator model with continuous mediators (ABCD and Rhea, n = 3,607) Gestational age (weeks) β −0.48 β −0.12 0.06 (0.02, 0.10) Birthweight (g) β −77 β 0.001 −0.04 (−0.09, 0.01) Total indirect effect 0.02 (−0.02, 0.057) Total direct effect (c′ path) 0.49 (0.23, 0.75) . . Sleep → Mediator (a-path) measure of association . Mediator → BMI (b-path) measure of association . Mediation (a × b) . Single mediator model with binary mediator (ABCD n = 2,947) Gestational diabetes at delivery (yes/no) OR 4.51 β 1.10 5.25 (1.35, 21.28) Total direct effect (c′ path) 0.52 (0.24, 0.80) Multiple paralell mediator model with continuous mediators (ABCD and Rhea, n = 3,607) Gestational age (weeks) β −0.48 β −0.12 0.06 (0.02, 0.10) Birthweight (g) β −77 β 0.001 −0.04 (−0.09, 0.01) Total indirect effect 0.02 (−0.02, 0.057) Total direct effect (c′ path) 0.49 (0.23, 0.75) Gestational sleep deprivation was defined as at least 5 and 6 h for Rhea and ABCD cohort, respectively. Mediation model based on SEM, adjusted for child sex and age at assessment (years). Bold-faced text indicates significant associations (p-value < 0.05). OR, odds ratio. Open in new tab Low birthweight was not a significant mediator. The effect of gestational sleep deprivation on birthweight was not significant (a-path), but a higher birthweight was associated with a higher offspring BMI (b-path). Apart from the BMI outcome, we also tested mediation for the other metabolic outcomes of interest. Gestational diabetes was a mediator for overweight/obesity, waist circumference, and per cent body fat, but not for DBP and SBP. Gestational age was a mediator for overweight/obesity and waist circumference. Low birthweight was not a mediator for the outcomes of interest (Supplementary Table S9). Discussion This is the first human epidemiological study showing that gestational sleep deprivation could be associated with offspring cardiometabolic profile. Children born to mothers with short sleep duration during pregnancy had higher adiposity and blood pressure levels with associations being more pronounced in girls than in boys and the effects becoming stronger with age. The effect estimates for each cohort separately were in the same direction, but stronger and significant in the ABCD cohort. The associations with adiposity were partly mediated by gestational diabetes and shorter gestational age. Both sleep duration and sleep quality are known to change during pregnancy [38]. A recent meta-analysis found that about half of pregnant women experience poor sleep quality and that median sleep quality decreases from the second to third trimester [39]. Studies in the general population, as well as in pregnant women, suggest that sleep disturbances may alter the neuroendocrine homeostasis of the body, with an increased activity of the sympathetic nervous system and hypothalamic–pituitary system, as well as the stress and pro-inflammatory responses which are associated with numerous health consequences [40, 41]. Syntheses of findings from epidemiological studies in general populations suggest that lack of sleep is associated with obesity and a wide range of adverse cardiometabolic outcomes affecting both adults and children [42–45]. Importantly, during pregnancy the adverse physiologic response to sleep deprivation may lead to a suboptimal intrauterine environment, with subsequent effects on the placenta function, direct maternal, and fetal effects, but also with long-term consequences [2, 40]. Gestational sleep disruption has been associated with gestational diabetes, pre-term delivery, and birthweight [11–16], factors also being associated with child’s risk of overweight/obesity and cardiometabolic status [17–21], thus may be involved in the causal pathway. In agreement with that, for the association between gestational sleep deprivation and offspring BMI, overweight, waist circumference, and per cent body fat, we found partly mediation by gestational diabetes. Mothers with gestational sleep deprivation during early pregnancy had higher odds of gestational diabetes during later pregnancy and consequently gestational diabetes was associated with higher offspring BMI. The underlying pathogenic mechanisms behind gestational diabetes and the abnormal metabolic risk profile in offspring are unknown, but epigenetic changes induced by exposure to maternal hyperglycemia during fetal life may be implicated in impaired insulin sensitivity in the offspring [46]. We also found that part of the association between gestational sleep deprivation and offspring adiposity in our cohort was mediated by gestational age; children of mothers with gestational sleep deprivation were born on average half a week earlier, and that was associated with a small increase in offspring BMI. Studies suggest the balance between pro- and anti-inflammatory cytokines may vary in each trimester, and sleep deprivation can adversely affect pro-inflammatory response with endothelial dysfunction in the placenta, which along with impaired glucose metabolism and can lead to preterm labor [14, 47, 48]. This causal pathway is further supported by another cohort study showing that obesity at the age of 2 years among children who were born extremely preterm was associated with perinatal systemic inflammation [49]. We found interaction by sex in our associations, with associations being more pronounced in girls than in boys. A sex-specific effect of poor sleep has also been observed by epidemiological studies in children, where sleep disruption was associated with more prominent effects on metabolic risk, adiposity, and altered lipid profile in girls compared with boys [25, 31, 32]. Also, during pregnancy sexual dimorphisms have been observed in the effects of maternal obesity on childhood growth [17]. A possible mechanism could be differences in placenta function between boys and girls, which are caused by differences in gene expression in response to maternal health [50]. The differences in adaptation between males and females may be context, species and stage specific, and therefore it is difficult to say whether one sex copes better than the other [50]. Our findings in human are not in line with studies in mice, where associations between sleep fragmentation were stronger in male offspring [22] and sleep deprivation had similar associations with blood pressure in both sexes [23, 24]. In a mouse study with female offspring the effects of gestational sleep deprivation were bigger in females that underwent an ovariectomy and lacked female hormones [24]. Future research could investigate if there is still interaction by sex when the children reach adolescence. Strengths and limitations Our study has several strengths. We were able to test longitudinal mediation in a large number of mother–infant pairs from different countries. By doing this, we were able to test potential mechanisms for the association between gestational sleep deprivation and adiposity. In the mediation analysis, the number of mothers with gestational sleep deprivation and gestational diabetes was low, but we still found a significant mediation effect with the minimal adjustment set. However, these results should be interpreted with caution due to the small sample size. Although our data are observational, the sequence of events and associations over time might implicate causal relationships. All data were collected prospectively and outcome measurements during childhood were all performed by research staff. Third, we tested the association in a pooled analysis from two cohorts, but we do also provide cohort specific estimations for the benefit of quantifying the heterogeneity between cohorts and plotting the associations. There were several limitations, mostly inherent to the cohorts’ study design. Our exposure variable of gestational sleep deprivation was composed from a self-administered questionnaire and therefore recall bias and possible under- or overreporting may occur. We measured sleep at two different points during pregnancy, during the third (Rhea) and second (ABCD) trimester, capturing two stages of pregnancy. Effects of sleep duration, as well as sleep duration itself, may vary during pregnancy, and that may, besides other unknown factors, explain the different associations between the two cohorts. Also, the phrasing of the sleep question differed between both cohorts. Therefore, we used different cutoffs in the main analysis, correcting for resting time during the day that was included in the ABCD-study. However, our sensitivity analysis where two different common cutoffs in both cohorts were used, showed the same associations. We have no details about the timing of sleep during the daytime and nighttime, for example, the effects of nocturnal sleep might be different from daytime naps, and we have no information about gestational weight gain in the ABCD-cohort. Moreover, there are important differences in demographics between the two cohorts, causing some heterogeneity in our analysis. There are higher rates of maternal smoking; obesity; gestational diabetes; and cesarean section in the Rhea cohort. The smaller numbers in the RHEA cohort (for the random effects meta-analysis n = 661 vs. n = 2,947 for ABCD cohort) resulted in limited power which might be one of the reasons for the non-significant findings in this cohort. However, effect estimates were in the same direction, specifically with regard to stronger associations in girls. Nevertheless, the random effects meta-analysis indicates low to moderate heterogeneity for most of the outcomes, and pooled analysis was adjusted for cohort and other relevant covariates. Due to numerical difficulties we were not able to provide a measure of risk (OR or RR) for overweight/obesity, instead we calculated Hazard Ratios assuming that the development of overweight/obesity happened at the exact time of the follow-up visit. SEMs allowed us to assess multiple potential mediators but it makes strong assumptions that the relations between all variables are unconfounded. For this reason, we consider the mediation analysis an explorative study and do not claim causality. Lastly, loss to follow-up over the years of childhood caused our analysis to have a lower rate of mothers with short sleep duration in the participant group versus non-participants. We hypothesize that this difference was most likely attributed to higher loss to follow-up rates in non-Greek or non-Dutch origins, as ethnicity was previously shown to be associated with shorter sleep duration in a Dutch population [51], and we corrected our analyses for that. Gestational sleep deprivation and clinical implications Pregnancy is a period where lifestyle interventions are encouraged and parents are more aware of their choices [52]. Healthy gestational sleep has several perinatal benefits, whereas based on our findings, it probably also has positive long-term effects on childhood cardiometabolic health. Primary prevention may be limited to few socioeconomic factors previously related to sleep deprivation, for example, ethnicity and occupation [53]. But also secondary prevention could have a great impact for mothers with sleep disturbances already in early pregnancy. Closer monitoring for glucose metabolism and preterm birth might be extra warranted in mothers with sleep deprivation during pregnancy. Although sleep needs may vary by age and gender, both the National Sleep Foundation and American Academy of Sleep Medicine and Sleep Research Society have recommended 7–9 h of sleep per 24 h for adults [54, 55]. In a sensitivity analysis, we found that the associations are stronger for more severe sleep deprivation (≤5 compared to ≤7 h). During some circumstances sleeping more than 9 h per night might be appropriate too and for other it is uncertain if this is associated with health risk. There are no official sleep recommendations for pregnant women, but we postulate based on our findings that sleep deprivation (meaning a sleep duration of less than 6 h) should be avoided at any stage during pregnancy. Future perspectives Future studies should be done to replicate our findings in other populations, different stages of pregnancy, and to further study the underlying mechanism. Also, the associations we found for gestational sleep deprivation with child adiposity and blood pressure should be further explored in relation not only to sleep deprivation, but also in relation to sleep quality during pregnancy. We tested potential mechanisms with an explorative mediation analysis. Further research on the effects of gestational sleep deprivation on gestational diabetes and shorter gestational age and subsequent childhood metabolic health are needed to investigate causality and opportunities for prevention. We tested for three potential perinatal mediators, however other potential mediators (e.g. childhood lifestyle and sleep) could exist during gestation and early life which may warrant further study. There is one published research protocol of a prospective cohort study that investigates the effects of circadian rhythm on birth and infant outcomes, which can replicate the studied associations [56]. Conclusion Our study is the first analysis on the association between maternal sleep duration during pregnancy and later childhood health. We used data from two ethnical and demographical diverse European cohorts and found that gestational sleep deprivation may be associated with increased risk for overweight and higher blood pressure in offspring, up until the age of 11 years, with more pronounced significant effects in girls than boys. Gestational diabetes and gestational age partly mediated these effects, pointing to altered glucose metabolism and inflammatory pathways as possible biological mechanisms underlying the observed associations. Acknowledgments We are thankful to all the mothers, fathers, and children who participated in the Rhea cohort and Amsterdam Born Children and their Development cohort. The ABCD study has been supported by grants from The Netherlands Organisation for Health Research and Development (ZonMW) and The Netherlands Heart Foundation. Research time of MH was supported by the municipal Amsterdam Healthy Weight Program (Amsterdamse Aanpak Gezond Gewicht). The Rhea project was financially supported by European projects (EU FP6-2003-Food-3-NewGeneris, EU FP6. STREP Hiwate, EU FP7 ENV.2007.1.2.2.2. Project No 211250 Escape, EU FP7-2008-ENV-1.2.1.4 Envirogenomarkers, EU FP7-HEALTH-2009-single stage CHICOS, EU FP7 ENV.2008.1.2.1.6. Proposal No 226285 ENRIECO, EUFP7-HEALTH-2012 Proposal No 308333 HELIX, FP7 European Union project, No. 264357 MeDALL), and the Ministry of health and social solidarity, Greece (Program of Prevention of obesity and neurodevelopmental disorders in preschool children, in Heraklion district, Crete, Greece: 2011–2014; Rhea Plus: Primary Prevention Program of Environmental Risk Factors for Reproductive Health, and Child Health: 2012–15). Dr Lida Chatzi was supported by the National Institute of Environmental Health Sciences (NIEHS/National Institutes of Health (NIH)) grants: R21ES029681, R01ES030691, R01ES029944, R01 ES030364, R21ES028903, and P30ES007048, and by NIH (UH3OD023287). The study was also supported by Seventh Framework Programme and Hartstichting. Disclosure Statements Financial disclosure: None. Non-financial disclosure: None. References 1. Gaillard R , et al. Lifestyle intervention strategies in early life to improve pregnancy outcomes and long-term health of offspring: a narrative review . J Dev Orig Heal Dis . 2019 ; 10 : 314 – 321 . Google Scholar Crossref Search ADS WorldCat 2. Gluckman PD , et al. Effect of in utero and early-life conditions on adult health and disease . N Engl J Med. 2008 ; 359 : 61 – 73 . Google Scholar Crossref Search ADS PubMed WorldCat 3. Davenport MH , et al. Impact of prenatal exercise on neonatal and childhood outcomes: a systematic review and meta-analysis . Br J Sports Med. 2018 ; 52 : 1386 – 1396 . Google Scholar Crossref Search ADS PubMed WorldCat 4. Ruchat SM , et al. Effectiveness of exercise interventions in the prevention of excessive gestational weight gain and postpartum weight retention: a systematic review and meta-analysis . Br J Sports Med. 2018 ; 52 : 1347 – 1356 . Google Scholar Crossref Search ADS PubMed WorldCat 5. Rogozińska E , et al. Effects of antenatal diet and physical activity on maternal and fetal outcomes: individual patient data meta-analysis and health economic evaluation . Health Technol Assess. 2017 ; 21 : 1 – 158 . Google Scholar Crossref Search ADS PubMed WorldCat 6. Tanentsapf I , et al. Systematic review of clinical trials on dietary interventions to prevent excessive weight gain during pregnancy among normal weight, overweight and obese women . BMC Pregnancy Childb. 2011 ; 11 : 81 . Google Scholar Crossref Search ADS WorldCat 7. Mindell JA , et al. Sleep patterns and sleep disturbances across pregnancy . Sleep Med. 2015 ; 16 : 483 – 488 . Google Scholar Crossref Search ADS PubMed WorldCat 8. Facco FL , et al. Sleep disturbances in pregnancy . Obstet Gynecol. 2010 ; 115 : 77 – 83 . Google Scholar Crossref Search ADS PubMed WorldCat 9. Ferraro ZM , et al. The potential value of sleep hygiene for a healthy pregnancy: a brief review . ISRN Family Med. 2014 ; 2014 : 928293 . Google Scholar Crossref Search ADS PubMed WorldCat 10. Reutrakul S , et al. Interactions between sleep, circadian function, and glucose metabolism: implications for risk and severity of diabetes . Ann N Y Acad Sci. 2014 ; 1311 : 151 – 173 . Google Scholar Crossref Search ADS PubMed WorldCat 11. Balserak BI . Sleep-disordered breathing in pregnancy . Am J Respir Crit Care Med. 2014 ; 190 : P1 – P2 . Google Scholar Crossref Search ADS PubMed WorldCat 12. Chang JJ , et al. Sleep deprivation during pregnancy and maternal and fetal outcomes: is there a relationship? Sleep Med Rev. 2010 ; 14 : 107 – 114 . Google Scholar Crossref Search ADS PubMed WorldCat 13. Micheli K , et al. Sleep patterns in late pregnancy and risk of preterm birth and fetal growth restriction . Epidemiology. 2011 ; 22 : 738 – 744 . Google Scholar Crossref Search ADS PubMed WorldCat 14. O’Keeffe M , et al. Sleep duration and disorders in pregnancy: implications for glucose metabolism and pregnancy outcomes . Int J Obes (Lond). 2013 ; 37 : 765 – 770 . Google Scholar Crossref Search ADS PubMed WorldCat 15. Oyiengo D , et al. Sleep disorders in pregnancy . Clin Chest Med. 2014 ; 35 : 571 – 587 . Google Scholar Crossref Search ADS PubMed WorldCat 16. Reutrakul S , et al. Short sleep duration and hyperglycemia in pregnancy: aggregate and individual patient data meta-analysis . Sleep Med Rev. 2018 ; 40 : 31 – 42 . Google Scholar Crossref Search ADS PubMed WorldCat 17. Oostvogels AJJM , et al. Does maternal pre-pregnancy overweight or obesity influence offspring’s growth patterns from birth up to 7 years? The ABCD-study . Early Hum Dev. 2017 ; 113 : 62 – 70 . Google Scholar Crossref Search ADS PubMed WorldCat 18. Nehring I , et al. Gestational diabetes predicts the risk of childhood overweight and abdominal circumference independent of maternal obesity . Diabet Med. 2013 ; 30 : 1449 – 1456 . Google Scholar Crossref Search ADS PubMed WorldCat 19. Gillman MW , et al. Maternal gestational diabetes, birth weight, and adolescent obesity . Pediatrics. 2003 ; 111 : e221 – e226 . Google Scholar Crossref Search ADS PubMed WorldCat 20. Boney CM , et al. Metabolic syndrome in childhood: association with birth weight, maternal obesity, and gestational diabetes mellitus . Pediatrics. 2005 ; 115 : e290 – e296 . Google Scholar Crossref Search ADS PubMed WorldCat 21. Sun D , et al. Birthweight and cardiometabolic risk patterns in multiracial children . Int J Obes (Lond). 2018 ; 42 : 20 – 27 . Google Scholar Crossref Search ADS PubMed WorldCat 22. Khalyfa A , et al. Sex dimorphism in late gestational sleep fragmentation and metabolic dysfunction in offspring mice . Sleep. 2015 ; 38 : 545 – 557 . Google Scholar Crossref Search ADS PubMed WorldCat 23. Raimundo JR , et al. Autonomic and renal alterations in the offspring of sleep-restricted mothers during late pregnancy . Clinics (Sao Paulo). 2016 ; 71 : 521 – 527 . Google Scholar Crossref Search ADS PubMed WorldCat 24. Argeri R , et al. Sleep restriction during pregnancy and its effects on blood pressure and renal function among female offspring . Physiol Rep. 2016 ; 4 : e12888 . Google Scholar Crossref Search ADS PubMed WorldCat 25. Kuula L , et al. Sleep and lipid profile during transition from childhood to adolescence . J Pediatr. 2016 ; 177 : 173 – 178.e1 . Google Scholar Crossref Search ADS PubMed WorldCat 26. Cespedes EM , et al. Longitudinal associations of sleep curtailment with metabolic risk in mid-childhood . Obesity (Silver Spring). 2014 ; 22 : 2586 – 2592 . Google Scholar Crossref Search ADS PubMed WorldCat 27. Taveras EM , et al. Chronic sleep curtailment and adiposity . Pediatrics. 2014 ; 133 : 1013 – 1022 . Google Scholar Crossref Search ADS PubMed WorldCat 28. Felső R , et al. Relationship between sleep duration and childhood obesity: systematic review including the potential underlying mechanisms . Nutr Metab Cardiovasc Dis. 2017 ; 27 : 751 – 761 . Google Scholar Crossref Search ADS PubMed WorldCat 29. Fatima Y , et al. Longitudinal impact of sleep on overweight and obesity in children and adolescents: a systematic review and bias-adjusted meta-analysis . Obes Rev. 2015 ; 16 : 137 – 149 . Google Scholar Crossref Search ADS PubMed WorldCat 30. Chaput JP , et al. Systematic review of the relationships between sleep duration and health indicators in the early years (0–4 years) . BMC Public Health. 2017 ; 17 : 855 . Google Scholar Crossref Search ADS PubMed WorldCat 31. El-Sheikh M , et al. Growth in body mass index from childhood into adolescence: the role of sleep duration and quality . J Early Adolesc. 2014 ; 34 : 1145 – 1166 . Google Scholar Crossref Search ADS WorldCat 32. Reither EN , et al. Ethnic variation in the association between sleep and body mass among US adolescents . Int J Obes (Lond). 2014 ; 38 : 944 – 949 . Google Scholar Crossref Search ADS PubMed WorldCat 33. Chatzi L , et al. Cohort profile: the mother-child cohort in crete, Greece (Rhea Study) . Int J Epidemiol. 2017 ; 46 : 1392 – 1393k . Google Scholar Crossref Search ADS PubMed WorldCat 34. van Eijsden M , et al. Cohort profile: the Amsterdam Born Children and their Development (ABCD) study . Int J Epidemiol. 2011 ; 40 : 1176 – 1186 . Google Scholar Crossref Search ADS PubMed WorldCat 35. Kushner RF , et al. Is the impedance index (ht2/R) significant in predicting total body water? Am J Clin Nutr. 1992 ; 56 : 835 – 839 . Google Scholar Crossref Search ADS PubMed WorldCat 36. Cole TJ , et al. Extended international (IOTF) body mass index cut-offs for thinness, overweight and obesity . Pediatr Obes. 2012 ; 7 : 284 – 294 . Google Scholar Crossref Search ADS PubMed WorldCat 37. Textor J , et al. DAGitty: a graphical tool for analyzing causal diagrams . Epidemiol. 2011 ; 22 : 745 . Google Scholar Crossref Search ADS WorldCat 38. Coble PA , et al. Childbearing in women with and without a history of affective disorder. II. Electroencephalographic sleep . Compr Psychiatry. 1994 ; 35 : 215 – 224 . Google Scholar Crossref Search ADS PubMed WorldCat 39. Sedov ID , et al. Sleep quality during pregnancy: a meta-analysis . Sleep Med Rev. 2018 ; 38 : 168 – 176 . Google Scholar Crossref Search ADS PubMed WorldCat 40. Izci Balserak B . Sleep disordered breathing in pregnancy . Breathe (Sheff). 2015 ; 11 : 268 – 277 . Google Scholar Crossref Search ADS PubMed WorldCat 41. Medic G , et al. Short- and long-term health consequences of sleep disruption . Nat Sci Sleep. 2017 ; 9 : 151 – 161 . Google Scholar Crossref Search ADS PubMed WorldCat 42. Cappuccio FP , et al. Meta-analysis of short sleep duration and obesity in children and adults . Sleep. 2008 ; 31 : 619 – 626 . Google Scholar Crossref Search ADS PubMed WorldCat 43. Grandner MA , et al. Habitual sleep duration associated with self-reported and objectively determined cardiometabolic risk factors . Sleep Med. 2014 ; 15 : 42 – 50 . Google Scholar Crossref Search ADS PubMed WorldCat 44. Holliday EG , et al. Short sleep duration is associated with risk of future diabetes but not cardiovascular disease: a prospective study and meta-analysis . PLoS One. 2013 ; 8 : e82305 . Google Scholar Crossref Search ADS PubMed WorldCat 45. Kruisbrink M , et al. Association of sleep duration and quality with blood lipids: a systematic review and meta-analysis of prospective studies . BMJ Open. 2017 ; 7 : e018585 . Google Scholar Crossref Search ADS PubMed WorldCat 46. Damm P , et al. Gestational diabetes mellitus and long-term consequences for mother and offspring: a view from Denmark . Diabetologia. 2016 ; 59 : 1396 – 1399 . Google Scholar Crossref Search ADS PubMed WorldCat 47. Okun ML , et al. Sleep disruption during pregnancy: how does it influence serum cytokines? J Reprod Immunol. 2007 ; 73 : 158 – 165 . Google Scholar Crossref Search ADS PubMed WorldCat 48. Holcberg G , et al. Increased production of tumor necrosis factor-alpha TNF-alpha by IUGR human placentae . Eur J Obstet Gynecol Reprod Biol. 2001 ; 94 : 69 – 72 . Google Scholar Crossref Search ADS PubMed WorldCat 49. Perrin EM , et al. Elevations of inflammatory proteins in neonatal blood are associated with obesity and overweight among 2-year-old children born extremely premature . Pediatr Res. 2018 ; 83 : 1110 – 1119 . Google Scholar Crossref Search ADS PubMed WorldCat 50. Tarrade A , et al. Placental contribution to nutritional programming of health and diseases: epigenetics and sexual dimorphism . J Exp Biol. 2015 ; 218 : 50 – 58 . Google Scholar Crossref Search ADS PubMed WorldCat 51. Anujuo K , et al. Contribution of short sleep duration to ethnic differences in cardiovascular disease: results from a cohort study in the Netherlands . BMJ Open. 2017 ; 7 : e017645 . Google Scholar Crossref Search ADS PubMed WorldCat 52. Hall WA , et al. Translating knowledge directly to childbearing women: a study of Canadian women’s preferences . Health Care Women Int. 2013 ; 34 : 363 – 379 . Google Scholar Crossref Search ADS PubMed WorldCat 53. Feinstein L , et al. Racial/ethnic disparities in sleep duration and sleep disturbances among pregnant and non-pregnant women in the United States . J Sleep Res. 2020 ; 00 : e13000 . Google Scholar OpenURL Placeholder Text WorldCat 54. Hirshkowitz M , et al. National Sleep Foundation’s updated sleep duration recommendations: final report . Sleep Health. 2015 ; 1 : 233 – 243 . Google Scholar Crossref Search ADS PubMed WorldCat 55. Watson NF , et al. Recommended amount of sleep for a healthy adult: a joint consensus statement of the American Academy of Sleep Medicine and Sleep Research Society . Sleep. 2015 ; 38 : 843 – 844 . Google Scholar Crossref Search ADS PubMed WorldCat 56. Kaur S , et al. Circadian rhythm and its association with birth and infant outcomes: research protocol of a prospective cohort study . BMC Pregnancy Childb. 2020 ; 20 : 96 . Google Scholar Crossref Search ADS WorldCat Author notes These authors contributed equally to this work. © Sleep Research Society 2020. Published by Oxford University Press on behalf of the Sleep Research Society. This is an Open Access article distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs licence (http://creativecommons.org/licenses/by-nc-nd/4.0/), which permits non-commercial reproduction and distribution of the work, in any medium, provided the original work is not altered or transformed in any way, and that the work is properly cited. For commercial re-use, please contact [email protected] © Sleep Research Society 2020. Published by Oxford University Press on behalf of the Sleep Research Society.

Matsangas, Panagiotis; Shattuck, Nita Lewis

doi: 10.1093/sleep/zsaa118pmid: 32531020

Abstract Study Objectives This field study (a) assessed sleep quality of sailors on the U.S. Navy (USN) ships while underway, (b) investigated whether the Pittsburgh Sleep Quality Index (PSQI) scores were affected by occupational factors and sleep attributes, and (c) assessed whether the PSQI could predict impaired psychomotor vigilance performance. Methods Longitudinal field assessment of fit-for-duty USN sailors performing their underway duties (N = 944, 79.0% males, median age 26 years). Participants completed questionnaires, wore actigraphs, completed logs, and performed the wrist-worn 3-min Psychomotor Vigilance Task (PVT). Results Sailors slept on average 6.60 ± 1.01 h/day with 86.9% splitting their sleep into more than one episode/day. The median PSQI Global score was 8 (interquartile range [IQR] = 5); 80.4% of the population were classified as “poor sleepers” with PSQI scores >5. PSQI scores were affected by sailor occupational group, rank, daily sleep duration, and number of sleep episodes/day. Sleep quality showed a U-shape association with daily sleep duration due to the confounding effect of split sleep. Sailors with PSQI scores >9 had 21.1% slower reaction times (p < 0.001) and 32.8%–61.5% more lapses combined with false starts (all p < 0.001) than sailors with PSQI scores ≤9. Compared to males and officers, females and enlisted personnel had 86% and 23% higher risk, respectively, of having PSQI scores >9. Sailors in the PSQI > 9 group had more pronounced split sleep. Conclusions Working on Navy ships is associated with elevated PSQI scores, a high incidence of poor sleep, and degraded psychomotor vigilance performance. The widely used PSQI score>5 criterion should be further validated in active-duty service member populations. sleep quality, psychomotor vigilance performance, watchstanding, naval operational environment, military personnel, military veterans Statement of Significance The Pittsburgh Sleep Quality Index (PSQI) is a widely used tool to assess sleep quality. PSQI has not been validated, however, for military populations. Based on a sample of fit-for-duty USN sailors performing their duties on ships, our results showed that sleep quality was associated with individual sailor occupational group/work schedule, rank, daily sleep duration, and number of sleep episodes/day. All occupational groups had a large percentage of poor sleepers (PSQI score > 5), whereas a higher criterion (PSQI score > 9) was associated with impaired psychomotor vigilance performance. These findings expand our knowledge regarding sleep quality at sea and the usefulness of PSQI in the military, but also stress the need for further validation of the PSQI > 5 criterion on military populations. Introduction As part of a multiyear effort, researchers at the Naval Postgraduate School have undertaken a series of studies to validate a number of widely used screening tools in the military operational environment. The first study focused on the utility of the Epworth Sleepiness Scale (ESS) [1] to determine if ESS scores were predictive of actigraphically determined sleep and psychomotor vigilance performance of active duty crewmembers working their normal duties on a U.S. Navy (USN) ship [2]. The next study assessed whether the ESS used with the Fatigue Severity Scale could differentiate between subjective sleepiness and subjective fatigue [3]. The current study focuses on the Pittsburgh Sleep Quality Index (PSQI), a self-rated screening questionnaire which assesses sleep quality and disturbances [4]. The clinimetric and clinical properties of the PSQI, using a cutoff criterion score of more than 5, suggest its utility both in psychiatric clinical practice and research activities to distinguish between good and poor sleepers [4]. Being the most commonly used general sleep measure [5, 6], the PSQI has been used and validated in various populations; in primary insomnia patients [7], in healthy control subjects [4], in various sleep and psychiatric disorders [4, 8], in samples with health conditions [9], in patients with chronic pain [10, 11], and in elderly populations [12, 13]. However [14], the validity of PSQI in active duty military personnel and veterans is not well investigated. Our review showed only two studies that focused explicitly on the psychometric properties of PSQI in military populations, mainly in military service veterans and active duty service members with sleep problems [5, 15, 16]. Along these lines, Troxel and colleagues noted that few studies in military populations have used the full, validated instrument [6]. Our review failed to identify any studies focusing explicitly on the operational factors affecting PSQI scores in active duty members working in the operational maritime and naval environment. In agreement with Troxel et al. [6] that the PSQI needs further validation in military populations, our field study has three goals. First, assess subjective sleep quality in a large sample of active duty service members on USN ships while performing their underway duties. Second, investigate the association between PSQI scores and work schedules used on USN ships. Third, determine whether PSQI scores can differentiate amongst levels of psychomotor vigilance performance. This study is part of a multiyear effort designed to systematically and empirically assess a wide range of watchstanding schedules which are used in the USN, measure the work and rest patterns of sailors in a variety of shipboard operational environments, and provide insight and guidance for future naval operations. Methods Participants Sailors assigned to seven surface combatants of the USN (one Nimitz-class aircraft carrier, one Ticonderoga-class cruiser, and five Arleigh Burke-class destroyers) were recruited and enrolled in studies of work and rest patterns. Data were collected in six periods (December 2012, May 2013, June and November 2014, June 2017, December 2017–January 2018). Two ships were conducting underway training exercises and five ships were operationally deployed. All sailors onboard during the study periods were eligible to participate. Study procedures were approved by the Institutional Review Board of the Naval Postgraduate School. Participants provided written informed consent. Based on their dominant work schedule, sailors were classified into one of four occupational groups. “Watchstanders” included sailors who “stood watch,” a period of time during which a sailor is assigned specific, detailed responsibilities on a recurring basis [17]. The daily schedule of the watchstanders is knitted around their watch during which time they cannot leave their post unless relieved of duty. The watchstander group included sailors working various watchstanding schedules, with ~70% on fixed schedules standing watch at the same time every day (e.g. 3 h-on/9 h-off, 4 h-on/8 h-off, 6 h-on/18 h-off) and ~30% on rotating schedules standing watch at different times every day (e.g. 5 h-on/10 h-off, 5 h-on/15 h-off). “Non-watchstanders” were divided into three sub-groups: “maintenance shiftworkers” included sailors performing maintenance on fixed 12-h shifts, with the day shift commencing early in the morning hours and the night shift commencing in early evening hours. Compared to the shifts of watchstanders, a maintenance shift is more flexible, that is, it includes more self-paced tasks and sailors can take brief rest periods if needed. “Galley workers” included sailors involved in food preparation who work between early morning and late evening in the ship’s kitchen or galley. In general, galley workers slept at night. Lastly, the “Dayworkers” group included sailors who worked during the morning to early evening hours and slept at night. Regardless of their occupational group, while underway all sailors are responsible for carrying out various duties during their time off watch/shift, for example, attending meetings, training, drills, or other work and operational commitments. Equipment and instruments Actigraphy Sleep was assessed by wrist-worn actigraphy and activity logs following existing recommendations [18, 19]. Specifically, we used information from activity logs to manually determine start and end times of rest/sleep intervals using the actigraphy data as the primary source for the sleep analysis. Within each rest/sleep interval, the actigraphically assessed sleep was automatically calculated. Rest/sleep episodes were distinctly different from sleep periods and were readily identified. The Motionlogger Watch (Ambulatory Monitoring, Inc. [AMI], Ardsley, NY) and the Spectrum (Philips-Respironics [PR], Bend, OR) actiwatch were used. Data for both devices were collected in 1-min epochs. AMI data (collected in the Zero-Crossing Mode) were scored using Action W version 2.7.2155 software using the Cole–Kripke algorithm with rescoring rules. The criterion for sleep and wake episodes was 5 min; the sleep latency criterion was no more than 1 min awake in a 20-min period (all default values for this software). PR data were scored using Actiware software version 6.0.0 (Phillips Respironics, Bend, OR) using the medium sensitivity threshold (40 counts per epoch), with 10 immobile minutes as the criterion for sleep onset and sleep end (all default values for this software). Previous research has shown that AMI data analyzed with Cole–Kripke and PR data analyzed with medium sensitivity parameters assess total sleep time for an approximately 8-h night sleep episode with 3 min precision [20]. Psychomotor vigilance performance Performance data were collected using a 3-min version of the original 10-min Psychomotor Vigilance Task—PVT [21] which was embedded in the AMI actigraphs [22, 23]. The PVT is a simple reaction time test where participants are required to press a button in response to a visual stimulus. The nominal interstimulus interval ranged from 2 to 10 s. The actigraphic display was lit (red backlight on) for one second and the letters “PUSH” were used as the visual stimuli; the response time was then displayed in milliseconds. Questionnaires The pre-study questionnaire included demographic information (age, gender, rate/rank, department, use of caffeinated beverages (e.g. tea, coffee, soft drinks, and energy drinks), use of nicotine products, having an exercise routine, taking medications—prescribed or over-the-counter). The end-of-study questionnaire asked the participants to indicate whether they stood watch during the underway and to complete the Pittsburg Sleep Quality Index—PSQI [4] to assess sleep quality. From the 24 PSQI items, 19 are self-rated and 5 items are rated by the bedpartner or the roommate. The self-rated questions yield seven component scores (sleep quality, sleep latency, duration, sleep efficiency, sleep disturbances, sleep medication use, and daytime dysfunction) rated from 0 (better) to 3 (worse). The total score, ranging from 0 (better) to 21 (worse), is the summation of the component scores. Individuals with a PSQI total score ≤5 are characterized as good sleepers, whereas scores >5 are associated with poor sleep quality. The PSQI has a sensitivity of 89.6% and a specificity of 86.5% (κ = 0.75, p < 0.001) in non-military populations, and an internal consistency α = 0.83 [4]. Even though the original version of PSQI referred to sleep quality during the previous month, the ecological validity of the tool, that is, subjects’ accuracy in recalling sleep quality, has been demonstrated for various reporting periods from 3 days to 1 month [24]. The period of recall for the PSQI we used was 2 weeks. Study design and procedures The information presented herein is a subset of measures from multiple field assessments on USN ships. Data were collected using a prospective naturalistic design with an underway data collection period of 7–18 days. Initially, sailors completed the pre-study questionnaire which included demographic questions. Then over the data collection period, sailors were asked to wear an actiwatch, complete their activity log once per day, and perform the 3-min PVT. Watchstanders and shiftworkers were asked to perform the PVT before and after each watch/shift; non-shiftworkers were asked to perform the PVT once after awakening and once before bedtime. The protocol of PVT data collection for the watchstanders/shiftworkers ensured that their cognitive performance was assessed during the operationally relevant watch/shift periods. At the end of the data collection, sailors completed the post-study questionnaire. At the beginning of each data collection, sailors had been assigned to the same daily schedule for at least 3 days. Analytical approach Initially, 944 sailors were enrolled (Figure 1). Sailors using medications (sleeping aids, anti-inflammatory, and anti-depressant drugs—n = 57) or with missing data (n = 15) were excluded. Therefore, the analysis was based on 872 sailors (655 with actigraphy data and 267 with PVT data). Figure 1. Open in new tabDownload slide Participation flow chart. Figure 1. Open in new tabDownload slide Participation flow chart. PVT performance was assessed by mean reaction time (RT), mean response speed (1/RT), fastest 10% RT (i.e. 10th percentile of RT), slowest 10% of 1/RT (i.e. 10th percentile of 1/RT), percentage of 335 and 500 ms lapses, percentage of false starts, and percentage of lapses combined with false starts. No imputation was applied to PVT data. Participants were included in the PVT data analysis only when they performed the PVT on at least 50% of the days in the data collection period; the pattern of missing tests did not show a systematic bias. Based on these criteria, PVT adherence was ~75% and was not associated with age, gender, or rank (all p > 0.05). Sleep analysis was based on sleep episode duration, awake period duration, daily sleep duration, and on the number of sleep episodes per day. The metric “average number of sleep episodes/day” is calculated as the ratio of the number of sleep episodes during the data collection period divided by the number of data collection days. Initially, we calculated the average number of sleep episodes/day for each participant. Next, we calculated the grand average number of sleep episodes/day for those sailors who napped during the data collection period. Sleep episodes recorded in sleep logs were used to impute missing actigraphic data and accounted for 1.7% of all sleep episodes. Imputation was applied to sleep data only when: (a) there was a gap in actigraphy data within which the activity log showed a sleep interval and (b) the pattern of actigraphy data, assisted by the activity logs, was such to assure confidence in the interpolation of the sleep interval. Sleep and PVT metrics were aggregated to get an average score for each individual over the entire study period. Therefore, both sleep and PVT metrics provided an overall estimate of sailor alertness and performance during the data collection period. Statistical analysis was conducted using JMP statistical software (JMP Pro 15, SAS Institute, Cary, NC). First, all variables underwent descriptive statistical analysis to identify anomalous entries and to determine demographic characteristics. Next, we compared the occupational groups in terms of demographic characteristics and PSQI scores. General linear model analysis was used to assess the predictor factors of the PSQI Global scores. Potential predictor factors included sailor occupational group, rank group, daily sleep duration, number of sleep episodes per day, and the interaction term between daily sleep duration and the number of sleep episodes per day. Lastly, partition analysis was used to explore the association between PVT response speed and PSQI Global score. Data normality was assessed with the Shapiro–Wilk W test. Correlations were assessed with Spearman’s rho. Fisher’s Exact test was used for pairwise comparisons. Tukey–Kramer Honest Significant Difference (HSD) test and Dunn method for joint ranking were used for multiple comparisons. An alpha level of 0.05 was used to determine statistical significance. Post hoc statistical significance was assessed using the Benjamini–Hochberg False Discovery Rate (BH-FDR) controlling procedure [25] with q = 0.20. Summaries of continuous data are reported as mean (M) ± standard deviation (SD) or median (MD)—interquartile range (IQR) as appropriate. Results Participants had a median age of 25 (IQR = 7) years, and were predominantly males (692, 79.4%) and enlisted personnel (731, 83.8%). In terms of demographic characteristics, the study sample did not differ substantively from the population of active duty service members in the USN [26]. Approximately 9% (n = 82) of the sailors were using prescription or over-the-counter medications, that is, allergy drugs (24, 2.77%), high blood pressure drugs (12, 1.38%), acid reflux drugs (11, 1.27%), anti-emetic drugs (8, 0.92%), migraine/headaches drugs (3, 0.35%), anti-viral drugs (2, 0.23%), and other (22, 2.54%). As assessed by actigraphy, the average duration of sleep episodes was 4.63 ± 1.40 h, whereas the average duration of awake periods was 11.6 ± 3.32 h. Participants slept an average of 6.60 ± 1.01 h daily (ranging from 1.83 to 9.52 h), with 569 (86.9%) sailors splitting their sleep into 1.5 episodes per day (median value with IQR = 0.58). The median PSQI Global score was 8 (IQR = 5), ranging from 1 to 18. PSQI scores indicated that 80.4% of the participants were “poor sleepers” (PSQI score > 5). From the 872 sailors, 666 were watchstanders and 206 non-watchstanders (i.e. 39 maintenance shiftworkers, 32 galley workers, and 135 dayworkers). Occupational groups did not differ in terms of gender (Fisher’s Exact test, p = 0.174), but watchstanders were on average ~2.6 years younger than dayworkers (Dunn method for joint ranking p = 0.012). Compared to the other occupational groups, watchstanders had the shortest sleep episodes (4.32 ± 1.23 h; Dunn method for joint ranking, all p < 0.01) and the shortest awake periods (11.0 ± 3.10 h; all p < 0.01) but their average daily sleep duration (6.51 ± 1.03 h) differed only from dayworkers (p < 0.001). Galley workers and watchstanders had the highest (worst) PSQI Global scores (Table 1). Compared to dayworkers, watchstanders had worse PSQI scores in terms of the Global score, sleep latency, sleep duration, habitual sleep efficiency, sleep quality, and daytime dysfunction. The same pattern was evident also in galley workers. Specifically, compared to dayworkers, galley workers had worse PSQI scores in terms of the Global score, sleep duration, and sleep quality. These results are shown in Table 1. Using the cutoff criterion of PSQI Global score >5, 84 (62.2%) dayworkers were identified as “poor sleepers” as compared to 30 (76.9%) maintenance shiftworkers (Fisher’s Exact test, p = 0.125 compared to dayworkers), 27 (84.4%) galley workers (p = 0.021), and 560 (84.1%) watchstanders (p < 0.001). Table 1. Sleep attributes and PSQI scores PSQI . Entire sample (N = 872) . Dayworkers (n = 135) . Maintenance shiftworkers (n = 39) . Galley workers (n = 32) . Watchstanders (n = 666) . Sleep attributes  Sleep episode duration (h), M ± SD|| 4.63 ± 1.40 5.75 ± 1.43 6.03 ± 1.23 5.37 ± 1.51 4.33 ± 1.23*3,†2,††3  Awake period duration (h), M ± SD|| 11.6 ± 3.32 13.2 ± 3.20 12.9 ± 3.38 15.9 ± 3.26¶1 11.0 ± 3.10*3,†2,††3  Daily sleep duration (h), M ± SD|| 6.60 ± 1.01 7.06 ± 0.86 6.74 ± 0.97 6.54 ± 0.83 6.51 ± 1.03*3  Sailors with split sleep, # (%)# 569 (86.9%) 67 (69.8%) 16 (61.5%) 9 (42.9%)¶1 477 (93.2%)*3,†3,††3  Sleep episodes per day (#), MD (IQR)||,¶¶ 1.5 (0.58) 1.29 (0.43) 1.44 (0.48) 1.22 (0.39) 1.55 (0.61)*3,††1 PSQI, MD (IQR)  Global score 8 (5) 7 (5) 8 (4) 9 (6.5) ¶2 9 (5)*1  Sleep latency 1 (1) 1 (2) 1 (1) 1.5 (2) 2 (1)*2  Sleep duration 2 (1) 1 (2) 1 (1) 2 (0.75) ¶3,§1 2 (0)*1,†2  Habitual sleep efficiency 0 (1) 0 (1) 0 (2) 0.5 (1) 1 (2)*1  Sleep disturbances 1 (1) 1 (1) 1 (1) 1 (1) 1 (1)  Subjective sleep quality 1 (1) 1 (1) 1 (1) 2 (1) ¶3 1 (1)*2,††1  Use of sleeping medication 0 (0) 0 (0) 0 (0) 0 (1) 0 (0)  Daytime dysfunction 1 (1) 1 (0) 1 (0) 1 (1) 1 (1)*3 PSQI . Entire sample (N = 872) . Dayworkers (n = 135) . Maintenance shiftworkers (n = 39) . Galley workers (n = 32) . Watchstanders (n = 666) . Sleep attributes  Sleep episode duration (h), M ± SD|| 4.63 ± 1.40 5.75 ± 1.43 6.03 ± 1.23 5.37 ± 1.51 4.33 ± 1.23*3,†2,††3  Awake period duration (h), M ± SD|| 11.6 ± 3.32 13.2 ± 3.20 12.9 ± 3.38 15.9 ± 3.26¶1 11.0 ± 3.10*3,†2,††3  Daily sleep duration (h), M ± SD|| 6.60 ± 1.01 7.06 ± 0.86 6.74 ± 0.97 6.54 ± 0.83 6.51 ± 1.03*3  Sailors with split sleep, # (%)# 569 (86.9%) 67 (69.8%) 16 (61.5%) 9 (42.9%)¶1 477 (93.2%)*3,†3,††3  Sleep episodes per day (#), MD (IQR)||,¶¶ 1.5 (0.58) 1.29 (0.43) 1.44 (0.48) 1.22 (0.39) 1.55 (0.61)*3,††1 PSQI, MD (IQR)  Global score 8 (5) 7 (5) 8 (4) 9 (6.5) ¶2 9 (5)*1  Sleep latency 1 (1) 1 (2) 1 (1) 1.5 (2) 2 (1)*2  Sleep duration 2 (1) 1 (2) 1 (1) 2 (0.75) ¶3,§1 2 (0)*1,†2  Habitual sleep efficiency 0 (1) 0 (1) 0 (2) 0.5 (1) 1 (2)*1  Sleep disturbances 1 (1) 1 (1) 1 (1) 1 (1) 1 (1)  Subjective sleep quality 1 (1) 1 (1) 1 (1) 2 (1) ¶3 1 (1)*2,††1  Use of sleeping medication 0 (0) 0 (0) 0 (0) 0 (1) 0 (0)  Daytime dysfunction 1 (1) 1 (0) 1 (0) 1 (1) 1 (1)*3 Statistical significance for differences: “1”: p < 0.05; “2”: p < 0.01; “3”: p < 0.001. *Difference between “Watchstanders” and “Dayworkers” groups. †Difference between “Watchstanders” and “Maintenance shiftworkers” groups. ††Difference between “Watchstanders” and “Galley workers” groups. ¶Difference between “Galley workers” and “Dayworkers” groups. §Difference between “Galley workers” and “Maintenance shiftworkers”. ||Multiple comparisons with non-parametric Dunn method for joint ranking. #Pairwise comparisons with Fisher’s Exact Test. Post hoc analysis for statistical significance with the BH-FDR controlling procedure. ¶¶For sailors with split sleep. Open in new tab Table 1. Sleep attributes and PSQI scores PSQI . Entire sample (N = 872) . Dayworkers (n = 135) . Maintenance shiftworkers (n = 39) . Galley workers (n = 32) . Watchstanders (n = 666) . Sleep attributes  Sleep episode duration (h), M ± SD|| 4.63 ± 1.40 5.75 ± 1.43 6.03 ± 1.23 5.37 ± 1.51 4.33 ± 1.23*3,†2,††3  Awake period duration (h), M ± SD|| 11.6 ± 3.32 13.2 ± 3.20 12.9 ± 3.38 15.9 ± 3.26¶1 11.0 ± 3.10*3,†2,††3  Daily sleep duration (h), M ± SD|| 6.60 ± 1.01 7.06 ± 0.86 6.74 ± 0.97 6.54 ± 0.83 6.51 ± 1.03*3  Sailors with split sleep, # (%)# 569 (86.9%) 67 (69.8%) 16 (61.5%) 9 (42.9%)¶1 477 (93.2%)*3,†3,††3  Sleep episodes per day (#), MD (IQR)||,¶¶ 1.5 (0.58) 1.29 (0.43) 1.44 (0.48) 1.22 (0.39) 1.55 (0.61)*3,††1 PSQI, MD (IQR)  Global score 8 (5) 7 (5) 8 (4) 9 (6.5) ¶2 9 (5)*1  Sleep latency 1 (1) 1 (2) 1 (1) 1.5 (2) 2 (1)*2  Sleep duration 2 (1) 1 (2) 1 (1) 2 (0.75) ¶3,§1 2 (0)*1,†2  Habitual sleep efficiency 0 (1) 0 (1) 0 (2) 0.5 (1) 1 (2)*1  Sleep disturbances 1 (1) 1 (1) 1 (1) 1 (1) 1 (1)  Subjective sleep quality 1 (1) 1 (1) 1 (1) 2 (1) ¶3 1 (1)*2,††1  Use of sleeping medication 0 (0) 0 (0) 0 (0) 0 (1) 0 (0)  Daytime dysfunction 1 (1) 1 (0) 1 (0) 1 (1) 1 (1)*3 PSQI . Entire sample (N = 872) . Dayworkers (n = 135) . Maintenance shiftworkers (n = 39) . Galley workers (n = 32) . Watchstanders (n = 666) . Sleep attributes  Sleep episode duration (h), M ± SD|| 4.63 ± 1.40 5.75 ± 1.43 6.03 ± 1.23 5.37 ± 1.51 4.33 ± 1.23*3,†2,††3  Awake period duration (h), M ± SD|| 11.6 ± 3.32 13.2 ± 3.20 12.9 ± 3.38 15.9 ± 3.26¶1 11.0 ± 3.10*3,†2,††3  Daily sleep duration (h), M ± SD|| 6.60 ± 1.01 7.06 ± 0.86 6.74 ± 0.97 6.54 ± 0.83 6.51 ± 1.03*3  Sailors with split sleep, # (%)# 569 (86.9%) 67 (69.8%) 16 (61.5%) 9 (42.9%)¶1 477 (93.2%)*3,†3,††3  Sleep episodes per day (#), MD (IQR)||,¶¶ 1.5 (0.58) 1.29 (0.43) 1.44 (0.48) 1.22 (0.39) 1.55 (0.61)*3,††1 PSQI, MD (IQR)  Global score 8 (5) 7 (5) 8 (4) 9 (6.5) ¶2 9 (5)*1  Sleep latency 1 (1) 1 (2) 1 (1) 1.5 (2) 2 (1)*2  Sleep duration 2 (1) 1 (2) 1 (1) 2 (0.75) ¶3,§1 2 (0)*1,†2  Habitual sleep efficiency 0 (1) 0 (1) 0 (2) 0.5 (1) 1 (2)*1  Sleep disturbances 1 (1) 1 (1) 1 (1) 1 (1) 1 (1)  Subjective sleep quality 1 (1) 1 (1) 1 (1) 2 (1) ¶3 1 (1)*2,††1  Use of sleeping medication 0 (0) 0 (0) 0 (0) 0 (1) 0 (0)  Daytime dysfunction 1 (1) 1 (0) 1 (0) 1 (1) 1 (1)*3 Statistical significance for differences: “1”: p < 0.05; “2”: p < 0.01; “3”: p < 0.001. *Difference between “Watchstanders” and “Dayworkers” groups. †Difference between “Watchstanders” and “Maintenance shiftworkers” groups. ††Difference between “Watchstanders” and “Galley workers” groups. ¶Difference between “Galley workers” and “Dayworkers” groups. §Difference between “Galley workers” and “Maintenance shiftworkers”. ||Multiple comparisons with non-parametric Dunn method for joint ranking. #Pairwise comparisons with Fisher’s Exact Test. Post hoc analysis for statistical significance with the BH-FDR controlling procedure. ¶¶For sailors with split sleep. Open in new tab Predictors of PSQI Global scores Next, we assessed the predictors of the PSQI Global scores (Table 2). Potential predictors included sailor work schedule group, rank group, daily sleep duration, number of sleep episodes per day, and the interaction term between daily sleep duration and number of sleep episodes per day. Daily sleep duration and the number of sleep episodes per day were not correlated (Spearman’s rho = 0.046, p = 0.245). The overall model was statistically significant, F(14, 640) = 8.06, p < 0.001. Adjusted for ship and gender, PSQI Global scores were associated with work schedule group (p = 0.004) and rank group (p < 0.001). Watchstanders and galley workers had higher (worse) PSQI Global scores than dayworkers (Dunnett’s test; p = 0.009 and p = 0.044 respectively), whereas officers had lower (better) PSQI Global scores compared to enlisted personnel. In terms of sleep attributes, daily sleep duration (p = 0.013), number of sleep episodes per day (p = 0.048), and the interaction between daily sleep duration and the number of sleep episodes per day (p = 0.011) were statistically significant predictors. The association between sleep attributes and PSQI Global scores becomes more evident in Figure 2. The upper diagram shows that PSQI scores have a U-shape association with daily sleep duration. Longer daily sleep duration may be associated with low (better) and high (worse) PSQI scores. The reason is the number of sleep episodes that contribute to the accumulation of the daily sleep duration (lower diagram). Longer daily sleep duration, which in the naval environment can be achieved with napping, is associated with worse sleep quality as assessed by PSQI Global score. In contrast, longer daily sleep duration accrued with fewer sleep episodes (i.e. sleep is consolidated in longer sleep episodes) is associated with better sleep quality (i.e. lower PSQI Global scores.) Table 2. Factors for PSQI Global scores Factor . Coefficient . 95% confidence interval . P-value . Gender (female) – – 0.605 Ship – – 0.153 Sleep episodes per day* −14.5 −28.8 to −0.128 0.048 Daily sleep duration* −9.033 −16.1 to −1.92 0.013 Daily sleep duration × Sleep episodes per day 7.26 1.66 to 12.9 0.011 Work schedule group – – 0.004  Control vs. Watchstanders −0.681 −1.33 to −0.032 0.040  Galley Workers vs. Watchstanders 1.11 0.070 to 2.14 0.036  Maintenance Shiftworkers vs. Watchstanders −0.862 −1.81 to −0.085 0.075 Enlisted personnel 0.790 0.462 to 1.12 <0.001 Factor . Coefficient . 95% confidence interval . P-value . Gender (female) – – 0.605 Ship – – 0.153 Sleep episodes per day* −14.5 −28.8 to −0.128 0.048 Daily sleep duration* −9.033 −16.1 to −1.92 0.013 Daily sleep duration × Sleep episodes per day 7.26 1.66 to 12.9 0.011 Work schedule group – – 0.004  Control vs. Watchstanders −0.681 −1.33 to −0.032 0.040  Galley Workers vs. Watchstanders 1.11 0.070 to 2.14 0.036  Maintenance Shiftworkers vs. Watchstanders −0.862 −1.81 to −0.085 0.075 Enlisted personnel 0.790 0.462 to 1.12 <0.001 Note: Analysis conducted on the 655 sailors with questionnaire and actigraphy data. *Square-root transformed values. Open in new tab Table 2. Factors for PSQI Global scores Factor . Coefficient . 95% confidence interval . P-value . Gender (female) – – 0.605 Ship – – 0.153 Sleep episodes per day* −14.5 −28.8 to −0.128 0.048 Daily sleep duration* −9.033 −16.1 to −1.92 0.013 Daily sleep duration × Sleep episodes per day 7.26 1.66 to 12.9 0.011 Work schedule group – – 0.004  Control vs. Watchstanders −0.681 −1.33 to −0.032 0.040  Galley Workers vs. Watchstanders 1.11 0.070 to 2.14 0.036  Maintenance Shiftworkers vs. Watchstanders −0.862 −1.81 to −0.085 0.075 Enlisted personnel 0.790 0.462 to 1.12 <0.001 Factor . Coefficient . 95% confidence interval . P-value . Gender (female) – – 0.605 Ship – – 0.153 Sleep episodes per day* −14.5 −28.8 to −0.128 0.048 Daily sleep duration* −9.033 −16.1 to −1.92 0.013 Daily sleep duration × Sleep episodes per day 7.26 1.66 to 12.9 0.011 Work schedule group – – 0.004  Control vs. Watchstanders −0.681 −1.33 to −0.032 0.040  Galley Workers vs. Watchstanders 1.11 0.070 to 2.14 0.036  Maintenance Shiftworkers vs. Watchstanders −0.862 −1.81 to −0.085 0.075 Enlisted personnel 0.790 0.462 to 1.12 <0.001 Note: Analysis conducted on the 655 sailors with questionnaire and actigraphy data. *Square-root transformed values. Open in new tab Figure 2. Open in new tabDownload slide Daily sleep duration and number of sleep episodes per day by PSQI Global score. Cubic splines were applied to generate the smooth lines (upper diagram: lambda = 1.90; lower diagram: lambda = 32.9). Figure 2. Open in new tabDownload slide Daily sleep duration and number of sleep episodes per day by PSQI Global score. Cubic splines were applied to generate the smooth lines (upper diagram: lambda = 1.90; lower diagram: lambda = 32.9). PVT performance and PSQI Global Scores Partition analysis results suggested that a PSQI Global score of 9 could be used as a cutoff criterion for grouping sailors in terms of their PVT response speed (LogWorth = 8.72). Figure 3 shows PVT response speed versus PSQI Global score. Based on the partition results, sailors with PVT data were classified into three groups. As shown in Table 3, sailors with PSQI Global scores of 5 or less (the conventional criterion for good sleepers) did not differ from sailors with a score between 5 and 9. Compared to sailors with PSQI Global scores of 9 or less, however, sailors with a PSQI Global score of more than 9 had reaction times that were 21.1% slower (p < 0.001), 32.8% more lapses of 355 ms combined with false starts (p < 0.001), and 61.5% more lapses of 500 ms combined with false starts (p < 0.001). Compared to males and officers, females and enlisted personnel had 86% and 23% higher risk, respectively, of having a PSQI Global score of more than 9. PSQI groups also differed in the number of sleep episodes per day with sailors in the PSQI > 9 group having more pronounced split sleep. Figure 3. Open in new tabDownload slide PVT response speed versus PSQI Global score. A cubic spline was applied to generate the smooth line (lambda = 19.3). Figure 3. Open in new tabDownload slide PVT response speed versus PSQI Global score. A cubic spline was applied to generate the smooth line (lambda = 19.3). Table 3. Comparison between PSQI groups Variable . PSQI ≤ 5 (n = 37) . 5 < PSQI ≤ 9 (n = 129) . PSQI > 9 (n = 101) . Age in years, MD (IQR)¶ 24 (8) 26 (8) 25 (6) Sex (females), # (%)§ 8 (21.6%) 22 (17.1%) 34 (33.7%)†2 Enlisted personnel, # (%)§ 26 (70.3%) 99 (76.7%) 93 (92.1%)*2,†2 Watchstanders, # (%)§ 34 (91.9%) 111 (86.1%) 93 (92.1%) Daily sleep duration (h), MD (IQR)¶ 6.68 (1.64) 6.54 (1.35) 6.63 (1.77) Sleep episodes per day (#), MD (IQR)¶ 1.25 (0.53) 1.42 (0.60) 1.70 (0.62)*3,†2 PVT metrics  Mean RT (ms), MD (IQR)¶ 293 (87.4) 303 (94.5) 369 (133)*2,†3  Mean 1/RT, M ± SD†† 4.00 ± 0.67 3.90 ± 0.70 3.38 ± 0.77*3,†3  Fastest 10% RT (ms), MD (IQR)¶ 191 (42.0) 198 (48.7) 226 (57.0)*3,†3  Slowest 10% 1/RT, M ± SD†† 2.40 ± 0.62 2.42 ± 0.64 1.99 ± 0.63*2,†3  False Starts (FS) (%), MD (IQR)¶ 1.00 (2.05) 1.31 (1.68) 1.16 (1.34)*2,†3  Lapses 500 ms (%), MD (IQR)¶ 6.04 (8.13) 6.01 (67.33) 10.4 (10.9)*2,†3  Lapses 355 ms (%), MD (IQR)¶ 12.3 (12.0) 14.1 (15.4) 26.6 (24.9)*3,†3  Lapses 500 ms + FS (%), MD (IQR)¶ 7.60 (8.85) 7.43 (7.11) 12.1 (10.6)*2,†3  Lapses 355 ms + FS (%), MD (IQR) ¶ 13.8 (11.2) 15.8 (16.4) 28.2 (25.6)*3,†3 Variable . PSQI ≤ 5 (n = 37) . 5 < PSQI ≤ 9 (n = 129) . PSQI > 9 (n = 101) . Age in years, MD (IQR)¶ 24 (8) 26 (8) 25 (6) Sex (females), # (%)§ 8 (21.6%) 22 (17.1%) 34 (33.7%)†2 Enlisted personnel, # (%)§ 26 (70.3%) 99 (76.7%) 93 (92.1%)*2,†2 Watchstanders, # (%)§ 34 (91.9%) 111 (86.1%) 93 (92.1%) Daily sleep duration (h), MD (IQR)¶ 6.68 (1.64) 6.54 (1.35) 6.63 (1.77) Sleep episodes per day (#), MD (IQR)¶ 1.25 (0.53) 1.42 (0.60) 1.70 (0.62)*3,†2 PVT metrics  Mean RT (ms), MD (IQR)¶ 293 (87.4) 303 (94.5) 369 (133)*2,†3  Mean 1/RT, M ± SD†† 4.00 ± 0.67 3.90 ± 0.70 3.38 ± 0.77*3,†3  Fastest 10% RT (ms), MD (IQR)¶ 191 (42.0) 198 (48.7) 226 (57.0)*3,†3  Slowest 10% 1/RT, M ± SD†† 2.40 ± 0.62 2.42 ± 0.64 1.99 ± 0.63*2,†3  False Starts (FS) (%), MD (IQR)¶ 1.00 (2.05) 1.31 (1.68) 1.16 (1.34)*2,†3  Lapses 500 ms (%), MD (IQR)¶ 6.04 (8.13) 6.01 (67.33) 10.4 (10.9)*2,†3  Lapses 355 ms (%), MD (IQR)¶ 12.3 (12.0) 14.1 (15.4) 26.6 (24.9)*3,†3  Lapses 500 ms + FS (%), MD (IQR)¶ 7.60 (8.85) 7.43 (7.11) 12.1 (10.6)*2,†3  Lapses 355 ms + FS (%), MD (IQR) ¶ 13.8 (11.2) 15.8 (16.4) 28.2 (25.6)*3,†3 Analysis conducted on the 267 sailors with questionnaire, actigraphy, and PVT data. Statistical significance for differences: “1”: p < 0.05; “2”: p < 0.01; “3”: p < 0.001. *Difference between “PSQI > 9” and “PSQI ≤ 5” groups. †Difference between “PSQI > 9” and “5 < PSQI ≤ 9” groups. ††Multiple comparisons with Tukey–Kramer Honest Significant Difference (HSD) test. ¶Multiple comparisons with Dunn method for joint ranking. §Pairwise comparisons with Fisher’s Exact Test. Post hoc analysis for statistical significance with the BH-FDR controlling procedure. Open in new tab Table 3. Comparison between PSQI groups Variable . PSQI ≤ 5 (n = 37) . 5 < PSQI ≤ 9 (n = 129) . PSQI > 9 (n = 101) . Age in years, MD (IQR)¶ 24 (8) 26 (8) 25 (6) Sex (females), # (%)§ 8 (21.6%) 22 (17.1%) 34 (33.7%)†2 Enlisted personnel, # (%)§ 26 (70.3%) 99 (76.7%) 93 (92.1%)*2,†2 Watchstanders, # (%)§ 34 (91.9%) 111 (86.1%) 93 (92.1%) Daily sleep duration (h), MD (IQR)¶ 6.68 (1.64) 6.54 (1.35) 6.63 (1.77) Sleep episodes per day (#), MD (IQR)¶ 1.25 (0.53) 1.42 (0.60) 1.70 (0.62)*3,†2 PVT metrics  Mean RT (ms), MD (IQR)¶ 293 (87.4) 303 (94.5) 369 (133)*2,†3  Mean 1/RT, M ± SD†† 4.00 ± 0.67 3.90 ± 0.70 3.38 ± 0.77*3,†3  Fastest 10% RT (ms), MD (IQR)¶ 191 (42.0) 198 (48.7) 226 (57.0)*3,†3  Slowest 10% 1/RT, M ± SD†† 2.40 ± 0.62 2.42 ± 0.64 1.99 ± 0.63*2,†3  False Starts (FS) (%), MD (IQR)¶ 1.00 (2.05) 1.31 (1.68) 1.16 (1.34)*2,†3  Lapses 500 ms (%), MD (IQR)¶ 6.04 (8.13) 6.01 (67.33) 10.4 (10.9)*2,†3  Lapses 355 ms (%), MD (IQR)¶ 12.3 (12.0) 14.1 (15.4) 26.6 (24.9)*3,†3  Lapses 500 ms + FS (%), MD (IQR)¶ 7.60 (8.85) 7.43 (7.11) 12.1 (10.6)*2,†3  Lapses 355 ms + FS (%), MD (IQR) ¶ 13.8 (11.2) 15.8 (16.4) 28.2 (25.6)*3,†3 Variable . PSQI ≤ 5 (n = 37) . 5 < PSQI ≤ 9 (n = 129) . PSQI > 9 (n = 101) . Age in years, MD (IQR)¶ 24 (8) 26 (8) 25 (6) Sex (females), # (%)§ 8 (21.6%) 22 (17.1%) 34 (33.7%)†2 Enlisted personnel, # (%)§ 26 (70.3%) 99 (76.7%) 93 (92.1%)*2,†2 Watchstanders, # (%)§ 34 (91.9%) 111 (86.1%) 93 (92.1%) Daily sleep duration (h), MD (IQR)¶ 6.68 (1.64) 6.54 (1.35) 6.63 (1.77) Sleep episodes per day (#), MD (IQR)¶ 1.25 (0.53) 1.42 (0.60) 1.70 (0.62)*3,†2 PVT metrics  Mean RT (ms), MD (IQR)¶ 293 (87.4) 303 (94.5) 369 (133)*2,†3  Mean 1/RT, M ± SD†† 4.00 ± 0.67 3.90 ± 0.70 3.38 ± 0.77*3,†3  Fastest 10% RT (ms), MD (IQR)¶ 191 (42.0) 198 (48.7) 226 (57.0)*3,†3  Slowest 10% 1/RT, M ± SD†† 2.40 ± 0.62 2.42 ± 0.64 1.99 ± 0.63*2,†3  False Starts (FS) (%), MD (IQR)¶ 1.00 (2.05) 1.31 (1.68) 1.16 (1.34)*2,†3  Lapses 500 ms (%), MD (IQR)¶ 6.04 (8.13) 6.01 (67.33) 10.4 (10.9)*2,†3  Lapses 355 ms (%), MD (IQR)¶ 12.3 (12.0) 14.1 (15.4) 26.6 (24.9)*3,†3  Lapses 500 ms + FS (%), MD (IQR)¶ 7.60 (8.85) 7.43 (7.11) 12.1 (10.6)*2,†3  Lapses 355 ms + FS (%), MD (IQR) ¶ 13.8 (11.2) 15.8 (16.4) 28.2 (25.6)*3,†3 Analysis conducted on the 267 sailors with questionnaire, actigraphy, and PVT data. Statistical significance for differences: “1”: p < 0.05; “2”: p < 0.01; “3”: p < 0.001. *Difference between “PSQI > 9” and “PSQI ≤ 5” groups. †Difference between “PSQI > 9” and “5 < PSQI ≤ 9” groups. ††Multiple comparisons with Tukey–Kramer Honest Significant Difference (HSD) test. ¶Multiple comparisons with Dunn method for joint ranking. §Pairwise comparisons with Fisher’s Exact Test. Post hoc analysis for statistical significance with the BH-FDR controlling procedure. Open in new tab Discussion The PSQI Global scores of both watchstanders and dayworkers in our sample of the USN sailors (median 9 and 7, respectively) were on average 40% higher than their military and civilian peers in other occupational environments [27–32]. That is, sleep quality of sailors οn the USN ships while underway is worse than other military and civilian occupations. The consistent differences between the USN sailors and workers in other settings, both in shift- and day-workers, are indicative of the effect of the occupational stressors which are idiosyncratic to the naval operational environment [33, 34]. PSQI scores were associated with sailors’ work schedules and rank. Watchstanders and galley workers had worse sleep quality compared to dayworkers, whereas enlisted personnel had worse sleep quality compared to officers. We postulate that the latter difference can be attributed to two reasons. First, enlisted personnel live in more crowded sleeping quarters compared to officers, a factor associated with less satisfied service members and increased noise in the compartment [35, 36]. Second, the prevalence of undiagnosed sleep apnea may be higher in enlisted personnel given that obesity is higher in enlisted personnel compared to officers [37, 38]. In terms of sleep attributes, our results shed light upon the complex relation between daily sleep duration, napping (i.e. splitting daily sleep in more than one sleep episodes per day), and sleep quality in the naval environment where sleep opportunities are limited. In such environments, napping can be a useful (and perhaps the only) method to accrue sleep, but subjective sleep quality is negatively affected if sleep is split into multiple episodes. Hence, the PSQI Global score showed a U-shaped association with daily sleep duration. That is, multiple sleep episodes (i.e. increases in duration of daily sleep that are accrued in multiple sleep episodes) are associated with worse sleep quality, whereas fewer sleep episodes are associated with better sleep quality. The U-shape association between daily sleep duration and outcomes of interest is evident in studies assessing mortality [39, 40] and in a recent study on resilience in military populations [41]. Seelig and colleagues attributed their findings to underlying disorders that are not accounted for in their study. Our findings, however, show that even splitting sleep in more than one sleep episode per day (a common phenomenon in the military operational environment) may confound the effect of sleep duration. Notably, our finding that daily sleep duration was associated with PSQI Global scores does not agree with two other studies [4, 14]. We postulate two possible explanations. First, some studies assessed only the linear association between daily sleep duration and PSQI scores without considering the confounding effect of split sleep. Second, PSQI provides an estimate of average sleep over a period of time, and, therefore, it is not sensitive to daily variability as assessed by polysomnographic studies [4]. Another interesting finding was the high prevalence of “poor sleepers” in our sample (~80%) which is comparable (89%) to a sample of post-deployed ADSMs and veterans of Operation Enduring Freedom and Operation Iraqi Freedom (OIF) [42], but higher than the prevalence (48.6%) in 1,957 servicemembers across all branches reported recently [6]. The identification of poor sleepers, however, was based on the widely used criterion of a PSQI Global score greater than 5 which has not been validated on military populations [6]. Given the elevated average PSQI scores found in military personnel, the use of elevated cutoff PSQI scores is likely better suited to differentiate military personnel with sleep disorders [15]. Also, PSQI Global scores were a predictor of degraded psychomotor performance. Compared to crewmembers with PSQI scores of 9 or less, individuals with PSQI scores greater than 9 experienced slower reaction times by ~20% and greater numbers of lapses combined with false starts by approximately 33% or more. Using the PSQI cutoff score of 9, approximately 35% of our entire sample was at risk of degraded psychomotor vigilance performance. Of note, the prevalence of PSQI scores greater than 9 in our sample is doubled compared to the corresponding 18% found in a survey across all branches of the military [6]. Our findings raise a number of issues. The first issue pertains to the utility of napping for the USN sailors while underway in light of our finding that split sleep has a detrimental effect on sleep quality. Even though consolidated sleep is preferable to splitting sleep into more than one sleep episode, our studies have shown that long rest opportunities are rare at sea, especially for watchstanders [34, 43]. Napping may be the only viable method to accrue sleep in an occupational environment saturated with various duties and events not controlled by the sailor (work and operational events which cannot be planned ahead). Along these lines, the USN Comprehensive Fatigue and Endurance Management Policy (CFEMP) recommends that under ordinary conditions, underway sailors should receive either one uninterrupted 7-h period of sleep or an uninterrupted 5-h period with a 2-h nap [44]. Second, our results on sleep quality emphasize the importance of appropriate sleeping conditions in berthing compartments. Sleep-related habitability factors like environmental conditions and bedding can have detrimental effects on sailor well-being [45]. We should also consider the long-term implications of our findings. Sailors in the USN live and work in underway conditions for long periods of time, even for weeks and months at a time. A typical deployment lasts 6 months or more, and sailors frequently experience multiple deployments during their career. Chronic exposure to insufficient sleep has been associated with sleep and circadian disorders, obesity, diabetes, and other health issues [43, 46, 47]. These effects may continue even after service members retire [48]. The current study has a number of limitations which may inform future efforts. First, all our sailors were deemed to be fit for duty, but we are not aware of their actual health status. We asked, however, about what medications they were receiving, both prescription and the over-the-counter medication. Second, all ranks on the ship were not equally represented in our study sample. This diversity may have introduced a social desirability bias in the responses between different groups, for example, officers versus enlisted personnel. Also, our overall study sample was fairly large but the occupational groups differed in size. The decision to compare the occupational groups regardless of their size was based on the importance of this comparison and the fact that our study samples were, in general, representative of the actual size of the occupational groups in the ships we studied. Future efforts, however, should include larger samples of maintenance shiftworkers and galley workers. Lastly, sailor work schedules when underway are also associated with the number of sailors available for each duty/work activity, training/experience, rank, ship organizational structure and mission, organizational unit (department, division) the sailor belongs in, work hours per day, whether working on a fixed or rotating schedule, etc. Future research should further investigate these important occupational factors. Furthermore, our study has some key strengths. All data were collected in the field while sailors were conducting their underway duties on a number of ships. The demographics of the participants are representative of the USN populations of sailors in terms of age, gender, and officer/enlisted personnel ratio. Taken together, our results show the challenges sailors face when working on ships. Chronic sleep deprivation, split sleep, and deteriorated sleep quality are characteristics endemic to the naval operational environment [34, 43]. Even though widely used, further research is needed to assess sleep quality in military operational settings and the association between PSQI scores and occupational attributes of the military environment. Funding The various studies in this paper were supported by Office of the Chief of Naval Operations (OPNAV), 21st Century Sailor Office (N17), Office of Naval Research (ONR) Naval Research Program, and the Naval Medical Research Center’s, Naval Advanced Medical Development Program. Author Contributions Conception by PM. Study design, data collection, and analysis by NLS and PM. Both authors interpreted data, edited the manuscript, and approved the final draft. Disclosure Statement Financial disclosure: PM and NLS have no financial relationships relevant to this article to disclose. Non-financial disclosure: PM and NLS declare no conflict of interest. References 1. Johns MW . A new method for measuring daytime sleepiness: the Epworth sleepiness scale . Sleep. 1991 ; 14 ( 6 ): 540 – 545 . Google Scholar Crossref Search ADS PubMed WorldCat 2. Shattuck NL , et al. Psychomotor vigilance performance predicted by Epworth Sleepiness Scale scores in an operational setting with the United States Navy . J Sleep Res. 2015 ; 24 ( 2 ): 174 – 180 . Google Scholar Crossref Search ADS PubMed WorldCat 3. Matsangas P , et al. Discriminating between fatigue and sleepiness in the naval operational environment . Behav Sleep Med. 2018 ; 16 ( 5 ): 427 – 436 . Google Scholar Crossref Search ADS PubMed WorldCat 4. Buysse DJ , et al. The Pittsburgh Sleep Quality Index: a new instrument for psychiatric practice and research . Psychiatry Res. 1989 ; 28 ( 2 ): 193 – 213 . Google Scholar Crossref Search ADS PubMed WorldCat 5. Mollayeva T , et al. The Pittsburgh Sleep Quality Index as a screening tool for sleep dysfunction in clinical and non-clinical samples: a systematic review and meta-analysis . Sleep Med Rev. 2016 ; 25 : 52 – 73 . Google Scholar Crossref Search ADS PubMed WorldCat 6. Troxel WM , et al. Sleep in the Military: Promoting Healthy Sleep among U.S. Servicemembers . Santa Monica, CA : RAND ; 2015 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC 7. Backhaus J , et al. Test-retest reliability and validity of the Pittsburgh Sleep Quality Index in primary insomnia . J Psychosom Res. 2002 ; 53 ( 3 ): 737 – 740 . Google Scholar Crossref Search ADS PubMed WorldCat 8. Gliklich RE , et al. Health status in patients with disturbed sleep and obstructive sleep apnea . Otolaryngol Head Neck Surg. 2000 ; 122 ( 4 ): 542 – 546 . Google Scholar PubMed OpenURL Placeholder Text WorldCat 9. Carpenter JS , et al. Psychometric evaluation of the Pittsburgh Sleep Quality Index . J Psychosom Res. 1998 ; 45 ( 1 ): 5 – 13 . Google Scholar Crossref Search ADS PubMed WorldCat 10. Menefee LA , et al. Self-reported sleep quality and quality of life for individuals with chronic pain conditions . Clin J Pain. 2000 ; 16 ( 4 ): 290 – 297 . Google Scholar Crossref Search ADS PubMed WorldCat 11. Smith MT , et al. Sleep quality and presleep arousal in chronic pain . J Behav Med. 2000 ; 23 ( 1 ): 1 – 13 . Google Scholar Crossref Search ADS PubMed WorldCat 12. Gentili A , et al. Test-retest reliability of the Pittsburgh Sleep Quality Index in nursing home residents . J Am Geriatr Soc. 1995 ; 43 ( 11 ): 1317 – 1318 . Google Scholar Crossref Search ADS PubMed WorldCat 13. Beaudreau SA , et al. Validation of the Pittsburgh Sleep Quality Index and the Epworth Sleepiness Scale in older black and white women . Sleep Med. 2012 ; 13 ( 1 ): 36 – 42 . Google Scholar Crossref Search ADS PubMed WorldCat 14. Grandner MA , et al. Criterion validity of the Pittsburgh Sleep Quality Index: Investigation in a non-clinical sample . Sleep Biol Rhythms. 2006 ; 4 ( 2 ): 129 – 139 . Google Scholar Crossref Search ADS PubMed WorldCat 15. Matsangas P , Mysliwiec V. The utility of the Pittsburgh Sleep Quality Index in US military personnel . Mil Psychol . 2018 ; 30 ( 4 ): 360 – 369 . Google Scholar Crossref Search ADS WorldCat 16. Babson KA , et al. Sleep quality among U.S. military veterans with PTSD: a factor analysis and structural model of symptoms . J Trauma Stress. 2012 ; 25 ( 6 ): 665 – 674 . Google Scholar Crossref Search ADS PubMed WorldCat 17. Department of the Navy . Standard Organization and Regulations of the U.S. Navy—OPNAV Instruction 3120.32D . Washington, DC : Office of the Chief of Naval Operations ; 2012 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC 18. Ancoli-Israel S , et al. The SBSM guide to actigraphy monitoring: clinical and research applications . Behav Sleep Med. 2015 ; 13 Suppl 1 : S4 – S38 . Google Scholar Crossref Search ADS PubMed WorldCat 19. Morgenthaler T , et al. Practice parameters for the use of actigraphy in the assessment of sleep and sleep disorders: an update for 2007 . Sleep. 2007 ; 30 ( 4 ): 519 – 529 . Google Scholar Crossref Search ADS PubMed WorldCat 20. Meltzer LJ , et al. Direct comparison of two new actigraphs and polysomnography in children and adolescents . Sleep. 2012 ; 35 ( 1 ): 159 – 166 . Google Scholar PubMed OpenURL Placeholder Text WorldCat 21. Dinges DF , et al. Microcomputer analyses of performance on a portable, simple visual RT task during sustained operations . Behav Res Methods Instrum Comput . 1985 ; 17 ( 6 ): 652 – 655 . Google Scholar Crossref Search ADS WorldCat 22. Matsangas P , et al. Agreement between the 3-minute Psychomotor Vigilance Task (PVT) embedded in a wrist-worn device and the laptop-based PVT . Proc Hum Factors Ergon Soc Annu Meet . 2018 ; 62 ( 1 ): 666 – 670 . Google Scholar Crossref Search ADS WorldCat 23. Matsangas P , et al. Preliminary validation study of the 3-min wrist-worn psychomotor vigilance test . Behav Res Methods. 2017 ; 49 ( 5 ): 1792 – 1801 . Google Scholar Crossref Search ADS PubMed WorldCat 24. Broderick JE , et al. Pittsburgh and Epworth sleep scale items: accuracy of ratings across different reporting periods . Behav Sleep Med. 2013 ; 11 ( 3 ): 173 – 188 . Google Scholar Crossref Search ADS PubMed WorldCat 25. Benjamini Y , et al. Controlling the false discovery rate: A practical and powerful approach to multiple testing . J R Stat Soc Ser B Stat Methodol . 1995 ; 57 : 289 – 300 . Google Scholar OpenURL Placeholder Text WorldCat 26. DoD . 2015 Demographics: Profile of the Military Community . www.militaryonesource.mil. Accessed April 7, 2017 . 27. Waage S , et al. Shift work disorder among oil rig workers in the North Sea . Sleep. 2009 ; 32 ( 4 ): 558 – 565 . Google Scholar Crossref Search ADS PubMed WorldCat 28. Shattuck NL , et al. Improving work and rest patterns of military personnel in operational settings with frequent unplanned events . Proc Hum Factors Ergon Soc Annu Meet . 2018 ; 62 ( 1 ): 772 – 776 . Google Scholar Crossref Search ADS WorldCat 29. Neylan TC , et al. Critical incident exposure and sleep quality in police officers . Psychosom Med. 2002 ; 64 ( 2 ): 345 – 352 . Google Scholar Crossref Search ADS PubMed WorldCat 30. Lajoie P , et al. A cross-sectional study of shift work, sleep quality and cardiometabolic risk in female hospital employees . BMJ Open. 2015 ; 5 ( 3 ): e007327 . Google Scholar Crossref Search ADS PubMed WorldCat 31. van Mark A , et al. The impact of shift work induced chronic circadian disruption on IL-6 and TNF-α immune responses . J Occup Med Toxicol . 2010 ; 5 : 18 . Google Scholar Crossref Search ADS PubMed WorldCat 32. De Almondes KM , et al. Sleep quality and daily lifestyle in workers with different working hours . Biol Rhythm Res . 2011 ; 42 ( 3 ): 231 – 245 . Google Scholar Crossref Search ADS WorldCat 33. Comperatore CA , et al. Enduring the shipboard stressor complex: a systems approach . Aviat Space Environ Med. 2005 ; 76 ( 6 Suppl ): B108 – B118 . Google Scholar PubMed OpenURL Placeholder Text WorldCat 34. Shattuck NL , et al. Sleep and fatigue issues in military operations. In: Vermetten E, Germain A, Neylan T, eds. Sleep and Combat related PTSD . New York, NY : Springer ; 2018 : 69 – 76 . Google Scholar Crossref Search ADS Google Preview WorldCat COPAC 35. Wilcove GL. 2002 Navy Quality of Life (QOL) survey: Shipboard life results. Millington, TN : Bureau of Naval Personnel ; 2006 . NPRST-AB-06-2. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC 36. Matsangas P , et al. Exploring sleep-related habitability issues in berthing spaces on U.S. Navy ships . Proc Hum Factors Ergon Soc Annu Meet . 2017 ; 61 ( 1 ): 450 – 454 . Google Scholar Crossref Search ADS WorldCat 37. Meadows SO , et al. 2015 Department of Defense Health Related Behaviors Survey (HRBS) . Santa Monica, CA : RAND Corporation ; 2018 . Google Scholar Crossref Search ADS Google Preview WorldCat COPAC 38. Reyes-Guzman CM , et al. Overweight and obesity trends among active duty military personnel: a 13-year perspective . Am J Prev Med. 2015 ; 48 ( 2 ): 145 – 153 . Google Scholar Crossref Search ADS PubMed WorldCat 39. Kripke DF , et al. Mortality associated with sleep duration and insomnia . Arch Gen Psychiatry. 2002 ; 59 ( 2 ): 131 – 136 . Google Scholar Crossref Search ADS PubMed WorldCat 40. Kripke DF , et al. Mortality related to actigraphic long and short sleep . Sleep Med . 2011 ; 12 ( 1 ): 28 – 33 . Google Scholar Crossref Search ADS PubMed WorldCat 41. Seelig AD , et al. Sleep and health resilience metrics in a large military cohort . Sleep. 2016 ; 39 ( 5 ): 1111 – 1120 . Google Scholar Crossref Search ADS PubMed WorldCat 42. Plumb TR , et al. Sleep disturbance is common among servicemembers and veterans of Operations Enduring Freedom and Iraqi Freedom . Psychol Serv. 2014 ; 11 ( 2 ): 209 – 219 . Google Scholar Crossref Search ADS PubMed WorldCat 43. Shattuck NL , et al. The role of sleep in human performance and well-being. In: Matthews MD, Schnyer D, eds. Human Performance Optimization: The Science and Ethics of Enhancing Human Capabilities . New York, NY : Oxford University Press ; 2019 : 200 – 233 . Google Scholar Crossref Search ADS Google Preview WorldCat COPAC 44. Department of the Navy . Comprehensive fatigue and endurance management policy (COMNAVSURFPAC/COMNAVSURFLANT Instruction 3120.2).https://federalnewsnetwork.com/wp-content/uploads/2017/12/113017_navyrestpolicy.pdf. 45. Matsangas P , Shattuck NL. Habitability in berthing compartments and well-being of sailors working on United States Navy surface ships . Hum Factors . 2020 . Google Scholar OpenURL Placeholder Text WorldCat 46. Mysliwiec V , et al. An unusual circadian rhythm in an active duty service member . Sleep Biol Rhythms . 2015 ; 14 ( 1 ): 113 – 115 . Google Scholar Crossref Search ADS WorldCat 47. Good CH , et al. Sleep in the United States Military . Neuropsychopharmacology. 2020 ; 45 ( 1 ): 176 – 191 . doi:10.1177/0018720820906050 Google Scholar Crossref Search ADS PubMed WorldCat 48. Monk TH , et al. Shiftworkers report worse sleep than day workers, even in retirement . J Sleep Res. 2013 ; 22 ( 2 ): 201 – 208 . Google Scholar Crossref Search ADS PubMed WorldCat Published by Oxford University Press on behalf of Sleep Research Society (SRS) 2020. This work is written by (a) US Government employee(s) and is in the public domain in the US. Published by Oxford University Press on behalf of Sleep Research Society (SRS) 2020.

Bélanger, Jocelyn, J;Raafat, Karima, A;Nisa, Claudia, F;Schumpe, Birga, M

doi: 10.1093/sleep/zsaa107pmid: 32474581

Abstract Study Objectives The present research examines the relationship between people’s frequent involvement in an activity they like and find important (i.e., a passion) and the quality of their sleep. Research on the dualistic model of passion has widely documented the relationship between individuals’ type of passion—harmonious versus obsessive—and the quality of their mental and physical health. However, research has yet to examine the relationship between passion and sleep quality. Building on prior research has shown that obsessive (vs harmonious) passion is related to depressive mood symptoms—an important factor associated with sleep problems—we hypothesized that obsessive passion would be associated with overall worse sleep quality, whereas harmonious passion would predict better sleep quality. Methods A sample of 1,506 Americans filled out an online questionnaire on sleep habits and passion. Sleep quality was measured using the Pittsburgh Sleep Quality Index. Hierarchical linear regressions and mediation analyses were carried out with results confirming our hypotheses. Results Obsessive passion for an activity was associated with worse sleep quality, whereas harmonious passion was associated with better sleep quality, adjusting for demographics, the type of passionate activity and its self-reported importance, alcohol and tobacco consumption, BMI, self-reported health, and diagnosed health conditions. The relationship between both types of passion and sleep quality was mediated by depressive mood symptoms. Conclusions Our study presents evidence of a strong relationship between sleep quality and passion, opening the door for future research to create new interventions to improve people’s sleep and, consequently, their well-being. sleep quality, dualistic model of passion, activity regulation, Pittsburgh Sleep Quality Index Statement of Significance Despite the significant amount of literature connecting sleep quality to mental and physical health, research has thus far been silent on the relationship between people’s passion for an activity (e.g., playing music, sports, and relationships) and sleep quality. This is an important gap in the research because more than 75% of people are highly passionate about a given activity and will invest an average of 9.16 h a week doing it. This study (N = 1,506) shows that passion for an activity is associated with sleep quality as measured by the Pittsburgh Sleep Quality Index. Obsessive passion for an activity was associated with worse sleep quality, whereas harmonious passion was associated with better sleep quality. These relationships were mediated by depressive mood symptoms. Introduction Sleep constitutes a vital component of psychological well-being. Lack of adequate sleep can affect levels of overall cognitive functioning [1], motor performance [2], and productivity [3]. A large body of research also suggests that sleep impacts an individual’s overall mental health. For example, some studies have shown an association between increased rumination and depressive mood and poor sleep quality [4, 5]. Others have shown that people’s ability to regulate negative emotions [6, 7], their level of impulsivity to negative stimuli [8], and their vulnerability to PTSD, anxiety [9–12], and substance abuse disorders [13] are all associated with poor sleep quality. Aside from the clear association between sleep and mental health, the lack of a good night’s sleep also has conspicuous effects on one’s physical health. Specifically, research has shown that sleep problems are associated with a host of physical conditions such as gastrointestinal problems, migraines, allergies, arthritis, and metabolic disorders [13]. Furthermore, researchers have reported strong correlations between sleep and kidney function [14], shorter sleep durations and hypertension [15], as well as sleep disorders and an increased risk of experiencing a stroke [16]. Despite the significant amount of literature that documents the aforementioned relationships, research has not looked into the influence of people’s passionate activities on sleep quality. Having a passion is an important source of meaning in people’s lives: it constitutes an integral part of individuals’ identities, impacting their behaviors, feelings, thoughts, interpersonal relationships, performance, and health [17, 18]. People spend decades, sometimes even a lifetime, developing and perfecting their passions; research indicates that more than 75% of people are highly passionate about a given activity and will invest an average of 9.16 h a week engaging in it [19]. Since passion makes up such a big part of an individual’s day, an extensive body of research has burgeoned in the last two decades, producing over 600 scientific papers [20]. Despite the vast literature that exists, research thus far has focused exclusively on the effects of passion when people are awake and going about their busy day. What if, however, passion had an influence beyond those hours and affect individuals’ ability to sleep? Addressing this question, the present research examines whether passion for an activity is associated with people’s quality of sleep and the psychological mechanism explaining this relationship. As we elucidate in the next section, not all passions are created equal and some passions could potentially be beneficial, and others detrimental, for sleep. The Dualistic Model of Passion Vallerand et al. [21] define passion as a “strong inclination toward an activity that people like, that they find important, and in which they invest time and energy” (p. 756). Previous research has proposed a dualistic model of passion, which distinguishes between two types of passion: harmonious and obsessive. These two types of passion can be differentiated in terms of how passionate activity is regulated with other life domains. Harmonious passion (HP) refers to a strong desire to engage in one’s favorite activity, whereby activity engagement remains under the person’s control [17, 18]. This means that harmoniously passionate individuals freely decide to engage in this particular activity, as there are no contingencies of self-esteem or self-worth attached to it [22]. As a result, harmoniously passionate individuals have the ability to balance their passionate activity with other life domains, thereby preventing conflicts between the two [21], which is conducive to experiencing more positive, and less negative, emotions [19, 23]. People who are obsessively passionate (OP) also love their activity and are as committed to it as harmoniously passionate individuals; however, their activity is beyond their control: individuals feel pressured to pursue it continuously [17]. Research [17] has shown that this lack of control “results from a controlled internalization of the activity into one’s identity” (p. 102) because the activity is attached to contingencies of self-worth [22]. This internalization and lack of control eventually lead to rigid engagement in the activity, as well as goal-conflict [17]. Consequently, for obsessively passionate people, the balance between passionate activity and other life domains becomes disproportionate, eventually leading to less positive emotions and depressive mood symptoms [19, 23, 24]. A large body of evidence has documented the relationship between the type of passion and health outcomes. For example, research has shown that HP is positively correlated with positive emotions [19], perceived vitality [25], feelings of flow and concentration during engagement in the activity [21, 26, 27], and hedonic and eudaimonic well-being [28–30]. OP, on the other hand, has been shown to be correlated with negative emotions [19, 23], basic need frustration [31, 32], lack of vitality [25], and rumination [21, 33, 34]. Research has also found that the passion scale predicts risky behavior in sports, rigid task engagement, and chronic injuries. For instance, individuals with OP (but not HP) for cycling continue their activity even under harsh conditions (e.g., subzero temperature and icy road conditions) [21]. Likewise, in a study with ballet dancers, OP was positively correlated with continuing to dance despite being injured, whereas HP was unrelated to such a phenomenon [35]. OP is also related to poor interpersonal relationships [36], self-neglect [28], emotional exhaustion/burnout [37], increased work-related stress, burnout [38], and depressive mood symptoms [24, 39]. The latter variable is of special interest to this research given that it has been associated with poor sleep quality [9, 11, 12]. The Present Research The present study aims to fill the gaps in both the passion and sleep literatures by examining the relationship between individuals’ type of passion and their ability to sleep. We propose that looking into individuals’ passions might be a gateway to understanding sleep quality. Specifically, we hypothesize that because individuals with high levels of OP tend to experience depressive mood symptoms [24, 39] and that such emotional states are associated with poor sleep [4, 5], they will report overall worse sleep quality indicated by higher scores on the Pittsburgh Sleep Quality Index (PSQI). On the other hand, because individuals with high levels of HP are less prone to experiencing such a negative emotional state, we predict they will report overall better sleep quality, indicated by lower scores on the PSQI. We posit that these associations will hold adjusting for the effects of other variables known to affect sleep, such as age, gender, body mass index (BMI), a number of diagnosed health conditions, and lifestyle choices (i.e., alcohol and tobacco use) [40–47]. Methods Participants The authors assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008. The study was approved by the Research Ethics Committee of New York University Abu Dhabi (no. 121-2018). One thousand seven hundred and forty-three Americans were recruited via a panel service to complete an anonymous online questionnaire on sleep habits. Participants were given the definition of passion and indicated whether they were passionate about an activity in particular (Yes vs No). Two hundred and thirty-seven participants (13% of the total sample) reported not having a passion, thus leaving 1,506 Americans (761 women; Mage = 42.12 years, SDage = 14.38 years) for our analyses. Written consent was obtained from the participants. The sample comprised of multiple ethnicities, including American Indian/Alaskan Native (0.4%), Asian (6.4%), black (9.9%), Latino (8.6%), Mixed/Native Hawaiian or Pacific Islander (2.7%), and white (72%). In terms of employment, 70.7% of participants indicated being currently employed. Procedures and measures Participants completed a short survey measuring their passion for an activity and quality of sleep. Written consent was obtained. Other measures related to alcohol and tobacco consumption, self-reported health, BMI, and diagnosed health problems were also included to adjust for their influence in our analyses given that they have shown to be associated with sleep quality [41, 48–52]. Passion Participants were instructed to list an “activity that you like, that is important to you, and in which you invest a significant amount of time on a regular basis” (the definition of passion) and then asked to respond to the passion scale [21] for this activity. The passion scale consists of 2 six-item subscales assessing harmonious (e.g., “My activity is well integrated in my life,” “My passion allows me to live a variety of experiences,” M = 5.47, SD = 1.05, α = .86) and obsessive (e.g., “If I could, I would dedicate myself only to my activity,” “I have difficulties controlling my urge to do my passion,” M = 3.47, SD = 1.49, α = .84) passion. Participants rated their agreement for each of these items on a seven-point scale ranging from 1 (not agree at all) to 7 (very strongly agree). Two coders grouped participants’ activities into categories used in prior research [31] (see Table 1 for a list of the activities listed by participants). Table 1. Types of Passionate Activities and the Percentage of Participants Engaging in Each Activity (N = 1,501) Activity categories . Examples of activity . Percentage of participants who engaged in activity . 1. Active leisure Baking, gardening, fishing 26.6 2. Individual sports/physical activity Weight-lifting, jogging, swimming 7.7 3. Work/education Part-time work, studying, teaching 11.8 4. Interpersonal relationships Taking care of friends, family, animals 21.5 5. Active arts Drawing, dancing, photography 9.1 6. Reading/writing Reading or writing a novel, blogging 8.1 7. Passive leisure Listening to music, watching a movie 2.1 8. Active music Playing an instrument, composing 6.7 9. Team sports Playing basketball, baseball 1.2 10. Religion Attending religious services, preaching 2.2 11. Other Working toward being fit and healthy, achieving self-growth 3.1 Activity categories . Examples of activity . Percentage of participants who engaged in activity . 1. Active leisure Baking, gardening, fishing 26.6 2. Individual sports/physical activity Weight-lifting, jogging, swimming 7.7 3. Work/education Part-time work, studying, teaching 11.8 4. Interpersonal relationships Taking care of friends, family, animals 21.5 5. Active arts Drawing, dancing, photography 9.1 6. Reading/writing Reading or writing a novel, blogging 8.1 7. Passive leisure Listening to music, watching a movie 2.1 8. Active music Playing an instrument, composing 6.7 9. Team sports Playing basketball, baseball 1.2 10. Religion Attending religious services, preaching 2.2 11. Other Working toward being fit and healthy, achieving self-growth 3.1 Open in new tab Table 1. Types of Passionate Activities and the Percentage of Participants Engaging in Each Activity (N = 1,501) Activity categories . Examples of activity . Percentage of participants who engaged in activity . 1. Active leisure Baking, gardening, fishing 26.6 2. Individual sports/physical activity Weight-lifting, jogging, swimming 7.7 3. Work/education Part-time work, studying, teaching 11.8 4. Interpersonal relationships Taking care of friends, family, animals 21.5 5. Active arts Drawing, dancing, photography 9.1 6. Reading/writing Reading or writing a novel, blogging 8.1 7. Passive leisure Listening to music, watching a movie 2.1 8. Active music Playing an instrument, composing 6.7 9. Team sports Playing basketball, baseball 1.2 10. Religion Attending religious services, preaching 2.2 11. Other Working toward being fit and healthy, achieving self-growth 3.1 Activity categories . Examples of activity . Percentage of participants who engaged in activity . 1. Active leisure Baking, gardening, fishing 26.6 2. Individual sports/physical activity Weight-lifting, jogging, swimming 7.7 3. Work/education Part-time work, studying, teaching 11.8 4. Interpersonal relationships Taking care of friends, family, animals 21.5 5. Active arts Drawing, dancing, photography 9.1 6. Reading/writing Reading or writing a novel, blogging 8.1 7. Passive leisure Listening to music, watching a movie 2.1 8. Active music Playing an instrument, composing 6.7 9. Team sports Playing basketball, baseball 1.2 10. Religion Attending religious services, preaching 2.2 11. Other Working toward being fit and healthy, achieving self-growth 3.1 Open in new tab The passion scale has been shown to be valid and reliable with respect to a variety of activities [21] and contexts [53–55]. In this study, a factor analysis with maximum likelihood and oblimin rotation was performed on the 12 items of the passion scale to confirm its two-factor structure. Results supported this prediction with eigenvalues of 4.38 and 2.65 explaining 32% and 18% of the variance, respectively. All factors loadings were above .57 and there were no cross-loadings. Pittsburgh Sleep Quality Index Participants rated the quality of their sleep in the past month using the validated PSQI [56] (Sum = 6.97, SD = 4.15). The PSQI includes 19 items used to assess seven different components of sleep: self-reported sleep quality, sleep latency, sleep duration, habitual sleep efficiency, sleep disturbances, use of sleeping medication, and daytime dysfunction. Participants responded to these items either using a four-point scale (e.g., “During the past month, how often have you had trouble sleeping because you cannot get to sleep within 30 minutes?,” 0 = Not during the past month; 1 = Less than once a week; 2 = Once or twice a week; 3 = Three or more times a week) or by indicating a specific period of time (e.g., “How long (in minutes) has it taken you to fall asleep each night?”). Scores from each item are summed to give a global sleep score, with higher scores indicating worse sleep quality [56]. The PSQI has significantly high internal consistency and has been shown to be a reliable measure for distinguishing between “good” and “poor” sleepers as well as assessing sleep quality and disturbance in general clinical or psychiatric settings. Alcohol and tobacco consumption Alcohol and tobacco consumption are factors known to reduce sleep quality [40–43], they were thus included in our model. Given that they were included in the analysis as covariates to take into consideration, but not as a primary focus of interest, we restricted the measurement of both variables to participants’ consumption of these products in the past month (Yes vs No). This binary approach to the measurement of smoking and drinking is a frequently used methodological choice [57–60]. In the sample, 50.1% and 28.1% of participants indicated they had consumed alcohol and tobacco in the past month, respectively. General health Prior research has shown that self-reported health is related to poor sleep quality [44]. To control for such variable, participants were asked to rate their general health using the standard version of the general self-rated health survey [61] (GSRH). The GSRH (M = 3.21, SD = .98) is a one-item survey that asks participants to rate their overall health (“In general, would you say your health is”) using a five-point scale ranging from 1 (poor) to 5 (excellent). The GSRH has previously been shown to be a reliable and valid measure of overall health status [61]. Diagnosed health conditions Participants were asked to indicate (no = 0, yes = 1) if they have ever been diagnosed with any one of 14 health problems presented to them using the question: “Have you ever been diagnosed with any of the following health problems?” The 14 health problems included were allergies, arthritis, cancer, chronic fatigue, chronic pulmonary disease or asthma, cirrhosis, gastrointestinal disease, heart disease, high cholesterol, kidney disease, stroke, tuberculosis, type 1 diabetes, and type 2 diabetes. Previous research has shown that these conditions either impair or are related to conditions that impair sleep [45–47]. Body mass index Participants were asked to indicate their height and weight. Each participant’s BMI was then calculated using the standard formula BMI = kg/m2, in which kg is the individual’s weight in kilograms and m2 is their height in meters squared (M = 27.33, SD = 6.74). This served as a control variable given the existing relationship between BMI and poor sleep quality [62–65]. Depressive mood symptoms The mediating variable in our analysis was measured using seven items taken from the Patient Health Questionnaire (M = 1.81, SD = .93, α = .90) which measured participants’ depressive symptomatology based on the DSM-IV diagnosis. Participants reported to what extent they were troubled by different problems such as “Feeling down, depressed or hopeless” and “Poor appetite or overeating.” Participants provided their responses on a five-point scale from 1 (not at all) to 5 (extremely). Results Hierarchical linear regressions were conducted to examine the relationship between passion and the PSQI, controlling for demographics, alcohol and tobacco consumption, BMI, self-reported health, and diagnosed health conditions. All variables were entered simultaneously. Preliminary analyses yielded skewness and kurtosis values ranging from –2 to 2, thus indicating univariate normality for the continuous variables [66]. To examine the possibility of multicollinearity, we calculated the variance inflation factor (VIF) and tolerance for the variables in the model. The VIF, which should be close to 1 and less than 10 [67–69], ranged from 1.02 to 1.57, thus indicating no multicollinearity. A similar conclusion was drawn from the tolerance values which should not be lower than .01 [70], they were all above 0.63. Five cases were removed because they had residuals that were three or more standard deviations away from the mean, leaving 1,501 cases for analysis. Moreover, in line with Tabachnick and Fidell’s [71] recommendations, an inspection of the residuals from the regression analysis showed that various assumptions of linearity, normality, and heteroscedasticity were met. We display means, standard deviations, and correlations for all measures in Supplementary File—Tables S1 and S2. We entered age, gender, employment, ethnicity, and the type of passionate activity in Step 1. In Step 2, we included measures related to BMI, alcohol and tobacco use, general and diagnosed health, and in Step 3, we included OP and HP. Step 1 explained a significant amount of variance, F(18, 1482) = 3.66, p < .001, ΔR2 = .04. The addition of the health-related variables in Step 2 significantly increased the explained variance, F(36, 1464) = 21.00, p < .001, ΔR2 = .19 and so did adding OP and HP in Step 3, F(38, 1462) = 10.19, p < .001, ΔR2 = .01. The full model predicted 25% of the variance (see Table 2). Table 2. Results of Hierarchical Multiple Regression Predicting the Pittsburgh Sleep Quality Index (N = 1,501) . F . R 2 . ΔR2 . Step 1—Demographics 3.66*** .04 .04 Step 2—Health-related measures 12.77*** .23 .19 Step 3—OP and HP 12.79*** .25 .01 . F . R 2 . ΔR2 . Step 1—Demographics 3.66*** .04 .04 Step 2—Health-related measures 12.77*** .23 .19 Step 3—OP and HP 12.79*** .25 .01 ***p < .001. Open in new tab Table 2. Results of Hierarchical Multiple Regression Predicting the Pittsburgh Sleep Quality Index (N = 1,501) . F . R 2 . ΔR2 . Step 1—Demographics 3.66*** .04 .04 Step 2—Health-related measures 12.77*** .23 .19 Step 3—OP and HP 12.79*** .25 .01 . F . R 2 . ΔR2 . Step 1—Demographics 3.66*** .04 .04 Step 2—Health-related measures 12.77*** .23 .19 Step 3—OP and HP 12.79*** .25 .01 ***p < .001. Open in new tab Step 1 Demographic variables Results showed a significant negative relationship between age and PSQI scores, indicating that younger participants in our sample were more likely to report worse sleep quality, B = −.03, SE = .008, t(1462) = −4.13, p < .001. This finding differs from previous studies showing a negative relationship between age and sleep quality [72] and may be an idiosyncrasy of our sample. On the other hand, there was a positive relationship between gender and PSQI scores, indicating that women were more likely to experience worse sleep quality than men, B = .67, SE = .19, t(1462) = 3.40, p = .001; a finding that is consistent with previous research showing that women are more likely to experience insomnia, longer sleep latency, and more sleepiness than men [73]. Ethnicity did not predict PSQI scores. Type of passionate activity Results showed a positive relationship between religion and PSQI scores, indicating that individuals who were passionate about religion were more likely to experience worse sleep quality, B = 1.86, SE = .67, t(1462) = 2.78, p < .01—an interesting finding given the little amount of research on this topic [74, 75]. No other activity type showed a significant correlation with sleep quality (all p values >.05). Step 2 Health Self-rated health as measured by the GSRH negatively predicted PSQI scores, indicating that individuals with worse general health experienced worse sleep quality, B = −1.41, SE = .11, t(1462) = −12.42, p < .001. Both allergies, B = .89, SE = .22, t(1462) = 4.09, p < .001, and arthritis, B = .75, SE = .32, t(1462) = 2.32, p < .05, were positively associated with PSQI scores. This is in line with previous research showing strong associations between allergies and sleep disturbances [76–78], as well as research documenting the connection between arthritis and poor sleep quality due to pain [79]. Both smoking, B = .76, SE = .21, t(1462) = 3.47, p = .001, and alcohol consumption, B = .48, SE = .19, t(1462) = 2.51, p < .05, positively predicted PSQI scores, indicating that individuals who had either smoked or consumed alcohol in the past month were more likely to experience worse sleep quality. None of the other health-related variables showed significant correlations with sleep quality (all p values >.05). Step 3 Passion In line with our predictions, OP was associated with a greater score on the PSQI, indicating worse sleep quality, B = .30, SE = .06, t(1462) = 4.47, p < .001, whereas HP was shown to be negatively, although marginally, related to it, indicating better sleep, B = −.17, SE = .09, t(1462) = −1.80, p = .07. These effects held adjusting for the influence of the control variables in the model (see Table 3 for all regression coefficients).1 Table 3. Summary of Multiple Regression Analyses for Variables Predicting PSQI (N =1,501) Variable . B . SE . T . p . 1. Obsessive passion (OP) .30 .06 4.47*** <.001 2. Harmonious passion (HP) −.17 .09 −1.80 .07 3. Age −.03 .008 −4.13*** <.001 4. Employment −.28 .22 −1.28 .19 5. Gender .67 .19 3.40*** .001 6. American Indian/Alaskan Native† −.23 1.49 −.15 .87 7. Asian −.27 .40 −.68 .49 8. Black .15 .32 .47 .63 9. Latino .02 .34 .08 .93 10. Mixed/Native Hawaiian or Pacific Islander −.55 .58 −.96 .33 11. Body mass index (kg/m2) .01 .01 .87 .38 12. General health −1.41 .11 −12.42*** <.001 13. Allergies .89 .22 4.09*** <.001 14. Alcohol consumption .48 .19 2.51* .01 15. Arthritis .75 .32 2.32* .02 16. Cancer −.27 .54 −.50 .61 17. Chronic fatigue .72 .56 1.28 .20 18. Chronic pulmonary disease or asthma .40 .46 .86 .38 19. Cirrhosis −2.34 2.74 −.85 .39 20. Gastrointestinal disease .19 .42 .44 .65 21. Heart disease .26 .58 .44 .65 22. High cholesterol .26 .27 .96 .33 23. Kidney disease −.92 .96 −.96 .33 24. Smoking (tobacco or other) .76 .21 3.47*** .001 25. Stroke 1.35 .85 1.58 .11 26. Tuberculosis 3.24 2.13 1.52 .12 27. Type 1 diabetes 1.30 .94 1.37 .16 28. Type 2 diabetes −.06 .40 −.15 .87 29. Individual sports/physical activity‡ .41 .38 1.06 .28 30. Work/education .02 .33 .06 .95 31. Interpersonal relationships .49 .28 1.76 .07 32. Active arts −.08 .36 −.22 .81 33. Reading/writing .33 .38 .87 .38 34. Passive leisure .83 .66 1.24 .21 35. Active music .71 .41 1.74 .08 36. Team sports 1.48 .88 1.67 .09 37. Religion 1.86 .67 2.78** .006 38. Other .88 .57 1.55 .12 Adjusted R 2 .23 F 12.71*** Variable . B . SE . T . p . 1. Obsessive passion (OP) .30 .06 4.47*** <.001 2. Harmonious passion (HP) −.17 .09 −1.80 .07 3. Age −.03 .008 −4.13*** <.001 4. Employment −.28 .22 −1.28 .19 5. Gender .67 .19 3.40*** .001 6. American Indian/Alaskan Native† −.23 1.49 −.15 .87 7. Asian −.27 .40 −.68 .49 8. Black .15 .32 .47 .63 9. Latino .02 .34 .08 .93 10. Mixed/Native Hawaiian or Pacific Islander −.55 .58 −.96 .33 11. Body mass index (kg/m2) .01 .01 .87 .38 12. General health −1.41 .11 −12.42*** <.001 13. Allergies .89 .22 4.09*** <.001 14. Alcohol consumption .48 .19 2.51* .01 15. Arthritis .75 .32 2.32* .02 16. Cancer −.27 .54 −.50 .61 17. Chronic fatigue .72 .56 1.28 .20 18. Chronic pulmonary disease or asthma .40 .46 .86 .38 19. Cirrhosis −2.34 2.74 −.85 .39 20. Gastrointestinal disease .19 .42 .44 .65 21. Heart disease .26 .58 .44 .65 22. High cholesterol .26 .27 .96 .33 23. Kidney disease −.92 .96 −.96 .33 24. Smoking (tobacco or other) .76 .21 3.47*** .001 25. Stroke 1.35 .85 1.58 .11 26. Tuberculosis 3.24 2.13 1.52 .12 27. Type 1 diabetes 1.30 .94 1.37 .16 28. Type 2 diabetes −.06 .40 −.15 .87 29. Individual sports/physical activity‡ .41 .38 1.06 .28 30. Work/education .02 .33 .06 .95 31. Interpersonal relationships .49 .28 1.76 .07 32. Active arts −.08 .36 −.22 .81 33. Reading/writing .33 .38 .87 .38 34. Passive leisure .83 .66 1.24 .21 35. Active music .71 .41 1.74 .08 36. Team sports 1.48 .88 1.67 .09 37. Religion 1.86 .67 2.78** .006 38. Other .88 .57 1.55 .12 Adjusted R 2 .23 F 12.71*** †White acted as the reference group. ‡Active leisure acted as the reference group. *p < .05; **p < .01; ***p < .001. Open in new tab Table 3. Summary of Multiple Regression Analyses for Variables Predicting PSQI (N =1,501) Variable . B . SE . T . p . 1. Obsessive passion (OP) .30 .06 4.47*** <.001 2. Harmonious passion (HP) −.17 .09 −1.80 .07 3. Age −.03 .008 −4.13*** <.001 4. Employment −.28 .22 −1.28 .19 5. Gender .67 .19 3.40*** .001 6. American Indian/Alaskan Native† −.23 1.49 −.15 .87 7. Asian −.27 .40 −.68 .49 8. Black .15 .32 .47 .63 9. Latino .02 .34 .08 .93 10. Mixed/Native Hawaiian or Pacific Islander −.55 .58 −.96 .33 11. Body mass index (kg/m2) .01 .01 .87 .38 12. General health −1.41 .11 −12.42*** <.001 13. Allergies .89 .22 4.09*** <.001 14. Alcohol consumption .48 .19 2.51* .01 15. Arthritis .75 .32 2.32* .02 16. Cancer −.27 .54 −.50 .61 17. Chronic fatigue .72 .56 1.28 .20 18. Chronic pulmonary disease or asthma .40 .46 .86 .38 19. Cirrhosis −2.34 2.74 −.85 .39 20. Gastrointestinal disease .19 .42 .44 .65 21. Heart disease .26 .58 .44 .65 22. High cholesterol .26 .27 .96 .33 23. Kidney disease −.92 .96 −.96 .33 24. Smoking (tobacco or other) .76 .21 3.47*** .001 25. Stroke 1.35 .85 1.58 .11 26. Tuberculosis 3.24 2.13 1.52 .12 27. Type 1 diabetes 1.30 .94 1.37 .16 28. Type 2 diabetes −.06 .40 −.15 .87 29. Individual sports/physical activity‡ .41 .38 1.06 .28 30. Work/education .02 .33 .06 .95 31. Interpersonal relationships .49 .28 1.76 .07 32. Active arts −.08 .36 −.22 .81 33. Reading/writing .33 .38 .87 .38 34. Passive leisure .83 .66 1.24 .21 35. Active music .71 .41 1.74 .08 36. Team sports 1.48 .88 1.67 .09 37. Religion 1.86 .67 2.78** .006 38. Other .88 .57 1.55 .12 Adjusted R 2 .23 F 12.71*** Variable . B . SE . T . p . 1. Obsessive passion (OP) .30 .06 4.47*** <.001 2. Harmonious passion (HP) −.17 .09 −1.80 .07 3. Age −.03 .008 −4.13*** <.001 4. Employment −.28 .22 −1.28 .19 5. Gender .67 .19 3.40*** .001 6. American Indian/Alaskan Native† −.23 1.49 −.15 .87 7. Asian −.27 .40 −.68 .49 8. Black .15 .32 .47 .63 9. Latino .02 .34 .08 .93 10. Mixed/Native Hawaiian or Pacific Islander −.55 .58 −.96 .33 11. Body mass index (kg/m2) .01 .01 .87 .38 12. General health −1.41 .11 −12.42*** <.001 13. Allergies .89 .22 4.09*** <.001 14. Alcohol consumption .48 .19 2.51* .01 15. Arthritis .75 .32 2.32* .02 16. Cancer −.27 .54 −.50 .61 17. Chronic fatigue .72 .56 1.28 .20 18. Chronic pulmonary disease or asthma .40 .46 .86 .38 19. Cirrhosis −2.34 2.74 −.85 .39 20. Gastrointestinal disease .19 .42 .44 .65 21. Heart disease .26 .58 .44 .65 22. High cholesterol .26 .27 .96 .33 23. Kidney disease −.92 .96 −.96 .33 24. Smoking (tobacco or other) .76 .21 3.47*** .001 25. Stroke 1.35 .85 1.58 .11 26. Tuberculosis 3.24 2.13 1.52 .12 27. Type 1 diabetes 1.30 .94 1.37 .16 28. Type 2 diabetes −.06 .40 −.15 .87 29. Individual sports/physical activity‡ .41 .38 1.06 .28 30. Work/education .02 .33 .06 .95 31. Interpersonal relationships .49 .28 1.76 .07 32. Active arts −.08 .36 −.22 .81 33. Reading/writing .33 .38 .87 .38 34. Passive leisure .83 .66 1.24 .21 35. Active music .71 .41 1.74 .08 36. Team sports 1.48 .88 1.67 .09 37. Religion 1.86 .67 2.78** .006 38. Other .88 .57 1.55 .12 Adjusted R 2 .23 F 12.71*** †White acted as the reference group. ‡Active leisure acted as the reference group. *p < .05; **p < .01; ***p < .001. Open in new tab Mediation Analyses We used the PROCESS macro (model 4; [80]) to test a mediation model whereby passion (i.e., OP and HP) predicts depressive mood symptoms, which in turn predicts PSQI, controlling for demographics, alcohol and tobacco consumption, BMI, self-reported health, and diagnosed health conditions. As shown in Figure 1, OP was positively related to depressive mood symptoms (B = 0.15, SE = 0.01, t = 10.29, p < 0.001; 95% confidence interval [CI] = [0.12 to 0.18]), whereas HP was negatively related to it (B = −0.11, SE = 0.02, t = −5.46, p < 0.001; 95% CI = [−0.15 to −0.07]). Our analysis also indicated that depressive mood symptoms predicted worse sleep quality (B = 1.95, SE = 0.10, t = 17.96, p < 0.001; 95% CI = [1.74 to 2.17]). Figure 1. Open in new tabDownload slide Results from path analysis. Unstandardized betas are shown. ***p < .001; † <.10 Figure 1. Open in new tabDownload slide Results from path analysis. Unstandardized betas are shown. ***p < .001; † <.10 Bootstrapped CI estimates of the indirect effect were calculated to test the mediating role of depressive mood symptoms between passion and sleep quality. In the present study, the 95% CI of the indirect effect was obtained with 5,000 bootstraps resamples [81]. Results indicated that depressive mood symptoms mediated the relationship between OP and sleep quality (B = 0.29, SE = 0.03; 95% CI = [0.23 to 0.36]), as well the relationship between HP and sleep quality (B = −0.22, SE = 0.04; 95% CI = [−0.31 to −0.13]).2 Discussion A large body of research has documented the importance of sleep for overall well-being [83, 84], as well as the vital role that passion plays in everyday life [17, 18]. However, research thus far has not looked into whether these two literatures are interconnected. The present study aimed to fill in these gaps by examining the relationship between sleep and passion. Specifically, our main goal was to examine how people regulate their passionate activities with other life domains is associated with the quality of their sleep. Based on prior work which showed that OP (vs HP) is positively related to depressive mood symptoms [24, 39], we predicted that HP would be associated with better sleep quality, whereas OP would be associated with worse overall sleep. The results of the study confirmed these predictions. OP was positively, whereas HP was negatively, related to depressive mood symptoms, which in turn positively predicted worse sleep quality. Importantly, these results held even after adjusting for several factors including age, gender, ethnicity, BMI, a number of other health-related variables (e.g., cancer, heart disease, tuberculosis, and perceived health), and the type of activity. This demonstrates the added value and relevance of passion to the study of sleep. These results not only serve as strong evidence of the relationship that exists between sleep and passion, but they also offer two important contributions to the literature. First, the study fills a gap in the sleep literature regarding the relationship between activity engagement and sleep quality. Indeed, previous research has focused on how the quantity of time and effort placed onto activities such as work and school are associated with an individual’s quality of sleep [85–88]. Generally, such research has found that activities that consume a great deal of people’s time and energy are associated with impaired sleep quality. However, our research shows that it also depends on the way the activity is regulated with other life domains. Specifically, by demonstrating the different associations that HP versus OP shares with sleep, our study helps illuminate that it is not the quantity, but the quality of activity engagement that matters to predict sleep quality. Harmoniously passionate individuals are able to balance between their passionate activities and other life domains, thereby experiencing more positive emotions, which in turn impact the quality of their sleep. Conversely, because obsessively passionate individuals experience conflicts with other life domains, they experience depressive mood symptoms that then compromise their ability to sleep well. Second, the current study expands on the research by examining how the influence of passion can extend beyond a person’s waking hours. Although previous research has shown that the way people pursue and regulate their passionate activity affects their performance [89] and well-being during the day [90], the results described herein suggest that harboring a passion for an activity has an influence around the clock, affecting the person’s well-being even long after they have stopped engaging in the activity itself and are trying sleep. Limitations and Future Directions Despite the new insights this research offers in relation to sleep and passion, there are several limitations that should be addressed in future research. First, the results obtained are cross-sectional, which prevents us from making any causal inferences about the relationship between sleep and passion. Future research could benefit from studies that instead choose to manipulate one of the variables and look into how the other is affected. Previous studies have shown that experimentally inducing different passion mindsets is possible [31]. If a causal inference can be drawn from these studies, they might open up a new avenue for looking into whether we can improve individuals’ sleep—and ultimately their overall health—by changing the quality of their engagement in relation to their passionate activities. Second, data were collected using a survey and self-report measures, both of which come with their own set of biases that may have affected the results. For instance, results on the PSQI, or other instruments in this research, may be affected by recall bias or participants’ ability to accurately report sleep-related information. Future studies should aim to replicate our study using objective measures—such as polysomnography and actigraphy to measure sleep disturbances in a laboratory setting—and seeing if the same relationship between sleep and passion shown here still holds. Lastly, it might also be interesting to carry out similar studies in a longitudinal setting. A longitudinal study might provide evidence of a cyclical relationship that exists between sleep and passion. For example, it could be the case that poor sleep quality leads to physical exhaustion, which could in turn lead to individuals neglecting other goals in order to focus the little energy they have left on their passionate activity—a cycle that may increase individuals’ OP, depressive mood symptoms, and as a result, the quality of their sleep. In such cases, breaking the cycle using the aforementioned interventions that help change people’s means of engaging in their passionate activities might prove effective. Finally, future research would do well to examine the relationship between passion, sleep quality, and chronic injuries. There is evidence suggesting that OP (vs HP) is related to chronic injuries and negative symptoms in activities such as dancing, yoga, and exercising [35,91,92]. One hypothesis worth investigating is whether these effects could be produced by poor sleep quality. Indeed, lack of sleep impairs cognitive functioning [1] and motor performance [2], which may impair athletes’ ability to perform highly skilled movements. These questions await further empirical examination. Conclusions The current study demonstrates the existence of a strong relationship between sleep quality and passion for an activity. The results add to a growing body of literature on sleep by demonstrating how the quality of one’s engagement in daily activities can influence overall sleep quality. If the activity is out of the person’s control (OP), less sleep can be expected compared to when people are in control (HP) of this activity. This is explained by greater depressive mood symptoms associated with obsessive (vs harmonious) passion. In addition to demonstrating that passion is a new predictor of sleep quality, the results expand on the research done on passion thus far by further demonstrating passion’s connectedness to domains that extend beyond the waking hours of the day. Further research should explore the relationship between sleep and passion with objective measures of sleep quality, as well as whether interventions to change people’s passion can be used to improve their sleep. Financial disclosure. None declared Nonfinancial disclosure. None declared. Conflict of interest statement. None declared. Footnotes 1 Results on the PSQI can be dichotomized into “good” (a score of 4 or lower) or “bad” sleep quality (5 or higher). A model similar to the one tested with the continuous PSQI scores was conducted with this dichotomous classification using logistic regression. Results were similar: all variables that were significant remained significant. BMI became significant as well. More importantly, OP and HP significantly predicted worse and better sleep quality, respectively. 2 Thoemmes [82] suggests that testing alternative models “to check whether one mediation model is superior to another is inadmissible” (p. 1). Thus, these tests were not conducted. References 1. Alhola P , et al. . Sleep deprivation: impact on cognitive performance . Neuropsychiatr Dis Treat. 2007 ; 3 ( 5 ): 553 – 567 . Google Scholar PubMed OpenURL Placeholder Text WorldCat 2. Pilcher JJ , et al. . Effects of sleep deprivation on performance: a meta-analysis . Sleep. 1996 ; 19 ( 4 ): 318 – 326 . Google Scholar Crossref Search ADS PubMed WorldCat 3. Rosekind MR , et al. . The cost of poor sleep: workplace productivity loss and associated costs . J Occup Environ Med. 2010 ; 52 ( 1 ): 91 – 98 . Google Scholar Crossref Search ADS PubMed WorldCat 4. Carney CE , et al. . Symptom-focused rumination and sleep disturbance . Behav Sleep Med. 2006 ; 4 ( 4 ): 228 – 241 . Google Scholar Crossref Search ADS PubMed WorldCat 5. Thomsen DK , et al. Rumination-relationship with negative mood and sleep quality . Pers Individ Dif . 2003 ; 34 ; 1293 – 1301 . Google Scholar Crossref Search ADS WorldCat 6. Baum KT , et al. . Sleep restriction worsens mood and emotion regulation in adolescents . J Child Psychol Psychiatry. 2014 ; 55 ( 2 ): 180 – 190 . Google Scholar Crossref Search ADS PubMed WorldCat 7. Mauss IB , et al. . Poorer sleep quality is associated with lower emotion-regulation ability in a laboratory paradigm . Cogn Emot. 2013 ; 27 ( 3 ): 567 – 576 . Google Scholar Crossref Search ADS PubMed WorldCat 8. Anderson C , et al. . Sleep deprivation lowers inhibition and enhances impulsivity to negative stimuli . Behav Brain Res. 2011 ; 217 ( 2 ): 463 – 466 . Google Scholar Crossref Search ADS PubMed WorldCat 9. Dinis J , et al. . Quality of sleep and depression in college students: a systematic review . Sleep Sci. 2018 ; 11 ( 4 ): 290 – 301 . Google Scholar Crossref Search ADS PubMed WorldCat 10. Geng F , et al. Bidirectional associations between insomnia, posttraumatic stress disorder, and depressive symptoms among adolescent earthquake survivors: a longitudinal multiwave cohort study . Sleep . 2019 ; 42 ( 11 ). doi:10.1093/sleep/zsz162 Google Scholar OpenURL Placeholder Text WorldCat 11. Gregory AM , et al. . Associations between sleep quality and anxiety and depression symptoms in a sample of young adult twins and siblings . J Psychosom Res. 2011 ; 71 ( 4 ): 250 – 255 . Google Scholar Crossref Search ADS PubMed WorldCat 12. Mohan J , et al. . Association between sleep time and depression: a cross-sectional study from countries in rural Northeastern China . J Int Med Res. 2017 ; 45 ( 3 ): 984 – 992 . Google Scholar Crossref Search ADS PubMed WorldCat 13. Stein MB , et al. . Impairment associated with sleep problems in the community: relationship to physical and mental health comorbidity . Psychosom Med. 2008 ; 70 ( 8 ): 913 – 919 . Google Scholar Crossref Search ADS PubMed WorldCat 14. Kim CW , et al. . Sleep duration and quality in relation to chronic kidney disease and glomerular hyperfiltration in healthy men and women . PLoS One. 2017 ; 12 ( 4 ): e0175298 . Google Scholar Crossref Search ADS PubMed WorldCat 15. Calhoun DA , et al. . Sleep and hypertension . Chest. 2010 ; 138 ( 2 ): 434 – 443 . Google Scholar Crossref Search ADS PubMed WorldCat 16. Hepburn M , et al. . Sleep medicine: stroke and sleep . Mo Med. 2018 ; 115 ( 6 ): 527 – 532 . Google Scholar PubMed OpenURL Placeholder Text WorldCat 17. Vallerand RJ . On passion for life activities: the dualistic model of passion . In: Zanna MP, ed. Advances in Experimental Social Psychology . Vol. 42. New York, NY: Academic Press ; 2010 ; 97 – 193 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC 18. Vallerand RJ. The Psychology of Passion, A Dualistic Model . New York, NY : Oxford University Press ; 2015 . Google Scholar Crossref Search ADS Google Preview WorldCat COPAC 19. Philippe FL , et al. . Passion does make a difference in people’s lives: a look at well‐being in passionate and non‐passionate individuals . Appl Psychol Health Well Being . 2009 ; 1 : 3 – 22 . Google Scholar Crossref Search ADS WorldCat 20. Curran T , et al. The psychology of passion: a meta-analytical review of a decade of research on intrapersonal outcomes . Motiv Emot . 2015 ; 39 ( 5 ): 631 – 655 . Google Scholar Crossref Search ADS WorldCat 21. Vallerand RJ , et al. . Les passions de l’ame: on obsessive and harmonious passion . J Pers Soc Psychol. 2003 ; 85 ( 4 ): 756 – 767 . Google Scholar Crossref Search ADS PubMed WorldCat 22. Mageau G , et al. . The role of self-esteem contingencies in the distinction between obsessive and harmonious passion . Eur J Soc Psychol . 2011 ; 41 ; 720 – 729 . Google Scholar Crossref Search ADS WorldCat 23. Mageau GA , et al. . Passion and gambling: investigating the divergent affective and cognitive consequences of gambling . J Appl Soc Psychol . 2005 ; 35 ( 1 ): 100 – 118 . Google Scholar Crossref Search ADS WorldCat 24. Houlfort N , et al. On passion and heavy work investment: personal and organizational outcomes . J Manag Psychol . 2014 ; 29 ( 1 ): 25 – 45 . Google Scholar Crossref Search ADS WorldCat 25. Lalande D , et al. . Obsessive passion: a compensatory response to unsatisfied needs . J Pers. 2017 ; 85 ( 2 ): 163 – 178 . Google Scholar Crossref Search ADS PubMed WorldCat 26. Lafrenière MA , et al. . Passion in sport: on the quality of the coach-athlete relationship . J Sport Exerc Psychol. 2008 ; 30 ( 5 ): 541 – 560 . Google Scholar Crossref Search ADS PubMed WorldCat 27. Vallerand RJ , et al. Passion in sport: a look at determinants and affective experiences . J Sport Exerc Psychol . 2006 ; 28 : 454 – 478 . Google Scholar Crossref Search ADS WorldCat 28. St-Louis AC , et al. . Passion for a cause: how it affects health and subjective well-being . J Pers. 2016 ; 84 ( 3 ): 263 – 276 . Google Scholar Crossref Search ADS PubMed WorldCat 29. Vallerand RJ , et al. . On the role of passion in performance . J Pers. 2007 ; 75 ( 3 ): 505 – 533 . Google Scholar Crossref Search ADS PubMed WorldCat 30. Vallerand RJ . On the synergy between hedonia and eudaimonia: the role of passion . In: Vittersø J, ed. Handbook of Eudaimonic Well-Being. International Handbooks of Quality-of-Life . Cham, Switzerland : Springer ; 2016 : 191 – 204 . Google Scholar Crossref Search ADS Google Preview WorldCat COPAC 31. Bélanger J , et al. Using implementation intentions to change passion: the role of environmental mastery and basic psychological needs . Motiv Sci . 2019 ; 5 ( 4 ): 343 – 356 . Google Scholar Crossref Search ADS WorldCat 32. Stenseng F , et al. The dark side of leisure: obsessive passion and its covariates and outcomes . Leisure Studies . 2011 ; 30 ( 1 ): 49 – 62 . Google Scholar Crossref Search ADS WorldCat 33. Ratelle CF , et al. . When passion leads to problematic outcomes: a look at gambling . J Gambl Stud. 2004 ; 20 ( 2 ): 105 – 119 . Google Scholar Crossref Search ADS PubMed WorldCat 34. Carpentier J , et al. Ruminations and flow: why do people with a more harmonious passion experience higher well-being? . J Happiness Stud . 2012 ; 13 : 501 – 518 . Google Scholar Crossref Search ADS WorldCat 35. Rip B , et al. The relationship between passion and injury in dance students . J Dance Med Sci . 2006 ; 10 ; 14 – 20 . Google Scholar OpenURL Placeholder Text WorldCat 36. Jowett S , et al. Passion for activities and relationship quality: a dyadic approach . J Pers Soc Psychol . 2013 ; 30 : 734 – 749 . Google Scholar OpenURL Placeholder Text WorldCat 37. Vallerand RJ , et al. . On the role of passion for work in burnout: a process model . J Pers. 2010 ; 78 ( 1 ): 289 – 312 . Google Scholar Crossref Search ADS PubMed WorldCat 38. Donahue EG , et al. . Passion for work and emotional exhaustion: the mediating role of rumination and recovery . Appl Psychol Health Well Being. 2012 ; 4 ( 3 ): 341 – 368 . Google Scholar Crossref Search ADS PubMed WorldCat 39. Rousseau FL , et al. Le rôle de la passion dans le bien-être subjectif des aînés . Revue Québécoise de psychologie . 2003 ; 24 ( 3 ): 197 – 211 . Google Scholar OpenURL Placeholder Text WorldCat 40. Phillips BA , et al. . Cigarette smoking and sleep disturbance . Arch Intern Med. 1995 ; 155 ( 7 ): 734 – 737 . Google Scholar Crossref Search ADS PubMed WorldCat 41. Jaehne A , et al. . How smoking affects sleep: a polysomnographical analysis . Sleep Med. 2012 ; 13 ( 10 ): 1286 – 1292 . Google Scholar Crossref Search ADS PubMed WorldCat 42. Thakkar MM , et al. . Alcohol disrupts sleep homeostasis . Alcohol. 2015 ; 49 ( 4 ): 299 – 310 . Google Scholar Crossref Search ADS PubMed WorldCat 43. Roehrs T , et al. . Sleep, sleepiness, sleep disorders and alcohol use and abuse . Sleep Med Rev. 2001 ; 5 ( 4 ): 287 – 297 . Google Scholar Crossref Search ADS PubMed WorldCat 44. Slavish DC , et al. . Rumination mediates the relationships between depressed mood and both sleep quality and self-reported health in young adults . J Behav Med. 2015 ; 38 ( 2 ): 204 – 213 . Google Scholar Crossref Search ADS PubMed WorldCat 45. Budhiraja R , et al. . Sleep disorders in chronic obstructive pulmonary disease: etiology, impact, and management . J Clin Sleep Med. 2015 ; 11 ( 3 ): 259 – 270 . Google Scholar Crossref Search ADS PubMed WorldCat 46. Foster GD , et al. .; Sleep AHEAD Research Group. Obstructive sleep apnea among obese patients with type 2 diabetes . Diabetes Care. 2009 ; 32 ( 6 ): 1017 – 1019 . Google Scholar Crossref Search ADS PubMed WorldCat 47. Heeren M , et al. . Active at night, sleepy all day–sleep disturbances in patients with hepatitis C virus infection . J Hepatol. 2014 ; 60 ( 4 ): 732 – 740 . Google Scholar Crossref Search ADS PubMed WorldCat 48. Dugas EN , et al. . Nicotine dependence and sleep quality in young adults . Addict Behav. 2017 ; 65 : 154 – 160 . Google Scholar Crossref Search ADS PubMed WorldCat 49. Medic G , et al. . Short- and long-term health consequences of sleep disruption . Nat Sci Sleep. 2017 ; 9 : 151 – 161 . Google Scholar Crossref Search ADS PubMed WorldCat 50. Stein MD , et al. . Disturbed sleep and its relationship to alcohol use . Subst Abus. 2005 ; 26 ( 1 ): 1 – 13 . Google Scholar Crossref Search ADS PubMed WorldCat 51. Vargas PA , et al. . Sleep quality and body mass index in college students: the role of sleep disturbances . J Am Coll Health. 2014 ; 62 ( 8 ): 534 – 541 . Google Scholar Crossref Search ADS PubMed WorldCat 52. Wetter DW , et al. . The relation between cigarette smoking and sleep disturbance . Prev Med. 1994 ; 23 ( 3 ): 328 – 334 . Google Scholar Crossref Search ADS PubMed WorldCat 53. Bonneville-Roussy A , et al. When passion leads to excellence: the case of musicians . Psychol Music . 2010 ; 39 ( 1 ): 123 – 138 . Google Scholar Crossref Search ADS WorldCat 54. Carbonneau N , et al. The role of passion for teaching in intrapersonal and interpersonal outcomes . J Educ Psychol . 2008 ; 100 ( 4 ): 977 – 987 . Google Scholar Crossref Search ADS WorldCat 55. Lafrenière MAK , et al. Self-esteem and passion for activities . Pers Individ Dif . 2011 ; 51 ; 541 – 544 . Google Scholar Crossref Search ADS WorldCat 56. Buysse DJ , et al. . The Pittsburgh Sleep Quality Index: a new instrument for psychiatric practice and research . Psychiatry Res. 1989 ; 28 ( 2 ): 193 – 213 . Google Scholar Crossref Search ADS PubMed WorldCat 57. Hedeker D , et al. . Analysis of binary outcomes with missing data: missing = smoking, last observation carried forward, and a little multiple imputation . Addiction. 2007 ; 102 ( 10 ): 1564 – 1573 . Google Scholar Crossref Search ADS PubMed WorldCat 58. Hu FB , et al. . Comparison of population-averaged and subject-specific approaches for analyzing repeated binary outcomes . Am J Epidemiol. 1998 ; 147 ( 7 ): 694 – 703 . Google Scholar Crossref Search ADS PubMed WorldCat 59. Pradhan PMS , et al. Tobacco use and associated factors among adolescent students in Dharan, Eastern Nepal: a cross-sectional questionnaire survey . BMJ Open . 2013 ; 3 ( 2 ): e002123 . Google Scholar Crossref Search ADS PubMed WorldCat 60. Preisser JS , et al. Analysis of smoking trends with incomplete longitudinal binary responses . J Am Stat Assoc . 2000 ; 95 ( 452 ): 1021 – 1031 . Google Scholar Crossref Search ADS WorldCat 61. DeSalvo KB , et al. . Assessing measurement properties of two single-item general health measures . Qual Life Res. 2006 ; 15 ( 2 ): 191 – 201 . Google Scholar Crossref Search ADS PubMed WorldCat 62. Schwartz AR , et al. . Obesity and obstructive sleep apnea: pathogenic mechanisms and therapeutic approaches . Proc Am Thorac Soc. 2008 ; 5 ( 2 ): 185 – 192 . Google Scholar Crossref Search ADS PubMed WorldCat 63. Romero-Corral A , et al. . Interactions between obesity and obstructive sleep apnea: implications for treatment . Chest. 2010 ; 137 ( 3 ): 711 – 719 . Google Scholar Crossref Search ADS PubMed WorldCat 64. Resta O , et al. . Sleep-related breathing disorders, loud snoring and excessive daytime sleepiness in obese subjects . Int J Obes Relat Metab Disord. 2001 ; 25 ( 5 ): 669 – 675 . Google Scholar Crossref Search ADS PubMed WorldCat 65. Gami AS , et al. . Obesity and obstructive sleep apnea . Endocrinol Metab Clin North Am. 2003 ; 32 ( 4 ): 869 – 894 . Google Scholar Crossref Search ADS PubMed WorldCat 66. Field AP. Discovering Statistics Using IBM SPSS Statistics . London, UK : SAGE ; 2013 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC 67. Hair JF Jr , et al. Multivariate Data Analysis . 3rd ed. New York, NY : Macmillan ; 1995 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC 68. Neter J , et al. Applied Linear Regression Models. 2nd ed. Homewood, IL. : Irwin ; 1989 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC 69. Stevens J. Applied Multivariate Statistics for the Social Sciences . Hillsdale, NJ : Lawrence Erlbaum Associates ; 2002 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC 70. Meyers LS , et al. Applied Multivariate Research: Design and Interpretation . Thousand Oaks, CA : Sage Publications ; 2006 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC 71. Tabachnick BG , et al. Multiple regression. In: Tabachnick BG, Fidell LS, eds. Using Multivariate Statistics . 4th ed. Boston, MA : Allyn and Bacon ; 2001 : 709 – 811 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC 72. Gadie A , et al. .; Cam-CAN. How are age-related differences in sleep quality associated with health outcomes? An epidemiological investigation in a UK cohort of 2406 adults . BMJ Open. 2017 ; 7 ( 7 ): e014920 . Google Scholar Crossref Search ADS PubMed WorldCat 73. Mallampalli MP , et al. . Exploring sex and gender differences in sleep health: a Society for Women’s Health Research Report . J Womens Health (Larchmt). 2014 ; 23 ( 7 ): 553 – 562 . Google Scholar Crossref Search ADS PubMed WorldCat 74. Krause N , et al. Is involvement in religion associated with better sleep quality? Pastoral Psychology . 2017 ; 66 ( 5 ): 595 – 608 . Google Scholar Crossref Search ADS WorldCat 75. Hill TD , et al. . Religious involvement as a social determinant of sleep: an initial review and conceptual model . Sleep Health. 2018 ; 4 ( 4 ): 325 – 330 . Google Scholar Crossref Search ADS PubMed WorldCat 76. Janson C , et al. . Increased prevalence of sleep disturbances and daytime sleepiness in subjects with bronchial asthma: a population study of young adults in three European countries . Eur Respir J. 1996 ; 9 ( 10 ): 2132 – 2138 . Google Scholar Crossref Search ADS PubMed WorldCat 77. Léger D , et al. . Allergic rhinitis and its consequences on quality of sleep: an unexplored area . Arch Intern Med. 2006 ; 166 ( 16 ): 1744 – 1748 . Google Scholar Crossref Search ADS PubMed WorldCat 78. Santos CB , et al. Allergic rhinitis and its effect on sleep, fatigue and daytime somnolence . Ann Allergy Immunol . 2006 ; 97 ( 5 ): 579 – 586 . Google Scholar Crossref Search ADS WorldCat 79. Westhovens R , et al. . Sleep problems in patients with rheumatoid arthritis . J Rheumatol. 2014 ; 41 ( 1 ): 31 – 40 . Google Scholar Crossref Search ADS PubMed WorldCat 80. Hayes AF . Introduction to Mediation, Moderation, and Conditional Process Analysis: A Regression-Based Approach. New York: The Guilford Press; 2018. 81. Preacher KJ , Hayes AF. Asymptotic and resampling strategies for assessing and comparing indirect effects in multiple mediator models. Behav Res Methods. 2008 ; 40 (3): 879 – 891 . Google Scholar Crossref Search ADS PubMed WorldCat 82. Thoemmes F . Reversing arrows in mediation models does not distinguish plausible models . Basic Appl Soc Psychol . 2015 ; 37 ( 4 ): 226 – 234 . Google Scholar Crossref Search ADS WorldCat 83. Mukherjee S , et al. .; American Thoracic Society ad hoc Committee on Healthy Sleep. An official American Thoracic Society statement: the importance of healthy sleep. Recommendations and future priorities . Am J Respir Crit Care Med. 2015 ; 191 ( 12 ): 1450 – 1458 . Google Scholar Crossref Search ADS PubMed WorldCat 84. Worley SL . The extraordinary importance of sleep: the detrimental effects of inadequate sleep on health and public safety drive an explosion of sleep research . P T. 2018 ; 43 ( 12 ): 758 – 763 . Google Scholar PubMed OpenURL Placeholder Text WorldCat 85. Afonso P , et al. . Impact of working hours on sleep and mental health . Occup Med (Lond). 2017 ; 67 ( 5 ): 377 – 382 . Google Scholar Crossref Search ADS PubMed WorldCat 86. Galloway M , et al. Nonacademic effects of homework in privileged, high-performing high schools . J Exp Educ . 2013 ; 81 ; 490 – 510 . Google Scholar Crossref Search ADS WorldCat 87. Nakashima M , et al. . Association between long working hours and sleep problems in white-collar workers . J Sleep Res. 2011 ; 20 ( 1 Pt 1 ): 110 – 116 . Google Scholar Crossref Search ADS PubMed WorldCat 88. Virtanen M , et al. . Long working hours and sleep disturbances: the Whitehall II prospective cohort study . Sleep. 2009 ; 32 ( 6 ): 737 – 745 . Google Scholar Crossref Search ADS PubMed WorldCat 89. Bélanger J , et al. Driven by fear: the effect of success and failure information on passionate individuals’ performance . J Pers Soc Psychol . 2013 ; 104 ( 1 ): 180 – 195 . Google Scholar Crossref Search ADS PubMed WorldCat 90. Vallerand RJ . The role of passion in sustainable psychological well-being . Psychol Well Being Theory Res Pract . 2012 ; 2 ; 1 – 21 . Google Scholar Crossref Search ADS WorldCat 91. Stephan Y , et al. Predictors of perceived susceptibility to sport-related injury among competitive runners: the role of previous experience, neuroticism, and passion for running . Appl Psychol . 2009 ; 58 ( 4 ): 672 – 689 . Google Scholar Crossref Search ADS WorldCat 92. Carbonneau N , et al. Is the practice of yoga associated with positive outcomes? The role of passion . J Posit Psychol . 2010 ; 5 ( 6 ): 452 – 465 . Google Scholar Crossref Search ADS WorldCat Author notes These authors have contributed equally to the manuscript. © Sleep Research Society 2020. Published by Oxford University Press on behalf of the Sleep Research Society. All rights reserved. For permissions, please e-mail [email protected]. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)

Hammam,, Nevin;Sadeghi,, Dorna;Carson,, Valerie;Tamana, Sukhpreet, K;Ezeugwu, Victor, E;Chikuma,, Joyce;, van Eeden, Charmaine;Brook, Jeffrey, R;Lefebvre, Diana, L;Moraes, Theo, J;Subbarao,, Padmaja;Becker, Allan, B;Turvey, Stuart, E;Sears, Malcolm, R;

Laganière,, Christine;Pennestri,, Marie-Hélène;Rassu, Anna, Laura;Barateau,, Lucie;Chenini,, Sofiène;Evangelista,, Elisa;Dauvilliers,, Yves;Lopez,, Régis

doi: 10.1093/sleep/zsaa105pmid: 32459316

Abstract Study Objectives Rhythmic movements (RMs) during sleep are frequent and often considered as benign in children. Disabling forms are diagnosed as RM disorder and may persist in adulthood. Whether RMs severely impact sleep architecture in patients with RM disorder remain unclear. We performed a case–control study to characterize the clinical and polysomnographic patterns of children and adults with a diagnosis of RM disorder in comparison to controls, and to assess the associations between the RMs and the sleep architecture. Methods All consecutive patients (n = 50; 27 children, 35 males) with RM disorder from a single sleep clinic (from 2006 to 2019) underwent a comprehensive clinical evaluation and a polysomnographic recording in comparison to 75 controls (42 children and 53 males). Results About 82% of children and adult patients had a complaint of disturbed nighttime sleep. Comorbid neurodevelopmental, affective or sleep disorders were found in 92% of patients. While RM sequences defined by video polysomnographic criteria were observed in 82% of patients (in wakefulness and in all sleep stages), no similar sequences were observed in controls. Patients had altered sleep continuity, with low sleep efficiency, increased wake time after sleep onset, and frequent periodic leg movements and apnea events. The severity of RMs was associated with disrupted nighttime sleep, even after controlling for comorbid motor and respiratory events. Conclusions RM disorder is a rare, highly comorbid and disabling condition both in children and adults with frequent disturbed nighttime sleep that may contribute to the burden of the disease. rhythmic movement disorder, bodyrocking, headbanging, polysomnography Statement of Significance Although rhythmic movements (RMs) during sleep are frequent, rhythmic movement disorder (RMD) is a rare condition with only 50 children and adult patients identified over a period of 13 years in a single sleep clinic. Patients with RMD had frequent complaints of nighttime disturbed sleep and objective altered sleep continuity in comparison with controls. RMs sequences were recorded during the video polysomnography in most patients, both during wakefulness and in all sleep stages. Despite frequent periodic leg movements and apnea events seen in patients with RMD, the severity of RM was independently associated with sleep disruption. Introduction Rhythmic movements (RMs) during sleep are repetitive, stereotyped, and rhythmic oscillations involving large groups of muscles predominately at sleep onset with potential persistence during sleep or that can arise out of consolidated sleep. They encompass a large spectrum of manifestations: body rocking, body rolling, head rolling, or head banging. RMs is thought to be highly prevalent in infants and to decrease with age [1–4], but persistence in adulthood is not uncommon [1, 5–9]. In contrast with RM, often considered as a benign phenomenon, rhythmic movement disorder (RMD) is a sleep-related motor disorder characterized by a severe RM phenotype, associated with significant functional impairment [1]. RMD is less frequent, probably around 1%–3% in children, these severe forms may also persist in adults, with a potential risk of injuries [9–15]. RMD remains a poorly studied sleep disorder. Little is known regarding sleep quality in those patients, with only few case series reporting associated disturbed nighttime sleep [5, 7, 14–23]. RMs frequently occur at bedtime leading to sleep onset insomnia, but also during sleep, possibly impairing sleep continuity and quality [6, 9, 14, 16, 24–27]. RMD may also be associated with other sleep disorders, disorders of arousal (DOA), restless legs syndrome (RLS), periodic leg movement (PLM) or sleep-related breathing disorders [25, 28–32]. All these conditions share an excessive nighttime sleep fragmentation. Whether objective disturbed nighttime sleep in patients with RMD is associated with RM severity or with frequent comorbid sleep disorders remains unknown. To the best of our knowledge, no case–control studies assessing the characteristics of disturbed nighttime sleep based on polysomnography parameters in patients with RMD were reported. We thus conducted a study aiming to (1) provide a clinical and polysomnographic evaluation of children and adult patients with RMD, (2) compare sleep parameters assessed by polysomnography in patients with RMD and controls, and (3) to study the associations between RMs characteristics and polysomnographic parameters of disturbed nighttime sleep in patients with RMD. Methods Population A total of 50 patients (70% males, median age 14.3, range 3.4–60.7) meeting the diagnostic criteria for RMD were retrospectively identified in a single sleep center from 2006 to 2019. The population included 27 children aged ≤16 years (24 males) and 23 patients >16 years old (11 males). The diagnosis of RMD was made according to the Third Revision of the International Classification of Sleep Disorders [1]. It required (1) a clinical history of repetitive, stereotyped, and RMs involving large muscle groups; (2) predominately related to sleep; (3) resulting in a significant complaint such as interference with normal sleep, impairment in daytime functioning or injuries; and (4) being not better explained by epilepsy or another movement disorder. RMs were the primary complaint that leaded to the medical investigations in 43 patients. The seven other patients were first referred for complaints of disturbed nighttime sleep, insomnia, non-restorative sleep, or daytime sleepiness, although they were aware of having these nighttime movements. We also recruited 75 community-dwelling participants (70.7% males, median age 15.3 years, range 3.9–62.4, including 33 > 16 years old), without past or present history of RM or other significant complaint of sleep disorder (with a ratio of 1.5 control per patient, matched for gender), through advertisements and local association networks from the general population. All participants or their parents signed informed consent, and this study was conducted in accordance with the Declaration of Helsinki and French Good Clinical Practices. Clinical evaluation RM phenotype (body rocking, body rolling, head rolling, or head banging), age at onset, frequency of RM episodes, and potential impairment associated with RM (RM-related injuries, sleep disruption, fatigue, excessive daytime sleepiness, and interferences with parent/bedpartner’s sleep) were systematically documented with a clinical interview, and retrieved from available medical charts. Information on positive family history of RM (first and second degrees) was detailed for 37 patients. Current comorbid neurodevelopmental (i.e. attention deficit hyperactivity disorder [ADHD], autism spectrum disorder, intellectual, and learning disabilities), affective (i.e. mood and anxiety disorders), and other sleep disorders including RLS and DOA were also reported. Polysomnographic assessment All participants underwent one-night video-PSG recording in the sleep laboratory. This included electroencephalogram (EEG; leads at C3/A2, Fp1/T1, T1/O1, O1/C3, C4/A1, Fp2/T2, T2/O2, O2/C4), electro-oculogram, chin electromyogram (EMG), and electrocardiogram. Respiration was monitored with a nasal cannula/pressure transducer system, mouth thermistor, chest and abdominal bands, and pulse oximeter. Leg movements were assessed with surface EMG electrodes placed on the right and left anterior tibialis muscles. Sleep stages, microarousals and respiratory events were manually scored according to standard criteria [33]. Sleep-stage scoring during RMs was unreliable because of pervasive motor artifacts in all channels (Figure 1). The epochs occupied by more than 15 s of RMs were scored as “movement time” [34]. The remaining epochs containing RM sequences were scored according to the underlying sleep stage, and RMs were scored as microarousal. Figure 1. Open in new tabDownload slide Polysomnographic aspects of RMs sequences. Bodyrocking episode emerging from wakefulness (A), headrolling episode emerging from REM sleep (B), and four successive bodyrolling episodes emerging from REM sleep (C). Figure 1. Open in new tabDownload slide Polysomnographic aspects of RMs sequences. Bodyrocking episode emerging from wakefulness (A), headrolling episode emerging from REM sleep (B), and four successive bodyrolling episodes emerging from REM sleep (C). Nighttime sleep architecture was assessed by total sleep time (TST), sleep efficiency (SE), sleep latency (SOL), wake after sleep onset (WASO) duration, and microarousal index. The epochs scored as “movement time” were not used for the calculation of SE and WASO. Sleep apnea syndrome (SAS) was defined as having an apnea–hypopnea index (AHI) >15/h in adults (>30/h for severe SAS) and >1.5/h in children [35]. PLMs were scored following the criteria set by the International Restless Legs Syndrome Study Group [36]. None of the participants were treated by a medication that could interfere with sleep or motor activities during the 2 weeks prior to the PSG recording. RMs scoring All the 50 video-PSG were analyzed at the time of diagnosis by one of the sleep physicians to assess RM sequences based on a careful inspection of the synchronized video recording. Due to our limited storage capacity in previous years, not all of the video recordings were systematically backed up, and only 20 full synchronized video recordings were available for re-analysis. However, when RM sequences were identified at the clinic, short video captures of the events were systematically saved as well as their presence in the sleep technician’s reports. A RM sequence was defined by a cluster of at least four body and/or head movements, with a frequency of 0.5–2 Hz according to the American Academy of Sleep Medicine criteria [37]. RMs were systematically associated with artifacts on EMG, EEG, respiratory, or ECG leads (Figure 1). We calculated the diagnostic value of the identification of these artifacts during the re-evaluation of 11 PSGs involving frequent body and/or head RMs (ranging from 21 to 108 sequences per PSG) by a second blind scorer (R.L.) of the video recording. Among the 539 RM sequences identified by the video-PSG, 518 (96.1%) were correctly identified by the PSG only (second rater). Only 19 (3.4%) artifacts scored as RMs on the PSG were not confirmed on the video recording. Overall, the agreement between the two scoring methods (video-PSG and PSG alone) was 92.8%. We further applied this scoring method to analyze the 30 remaining PSG recordings without the full video recording available. However, the PSG scoring alone could not identify the muscle group involved in the RM sequence (e.g. body or head movements), nor their subtype (e.g. rolling, banging, rocking movements). Accordingly, various parameters related to RM have been detailed for the 20 video-PSG available: RM subtype (i.e. body or head RM), total number of sequences, duration, total time spent in RM, occurrence from sleep stages and wakefulness, and distribution (i.e. before sleep onset, in the first, second, and third parts of the night). We calculated the RM index defined by the number of RM sequences per hour of time in bed, as well as the RM duration defined as the percentage of time in bed occupied by RMs, according to a recent work [38]. Statistical analysis Clinical and RM characteristics were described using median and range for continuous variables, and percentages for categorical variables. Independent t-tests were used to assess case–control differences for continuous PSG variables and chi-squares for dichotomous variables. Comparisons between cases and controls were made according to the two age groups (≤16 and >16 years old) and the full population. Partial Pearson’s correlations were used to assess association between RM characteristics and other PSG parameters while controlling for age and comorbid sleep disorders. All statistical analyses were made using SPSS version 24 for Windows. Statistical significance was set at p < 0.05. Results Clinical characteristics of RMD patients A total of 36 (72%) patients reported body RMs (29 with body rolling, 6 with body rocking, and 1 with unspecified body RM), 26 (52%) head RMs (11 with head rolling, 10 with head banging, 4 with both phenotypes, and 1 with unspecified head RM), and 12 patients (24%) both head and body RMs. A total of 32 patients reported a daily occurrence of RM. A complaint of disturbed nighttime sleep was found in 82% of patients (81% in children and 83% in adults), altered daytime functioning in 80%, and parent’s or siblings’ sleep disruption in 72%. Fifteen patients had a past history of injurious RMs. All the patients, except one, reported an age at onset before 12 year old, with 32 (67%) before 3 year old. The clinical RM phenotype was comparable between children and adults, except for a higher sleep parent’s disruption and a younger age at onset in children (Table 1). Table 1. Clinical and polysomnographic RMs phenotype in children and adults with RMD . Total (n = 50) . Adults (n = 23) . Children (n = 27) . . . Mean (SD) or % . Mean (SD) or % . Mean (SD) or % . P-value . Clinical characteristics (patient report)  Body RMs 72% 70% 74% 0.723  Head RMs 52% 52% 52% 0.982  Body and head RMs 24% 22% 26% 0.730  Familial history of RM (yes) 43% 44%* 43%* 0.957  Daily occurrence of RM (yes) 74% 68%* 79%* 0.423  Age at onset (>3 years old) 33% 52%* 19% 0.014  History of RM-related injuries (yes) 30% 22% 37% 0.239  Sleep disruption (yes) 82% 83% 81% 0.918  Altered daytime functioning (yes) 80% 87% 74% 0.256  Impairment in family functioning (yes) 72% 48% 93% <0.001 Comorbidities  Restless legs syndrome 32% 52% 15% 0.005  Disorder of arousal 26% 17% 33% 0.200  Neurodevelopmental disorder 46% 26% 63% 0.009  Affective disorder 26% 39% 19% 0.106 Polysomnographic RMs phenotype  RM sequences (yes) 82% 83% 81% 0.918  RM duration (min) 43.82 (61.70) 54.65 (74.94) 34.59 (47.22) 0.256  RM duration index (% of time in bed) 8.79 (12.60) 11.58 (15.69) 6.40 (8.83) 0.169  RM sequences (Time in bed) 54.18 (84.09) 68.52 (94.57) 41.96 (73.64) 0.270  RM index (/h of time in bed) 6.61 (10.72) 8.90 (12.79) 4.65 (8.33) 0.164  RM sequences (wakefulness) 36.72 (63.15) 45.35 (70.77) 29.37 (56.18) 0.378  RM sequences (N1 sleep) 5.92 (12.63) 7.26 (14.72) 4.78 (10.69) 0.494  RM sequences (N2 sleep) 6.78 (14.02) 8.48 (15.55) 5.33 (12.69) 0.435  RM sequences (N3 sleep) 0.46 (1.55) 0.70 (2.10) 0.26 (0.86) 0.328  RM sequences (REM sleep) 4.3 (13.50) 6.74 (17.73) 2.22 (8.24) 0.242  RM sequences before sleep onset 6.66 (13.36) 8.65 (16.48) 4.96 (10.05) 0.336  RM sequences first third 16.24 (25.06) 21.30 (27.48) 11.93 (22.42) 0.190  RM sequences second third 21.14 (41.23) 24.39 (39.11) 18.37 (43.49) 0.612  RM sequences last third 16.80 (29.51) 22.83 (38.09) 11.67 (18.84) 0.185 . Total (n = 50) . Adults (n = 23) . Children (n = 27) . . . Mean (SD) or % . Mean (SD) or % . Mean (SD) or % . P-value . Clinical characteristics (patient report)  Body RMs 72% 70% 74% 0.723  Head RMs 52% 52% 52% 0.982  Body and head RMs 24% 22% 26% 0.730  Familial history of RM (yes) 43% 44%* 43%* 0.957  Daily occurrence of RM (yes) 74% 68%* 79%* 0.423  Age at onset (>3 years old) 33% 52%* 19% 0.014  History of RM-related injuries (yes) 30% 22% 37% 0.239  Sleep disruption (yes) 82% 83% 81% 0.918  Altered daytime functioning (yes) 80% 87% 74% 0.256  Impairment in family functioning (yes) 72% 48% 93% <0.001 Comorbidities  Restless legs syndrome 32% 52% 15% 0.005  Disorder of arousal 26% 17% 33% 0.200  Neurodevelopmental disorder 46% 26% 63% 0.009  Affective disorder 26% 39% 19% 0.106 Polysomnographic RMs phenotype  RM sequences (yes) 82% 83% 81% 0.918  RM duration (min) 43.82 (61.70) 54.65 (74.94) 34.59 (47.22) 0.256  RM duration index (% of time in bed) 8.79 (12.60) 11.58 (15.69) 6.40 (8.83) 0.169  RM sequences (Time in bed) 54.18 (84.09) 68.52 (94.57) 41.96 (73.64) 0.270  RM index (/h of time in bed) 6.61 (10.72) 8.90 (12.79) 4.65 (8.33) 0.164  RM sequences (wakefulness) 36.72 (63.15) 45.35 (70.77) 29.37 (56.18) 0.378  RM sequences (N1 sleep) 5.92 (12.63) 7.26 (14.72) 4.78 (10.69) 0.494  RM sequences (N2 sleep) 6.78 (14.02) 8.48 (15.55) 5.33 (12.69) 0.435  RM sequences (N3 sleep) 0.46 (1.55) 0.70 (2.10) 0.26 (0.86) 0.328  RM sequences (REM sleep) 4.3 (13.50) 6.74 (17.73) 2.22 (8.24) 0.242  RM sequences before sleep onset 6.66 (13.36) 8.65 (16.48) 4.96 (10.05) 0.336  RM sequences first third 16.24 (25.06) 21.30 (27.48) 11.93 (22.42) 0.190  RM sequences second third 21.14 (41.23) 24.39 (39.11) 18.37 (43.49) 0.612  RM sequences last third 16.80 (29.51) 22.83 (38.09) 11.67 (18.84) 0.185 *Data not reported for all patients (familial history: adults n = 16, children n = 21; daily occurrence: adults n = 19, children n = 24; late onset: adults n = 21). RM, rhythmic movements; REM, rapid eye movement. Open in new tab Table 1. Clinical and polysomnographic RMs phenotype in children and adults with RMD . Total (n = 50) . Adults (n = 23) . Children (n = 27) . . . Mean (SD) or % . Mean (SD) or % . Mean (SD) or % . P-value . Clinical characteristics (patient report)  Body RMs 72% 70% 74% 0.723  Head RMs 52% 52% 52% 0.982  Body and head RMs 24% 22% 26% 0.730  Familial history of RM (yes) 43% 44%* 43%* 0.957  Daily occurrence of RM (yes) 74% 68%* 79%* 0.423  Age at onset (>3 years old) 33% 52%* 19% 0.014  History of RM-related injuries (yes) 30% 22% 37% 0.239  Sleep disruption (yes) 82% 83% 81% 0.918  Altered daytime functioning (yes) 80% 87% 74% 0.256  Impairment in family functioning (yes) 72% 48% 93% <0.001 Comorbidities  Restless legs syndrome 32% 52% 15% 0.005  Disorder of arousal 26% 17% 33% 0.200  Neurodevelopmental disorder 46% 26% 63% 0.009  Affective disorder 26% 39% 19% 0.106 Polysomnographic RMs phenotype  RM sequences (yes) 82% 83% 81% 0.918  RM duration (min) 43.82 (61.70) 54.65 (74.94) 34.59 (47.22) 0.256  RM duration index (% of time in bed) 8.79 (12.60) 11.58 (15.69) 6.40 (8.83) 0.169  RM sequences (Time in bed) 54.18 (84.09) 68.52 (94.57) 41.96 (73.64) 0.270  RM index (/h of time in bed) 6.61 (10.72) 8.90 (12.79) 4.65 (8.33) 0.164  RM sequences (wakefulness) 36.72 (63.15) 45.35 (70.77) 29.37 (56.18) 0.378  RM sequences (N1 sleep) 5.92 (12.63) 7.26 (14.72) 4.78 (10.69) 0.494  RM sequences (N2 sleep) 6.78 (14.02) 8.48 (15.55) 5.33 (12.69) 0.435  RM sequences (N3 sleep) 0.46 (1.55) 0.70 (2.10) 0.26 (0.86) 0.328  RM sequences (REM sleep) 4.3 (13.50) 6.74 (17.73) 2.22 (8.24) 0.242  RM sequences before sleep onset 6.66 (13.36) 8.65 (16.48) 4.96 (10.05) 0.336  RM sequences first third 16.24 (25.06) 21.30 (27.48) 11.93 (22.42) 0.190  RM sequences second third 21.14 (41.23) 24.39 (39.11) 18.37 (43.49) 0.612  RM sequences last third 16.80 (29.51) 22.83 (38.09) 11.67 (18.84) 0.185 . Total (n = 50) . Adults (n = 23) . Children (n = 27) . . . Mean (SD) or % . Mean (SD) or % . Mean (SD) or % . P-value . Clinical characteristics (patient report)  Body RMs 72% 70% 74% 0.723  Head RMs 52% 52% 52% 0.982  Body and head RMs 24% 22% 26% 0.730  Familial history of RM (yes) 43% 44%* 43%* 0.957  Daily occurrence of RM (yes) 74% 68%* 79%* 0.423  Age at onset (>3 years old) 33% 52%* 19% 0.014  History of RM-related injuries (yes) 30% 22% 37% 0.239  Sleep disruption (yes) 82% 83% 81% 0.918  Altered daytime functioning (yes) 80% 87% 74% 0.256  Impairment in family functioning (yes) 72% 48% 93% <0.001 Comorbidities  Restless legs syndrome 32% 52% 15% 0.005  Disorder of arousal 26% 17% 33% 0.200  Neurodevelopmental disorder 46% 26% 63% 0.009  Affective disorder 26% 39% 19% 0.106 Polysomnographic RMs phenotype  RM sequences (yes) 82% 83% 81% 0.918  RM duration (min) 43.82 (61.70) 54.65 (74.94) 34.59 (47.22) 0.256  RM duration index (% of time in bed) 8.79 (12.60) 11.58 (15.69) 6.40 (8.83) 0.169  RM sequences (Time in bed) 54.18 (84.09) 68.52 (94.57) 41.96 (73.64) 0.270  RM index (/h of time in bed) 6.61 (10.72) 8.90 (12.79) 4.65 (8.33) 0.164  RM sequences (wakefulness) 36.72 (63.15) 45.35 (70.77) 29.37 (56.18) 0.378  RM sequences (N1 sleep) 5.92 (12.63) 7.26 (14.72) 4.78 (10.69) 0.494  RM sequences (N2 sleep) 6.78 (14.02) 8.48 (15.55) 5.33 (12.69) 0.435  RM sequences (N3 sleep) 0.46 (1.55) 0.70 (2.10) 0.26 (0.86) 0.328  RM sequences (REM sleep) 4.3 (13.50) 6.74 (17.73) 2.22 (8.24) 0.242  RM sequences before sleep onset 6.66 (13.36) 8.65 (16.48) 4.96 (10.05) 0.336  RM sequences first third 16.24 (25.06) 21.30 (27.48) 11.93 (22.42) 0.190  RM sequences second third 21.14 (41.23) 24.39 (39.11) 18.37 (43.49) 0.612  RM sequences last third 16.80 (29.51) 22.83 (38.09) 11.67 (18.84) 0.185 *Data not reported for all patients (familial history: adults n = 16, children n = 21; daily occurrence: adults n = 19, children n = 24; late onset: adults n = 21). RM, rhythmic movements; REM, rapid eye movement. Open in new tab At time of study, 23 patients (46%, 17 children) had a comorbid neurodevelopmental disorder, 13 (26%) with ADHD, 2 (4%) with autism spectrum disorder, 5 (10%) with intellectual disabilities, and 15 (30%) with specific learning disorders. Fourteen patients (28%, five children) had an affective disorder, 13 (26%) with anxiety and 6 (12%) with depressive disorder. Sixteen patients (32%, four children) had RLS more than twice a week, with a median age at RLS onset at 17 years (i.e. after RMD onset in 13 patients, 81.2%). Disorder of arousal (i.e. sleepwalking or sleep terrors) was found in 13 patients (26%, 9 children). Finally, only four (8%, three children and one adult) patients with RMD had no other sleep disorder, comorbid neurodevelopmental, or affective disorder (Figure 2). Figure 2. Open in new tabDownload slide Comorbid neurodevelopmental, affective, and sleep disorders in patients with RMD. Figure 2. Open in new tabDownload slide Comorbid neurodevelopmental, affective, and sleep disorders in patients with RMD. Sixteen patients among 37 (43%) reported a positive family history of RMD, 12 with first degree and 4 with second degree relatives, including 3 RMD families with at least 3 affected subjects. No differences between sporadic and familial RM were found for demographic characteristics, comorbid neurodevelopmental, and affective or other sleep disorders. RMs during the polysomnography assessment During the PSG, 41 patients (82%) had at least one RM sequence, 32 (78.0%) with more than 10, and 9 (18%) more than 100 episodes. No RMs were seen in the controls. Altogether, 2,709 RM sequences were analyzed from the PSGs, with a mean duration of 60.75 s per episode, ranging from 3 s to 29.6 min. RM sequences often emerged from wakefulness (67.8% of sequences), and seven (14.0%) patients had only wakefulness-related episodes. The remaining RM sequences emerged from sleep: 33.9% in N1, 38.8% in N2, 2.6% in N3, and 24.6% in rapid eye movement (REM) sleep. Two adult patients had RM sequences only in REM sleep, patients with no comordid condition, no positive RMD family history, no REM sleep without atonia, and no REM sleep behavior disorder. The episodes were regularly distributed throughout the night, 30% during the first part (including 12.3% before sleep onset), 39% during the second part, and 31% during the last part of the night (including 3.1% of episodes after sleep offset). The mean RM index was 6.6 per hour of PSG recording and the mean total time spent in RM 43.8 min (8.8% of time in bed). We found no differences between adults and children for the RMs PSG characteristics (Table 1). Among the 1,632 RM sequences analyzed in a subgroup of patients with available video-PSG, a rolling subtype was found in 1,243 (76.2%) and a rocking/banging subtype in 389 (23.8%). Among the 175 RM sequences emerging from REM sleep, only one headbanging episode was found (0.6%), the remaining sequences were of rolling subtype. In contrast, the rocking and banging subtypes were observed in 26.6% of wakefulness and non-rapid eye movement (NREM) sleep RM sequences. Polysomnographic findings in patients and controls Compared to adult controls, adult patients with RMD had reduced TST and SE, and increased WASO duration, SOL, and microarousal index (Table 2). PLMS index was higher in patients compared to controls (PLMS index > 5/h: 40% vs. 11%, χ2 = 14.350, p < 0.001; PLMS index > 15/h: 22% vs. 1%, χ2 = 14.765, p < 0.001). Patients with moderate OSAS were comparable between the groups; however patients were more likely to have severe OSAS (i.e. AHI > 30/h: 22% vs. 2% χ2 = 7.098, p = 0.008). Table 2. Comparison of polysomnographic characteristics in children and adults with RM disorder and controls . Children . Adults . . Controls (n = 42) . RMD (n = 27) . . Controls (n = 33) . RMD (n = 23) . . . Mean . SD . Mean . SD . P-value . Mean . SD . Mean . SD . P-value . Gender male* 35 83.3% 24 88.9% 0.552 18 54.5% 11 47.8% 0.621 Age (years) 10.46 3.71 8.72 3.52 0.056 35.04 11.87 34.79 11.92 0.940 TST (min) 454.31 66.58 436.78 76.64 0.318 418.51 45.11 324.13 85.78 <0.001 Sleep efficiency (%) 86.27 8.52 80.92 11.97 0.033 85.30 6.85 67.62 16.44 <0.001 Sleep onset latency (min) 29.88 68.56 27.22 21.60 0.846 17.18 11.14 38.30 41.59 0.026 WASO (min) 46.69 34.05 75.00 66.70 0.049 44.54 26.91 114.83 68.85 <0.001 N1 (%) 3.89 2.57 5.17 3.66 0.093 4.19 2.50 8.31 9.50 0.053 N2 (%) 45.54 8.33 45.44 10.05 0.965 57.34 5.70 53.53 11.32 0.103 N3 (%) 31.08 8.94 28.55 8.17 0.240 19.20 5.81 18.17 7.00 0.552 REM (%) 19.47 5.08 20.82 6.06 0.322 19.08 4.46 19.98 6.36 0.537 Microarousal index (/h) 6.31 3.38 12.87 6.35 <0.001 11.90 4.91 19.70 9.93 0.002 AHI (/h) 1.34 1.56 3.75 7.24 0.041 3.18 4.33 7.40 12.58 0.079 Desaturation index (/h) 1.35 1.99 3.77 4.90 0.020 2.82 4.26 9.20 15.20 0.061 PLMS index (/h) 1.02 1.36 6.52 8.02 0.002 2.34 3.66 12.17 18.64 0.020 . Children . Adults . . Controls (n = 42) . RMD (n = 27) . . Controls (n = 33) . RMD (n = 23) . . . Mean . SD . Mean . SD . P-value . Mean . SD . Mean . SD . P-value . Gender male* 35 83.3% 24 88.9% 0.552 18 54.5% 11 47.8% 0.621 Age (years) 10.46 3.71 8.72 3.52 0.056 35.04 11.87 34.79 11.92 0.940 TST (min) 454.31 66.58 436.78 76.64 0.318 418.51 45.11 324.13 85.78 <0.001 Sleep efficiency (%) 86.27 8.52 80.92 11.97 0.033 85.30 6.85 67.62 16.44 <0.001 Sleep onset latency (min) 29.88 68.56 27.22 21.60 0.846 17.18 11.14 38.30 41.59 0.026 WASO (min) 46.69 34.05 75.00 66.70 0.049 44.54 26.91 114.83 68.85 <0.001 N1 (%) 3.89 2.57 5.17 3.66 0.093 4.19 2.50 8.31 9.50 0.053 N2 (%) 45.54 8.33 45.44 10.05 0.965 57.34 5.70 53.53 11.32 0.103 N3 (%) 31.08 8.94 28.55 8.17 0.240 19.20 5.81 18.17 7.00 0.552 REM (%) 19.47 5.08 20.82 6.06 0.322 19.08 4.46 19.98 6.36 0.537 Microarousal index (/h) 6.31 3.38 12.87 6.35 <0.001 11.90 4.91 19.70 9.93 0.002 AHI (/h) 1.34 1.56 3.75 7.24 0.041 3.18 4.33 7.40 12.58 0.079 Desaturation index (/h) 1.35 1.99 3.77 4.90 0.020 2.82 4.26 9.20 15.20 0.061 PLMS index (/h) 1.02 1.36 6.52 8.02 0.002 2.34 3.66 12.17 18.64 0.020 *Values expressed as number and percentages. TST, total sleep time; RMD, rhythmic movement disorder; WASO, wake after sleep onset; REM, rapid eye movement; AHI, apnea–hypopnea index; PLMS, periodic leg movement during sleep. Open in new tab Table 2. Comparison of polysomnographic characteristics in children and adults with RM disorder and controls . Children . Adults . . Controls (n = 42) . RMD (n = 27) . . Controls (n = 33) . RMD (n = 23) . . . Mean . SD . Mean . SD . P-value . Mean . SD . Mean . SD . P-value . Gender male* 35 83.3% 24 88.9% 0.552 18 54.5% 11 47.8% 0.621 Age (years) 10.46 3.71 8.72 3.52 0.056 35.04 11.87 34.79 11.92 0.940 TST (min) 454.31 66.58 436.78 76.64 0.318 418.51 45.11 324.13 85.78 <0.001 Sleep efficiency (%) 86.27 8.52 80.92 11.97 0.033 85.30 6.85 67.62 16.44 <0.001 Sleep onset latency (min) 29.88 68.56 27.22 21.60 0.846 17.18 11.14 38.30 41.59 0.026 WASO (min) 46.69 34.05 75.00 66.70 0.049 44.54 26.91 114.83 68.85 <0.001 N1 (%) 3.89 2.57 5.17 3.66 0.093 4.19 2.50 8.31 9.50 0.053 N2 (%) 45.54 8.33 45.44 10.05 0.965 57.34 5.70 53.53 11.32 0.103 N3 (%) 31.08 8.94 28.55 8.17 0.240 19.20 5.81 18.17 7.00 0.552 REM (%) 19.47 5.08 20.82 6.06 0.322 19.08 4.46 19.98 6.36 0.537 Microarousal index (/h) 6.31 3.38 12.87 6.35 <0.001 11.90 4.91 19.70 9.93 0.002 AHI (/h) 1.34 1.56 3.75 7.24 0.041 3.18 4.33 7.40 12.58 0.079 Desaturation index (/h) 1.35 1.99 3.77 4.90 0.020 2.82 4.26 9.20 15.20 0.061 PLMS index (/h) 1.02 1.36 6.52 8.02 0.002 2.34 3.66 12.17 18.64 0.020 . Children . Adults . . Controls (n = 42) . RMD (n = 27) . . Controls (n = 33) . RMD (n = 23) . . . Mean . SD . Mean . SD . P-value . Mean . SD . Mean . SD . P-value . Gender male* 35 83.3% 24 88.9% 0.552 18 54.5% 11 47.8% 0.621 Age (years) 10.46 3.71 8.72 3.52 0.056 35.04 11.87 34.79 11.92 0.940 TST (min) 454.31 66.58 436.78 76.64 0.318 418.51 45.11 324.13 85.78 <0.001 Sleep efficiency (%) 86.27 8.52 80.92 11.97 0.033 85.30 6.85 67.62 16.44 <0.001 Sleep onset latency (min) 29.88 68.56 27.22 21.60 0.846 17.18 11.14 38.30 41.59 0.026 WASO (min) 46.69 34.05 75.00 66.70 0.049 44.54 26.91 114.83 68.85 <0.001 N1 (%) 3.89 2.57 5.17 3.66 0.093 4.19 2.50 8.31 9.50 0.053 N2 (%) 45.54 8.33 45.44 10.05 0.965 57.34 5.70 53.53 11.32 0.103 N3 (%) 31.08 8.94 28.55 8.17 0.240 19.20 5.81 18.17 7.00 0.552 REM (%) 19.47 5.08 20.82 6.06 0.322 19.08 4.46 19.98 6.36 0.537 Microarousal index (/h) 6.31 3.38 12.87 6.35 <0.001 11.90 4.91 19.70 9.93 0.002 AHI (/h) 1.34 1.56 3.75 7.24 0.041 3.18 4.33 7.40 12.58 0.079 Desaturation index (/h) 1.35 1.99 3.77 4.90 0.020 2.82 4.26 9.20 15.20 0.061 PLMS index (/h) 1.02 1.36 6.52 8.02 0.002 2.34 3.66 12.17 18.64 0.020 *Values expressed as number and percentages. TST, total sleep time; RMD, rhythmic movement disorder; WASO, wake after sleep onset; REM, rapid eye movement; AHI, apnea–hypopnea index; PLMS, periodic leg movement during sleep. Open in new tab Comparing the children groups, patients with RMD had a lower SE, an increased WASO duration, and higher AHI, oxygen desaturation, PLMS, and microarousal indices (Table 2). Children with RMD had higher PLMS (PLMS index>5/h: 41% vs. 2%, χ2 = 16.833, p < 0.001) with similar trend for AHI (without between-group significant differences for having an AHI > 1.5/h: 44% vs. 26%, χ2 = 2.464, p = 0.116). Considering all patients (children and adults), similar differences were observed for TST, SE, WASO, and microarousal index between patients and controls. An increased percentage of N1 sleep, and higher AHI, PLMS, and oxygen desaturation indexes were also found in patients compared to controls. Disturbed nighttime sleep and its association with RM phenotype Considering all patients, RMs index and duration correlated negatively with TST, SE, and positively with WASO, after controlling for age, AHI, PLMS, and RLS (Table 3).The number of RM sequences in N2 and N3 sleep was positively correlated with the microarousal index, and the number of sequences occurring before sleep onset with a longer SOL. Table 3. Correlations between general sleep architecture parameters and RM phenotype in adults and children with RMD. . Total sleep time . Sleep efficiency . Sleep onset latency . Wake after sleep onset . Microarousal index . . Pearson r . P-value . Pearson r . P-value . Pearson r . P-value . Pearson r . P-value . Pearson r . P-value . RM index (/h) −0.400 0.006 −0.439 0.002 0.053 0.726 0.400 0.006 0.065 0.667 RM duration index (%) −0.332 0.024 −0.412 0.004 0.039 0.799 0.416 0.004 0.152 0.314 RM in wakefulness −0.489 0.001 −0.540 <0.001 0.104 0.492 0.476 0.001 −0.028 0.855 RM in N1 sleep −0.226 0.132 −0.263 0.078 0.019 0.899 0.226 0.130 0.043 0.779 RM in N2 sleep −0.099 0.513 −0.167 0.267 −0.044 0.770 0.258 0.083 0.385 0.008 RM in N3 sleep 0.257 0.085 0.210 0.161 −0.084 0.580 −0.096 0.524 0.387 0.008 RM in REM sleep 0.145 0.337 0.167 0.266 −0.163 0.280 −0.052 0.732 0.255 0.087 RM before sleep onset −0.256 0.085 −0.231 0.123 0.365 0.013 0.019 0.898 −0.137 0.365 . Total sleep time . Sleep efficiency . Sleep onset latency . Wake after sleep onset . Microarousal index . . Pearson r . P-value . Pearson r . P-value . Pearson r . P-value . Pearson r . P-value . Pearson r . P-value . RM index (/h) −0.400 0.006 −0.439 0.002 0.053 0.726 0.400 0.006 0.065 0.667 RM duration index (%) −0.332 0.024 −0.412 0.004 0.039 0.799 0.416 0.004 0.152 0.314 RM in wakefulness −0.489 0.001 −0.540 <0.001 0.104 0.492 0.476 0.001 −0.028 0.855 RM in N1 sleep −0.226 0.132 −0.263 0.078 0.019 0.899 0.226 0.130 0.043 0.779 RM in N2 sleep −0.099 0.513 −0.167 0.267 −0.044 0.770 0.258 0.083 0.385 0.008 RM in N3 sleep 0.257 0.085 0.210 0.161 −0.084 0.580 −0.096 0.524 0.387 0.008 RM in REM sleep 0.145 0.337 0.167 0.266 −0.163 0.280 −0.052 0.732 0.255 0.087 RM before sleep onset −0.256 0.085 −0.231 0.123 0.365 0.013 0.019 0.898 −0.137 0.365 Controlling for age, AHI, PLMS, and RLS. RM, rhythmic movement; RMD, rhythmic movement disorder; WASO, wake after sleep onset; AHI, apnea–hyperpnoea index; PLMS, periodic leg movement during sleep; RLS, restless legs syndrome. Open in new tab Table 3. Correlations between general sleep architecture parameters and RM phenotype in adults and children with RMD. . Total sleep time . Sleep efficiency . Sleep onset latency . Wake after sleep onset . Microarousal index . . Pearson r . P-value . Pearson r . P-value . Pearson r . P-value . Pearson r . P-value . Pearson r . P-value . RM index (/h) −0.400 0.006 −0.439 0.002 0.053 0.726 0.400 0.006 0.065 0.667 RM duration index (%) −0.332 0.024 −0.412 0.004 0.039 0.799 0.416 0.004 0.152 0.314 RM in wakefulness −0.489 0.001 −0.540 <0.001 0.104 0.492 0.476 0.001 −0.028 0.855 RM in N1 sleep −0.226 0.132 −0.263 0.078 0.019 0.899 0.226 0.130 0.043 0.779 RM in N2 sleep −0.099 0.513 −0.167 0.267 −0.044 0.770 0.258 0.083 0.385 0.008 RM in N3 sleep 0.257 0.085 0.210 0.161 −0.084 0.580 −0.096 0.524 0.387 0.008 RM in REM sleep 0.145 0.337 0.167 0.266 −0.163 0.280 −0.052 0.732 0.255 0.087 RM before sleep onset −0.256 0.085 −0.231 0.123 0.365 0.013 0.019 0.898 −0.137 0.365 . Total sleep time . Sleep efficiency . Sleep onset latency . Wake after sleep onset . Microarousal index . . Pearson r . P-value . Pearson r . P-value . Pearson r . P-value . Pearson r . P-value . Pearson r . P-value . RM index (/h) −0.400 0.006 −0.439 0.002 0.053 0.726 0.400 0.006 0.065 0.667 RM duration index (%) −0.332 0.024 −0.412 0.004 0.039 0.799 0.416 0.004 0.152 0.314 RM in wakefulness −0.489 0.001 −0.540 <0.001 0.104 0.492 0.476 0.001 −0.028 0.855 RM in N1 sleep −0.226 0.132 −0.263 0.078 0.019 0.899 0.226 0.130 0.043 0.779 RM in N2 sleep −0.099 0.513 −0.167 0.267 −0.044 0.770 0.258 0.083 0.385 0.008 RM in N3 sleep 0.257 0.085 0.210 0.161 −0.084 0.580 −0.096 0.524 0.387 0.008 RM in REM sleep 0.145 0.337 0.167 0.266 −0.163 0.280 −0.052 0.732 0.255 0.087 RM before sleep onset −0.256 0.085 −0.231 0.123 0.365 0.013 0.019 0.898 −0.137 0.365 Controlling for age, AHI, PLMS, and RLS. RM, rhythmic movement; RMD, rhythmic movement disorder; WASO, wake after sleep onset; AHI, apnea–hyperpnoea index; PLMS, periodic leg movement during sleep; RLS, restless legs syndrome. Open in new tab Discussion Here, we performed a case–control study and showed that objective disturbed nighttime sleep was more frequent in children and adult patients with RMD than in respective controls. RMD was highly associated with neurodevelopmental, affective, and other sleep disorders; however, RM characteristics were independently associated with markers of sleep fragmentation. Despite a well-known high prevalence of RMs in the general population in children [3, 4], only 50 patients with diagnostic criteria for RMD were identified over 13 years in a single reference national sleep center with half of our population being adults. This observation may confirm that RMs are frequent and mild in children and rarely require an evaluation in a specialized sleep disorder clinic. On the other hand, RMD is a less frequent and understudied condition, being the most serious and debilitating form of RMs. Symptoms of RMD were often comparable between children and adults suggesting a similar underlying pathology. While RMs in children tend to resolve spontaneously, without consequences, our sample of children and adults had persistent RMs with nighttime and or daytime impairment [39]. We observed a male RMD preponderance, but exclusively in children (89%) that may reflect a bias selection as most of epidemiological studies reported no association between RM, RMD, and gender [4, 40]. The diagnosis of RMs should rely on visual analysis of video synchronized to PSG or home videosomnography [2, 38, 41, 42]. Our procedure of assessing RMs first in video-PSG, and second in PSG only (blinded of the video) confirmed the excellent concordance (92.8%) between both methods, with highly stereotyped motor artifacts observed on most of channels during RM sequences. However, video-PSG remains mandatory to differentiate the various subtypes of RM. None of the controls and 41 patients (82.0%) displayed RMs during one night video-PSG recording, suggesting that some patients may suppress such movements in the laboratory setting [43], or that RMs may have a night-to-night variability as described for PLMs [44]. Most of RMs emerged from wakefulness, with a majority of sequences being observed after sleep onset. The repartition of the RM sequences was spread out across the night, not predominantly in the first part of the night that further challenge the historical view of RMD as a predormitum sleep-wake transition phenomenon [19]. In NREM sleep, we confirmed that RMs occurred mainly during light N1 and N2 sleep, with rare episodes emerging from N3 sleep [43]. The decrease in the frequency of RMs with NREM stages in proportion to the depth of sleep has previously been reported [43]. In two patients, RMs occurred only in REM sleep episodes, without associated REM sleep behavior disorder or other comorbidities as already reported [6, 9, 18, 19, 27]. We also confirmed the tendency toward rolling RMs subtype in REM sleep [43]. We found that both children and adults with RMD had an excessive sleep fragmentation, characterized by a reduced SE, and a higher microarousal index compared to controls, in line with a 9 year old boy report studying interactions between arousal mechanisms and RMs through the analysis of cyclic alternating patterns [45]. Our results revealed that differences on disturbed nighttime sleep were larger in the adult subsample, with almost 20% reduced SE and a twice-longer WASO duration, compared to adult controls. A shorter TST and an increased SOL were also found in this age group. These results were associated with both the RM index and duration although RMs were not scored as wakefulness but as “movement time.” In our study, RMD often co-occurred with affective, neurodevelopmental, or other sleep disorder. Comorbid neurodevelopmental disorders were found especially in children with RMs that confirms earlier studies [8, 25, 27, 43]. In the general population, RMs and RMD decrease as function of age, as for neurodevelopmental disorders. These findings would suggest that RMs reflect a delayed maturation of motor control during sleep, with possible implication of the vestibular system and the central pattern generator network, as hypothesized [8, 46]. We also found a high frequency of affective disorders, namely depression and anxiety disorders in both adults and children with RMD [47]. These psychiatric conditions are known to interfere with sleep initiation and maintenance. In this context, RMD could be considered as a learned behavior to facilitate sleep onset and return to sleep following awakenings, by reproducing parental rocking known to sooth infants and decrease arousal in experimental paradigms. Whether RMs directly disrupt sleep continuity or are soothing behaviors that partially prevent from more severe sleep fragmentation remains unknown. We found a high frequency of RLS, PLMS, and sleep-disordered breathing in patients with RMD, also underlined by previous studies, that may worsen the disrupted nighttime sleep reported this population [29, 32, 48, 49]. We found that RM severity was associated with poorer SE, increased WASO duration, longer SOL, and increased microarousal index even after controlling for RLS, PLMS, and AHI. Altogether, these results suggest that the severity of RMD influences PSG continuity and architecture parameters, regardless of age and comorbid sleep disorders. Only four patients (8%) of the population of patients with RMD had no comorbidities suggesting that a primary (idiopathic) RMD would be a rare condition. In contrast, we found a high frequency of a positive family history of RM (43%), without differences in phenotyping between sporadic and familial RMD. Familial occurrence of RM has been occasionally reported with a frequency from 8.0% to 20.0% [8, 25, 43, 50, 51]. The higher familial occurrence we observed does not mean that RMD has a genetic origin, given that RM could be a learned behavior from the parents or siblings. Another study reported a multigenerational family with comorbid RMD and insomnia and proposed that there might be a common genetic predisposition for both disorders [51]. However, to our knowledge, no genetic investigations on RMD have not been yet performed. Some limitations need to be acknowledged. Our clinical population from in a tertiary sleep clinic was well-characterized but relatively small with potential for bias toward a most severe condition than the one from the general population that prevent the generalization of our main results. Due to the small sample size, we were unable to perform sub-analyses comparing clinical and polysomnographic characteristics based on the presence of associated comorbidities. Due to the artifacts preventing a precise analysis of the beginning and the end of the various events on the PSG, we were not able to specify the sequence of occurrence of micro-arousals, leg movements or respiratory events related to RMs. This limitation may prevent understanding of how RMs interact with other events during sleep and cause disturbed nighttime sleep. Furthermore, we could not exclude that other comorbid conditions such as neurodevelopmental or affective disorders, often associated with sleep disturbances, would have modified the association between RMD and disturbed nighttime sleep. As patients were retrospectively identified, with limited storage capacity in past years, not all of the video recordings were systematically backed up. However, the agreement between the standard method for the diagnosis of RMs on video-PSG versus on PSG only was high, in favor of the reliability of the full reported data. Finally, we did not use standardized psychometric tools to quantify the complaints of disturbed nighttime sleep and daytime functioning in patients. In conclusion, we confirmed that RMD is a potential disabling condition with objective disturbed nighttime sleep. The severity of RMD and frequent associated sleep-related motor and respiratory disorders may aggravate the global altered sleep continuity. Patients with a primary RMD complaint should be screened for another sleep disorder but also for non-sleep-related comorbidities. Disclosure Statements Financial Disclosure: C.L. received McGill University International Graduate Mobility Award (T244294C0G). M.P. received Fonds de recherche du Québec-Santé. A.R. and E.E.travel to congress from Laidet medical. L.B. received honoraria for speaking from UCB Pharma. S.C.—None. Y.D. received honoraria for speaking and board engagements from UCB Pharma, Jazz, Bioprojet, Theranexus, Takeda, Idorsia. R.L. received honoraria for speaking from UCB Pharma, Shire; travel to congress from Laidet medical. Non-financial disclosure: None declared. References 1. American Academy of Sleep Medicine . International Classification of Sleep Disorders, Third Edition: Diagnostic and Coding Manual . Darien, IL : American Academy of Sleep Medicine ; 2014 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC 2. Gogo E , et al. Objectively confirmed prevalence of sleep-related rhythmic movement disorder in pre-school children . Sleep Med. 2019 ; 53 : 16 – 21 . Google Scholar Crossref Search ADS PubMed WorldCat 3. Klackenberg G . A prospective longitudinal study of children: data on psychic health and development up to 8 years of age . Acta Paediatr Scand . 1971 ; 224 ( Suppl ): 73 – 83 . Google Scholar OpenURL Placeholder Text WorldCat 4. Laberge L , et al. Development of parasomnias from childhood to early adolescence . Pediatrics. 2000 ; 106 ( 1 Pt 1 ): 67 – 74 . Google Scholar Crossref Search ADS PubMed WorldCat 5. Alves RS , et al. Jactatio capitis nocturna with persistence in adulthood. case report . Arq Neuropsiquiatr. 1998 ; 56 ( 3B ): 655 – 657 . Google Scholar Crossref Search ADS PubMed WorldCat 6. Gagnon P , et al. Repetitive head movements during REM sleep . Biol Psychiatry. 1985 ; 20 ( 2 ): 176 – 178 . Google Scholar Crossref Search ADS PubMed WorldCat 7. Reimão R , et al. [Jactatio capitis nocturnus: report of a case in an adult] . Arq Neuropsiquiatr. 1985 ; 43 ( 1 ): 86 – 90 . Google Scholar Crossref Search ADS PubMed WorldCat 8. Stepanova I , et al. Rhythmic movement disorder in sleep persisting into childhood and adulthood . Sleep. 2005 ; 28 ( 7 ): 851 – 857 . Google Scholar Crossref Search ADS PubMed WorldCat 9. Whyte J , et al. A self-destructive variant of jactatio capitis nocturna . J Nerv Ment Dis. 1991 ; 179 ( 1 ): 49 – 50 . Google Scholar Crossref Search ADS PubMed WorldCat 10. Bramble D . Two cases of severe head-banging parasomnias in peripubertal males resulting from otitis media in toddlerhood . Child Care Health Dev. 1995 ; 21 ( 4 ): 247 – 253 . Google Scholar Crossref Search ADS PubMed WorldCat 11. Spalter HF , et al. Cataracts following chronic headbanging . Arch Ophthal . 1970 ; 83 : 182 – 186 . Google Scholar Crossref Search ADS PubMed WorldCat 12. Stuck KJ , et al. Large skull defect in a headbanger . Pediatr Radiol. 1979 ; 8 ( 4 ): 257 – 258 . Google Scholar Crossref Search ADS PubMed WorldCat 13. Vogel W , et al. Citalopram for head-banging . J Am Acad Child Adolesc Psychiatry. 2000 ; 39 ( 5 ): 544 – 545 . Google Scholar Crossref Search ADS PubMed WorldCat 14. Gupta R , et al. Head banging persisting during adolescence: a case with polysomnographic findings . J Neurosci Rural Pract. 2014 ; 5 ( 4 ): 405 – 408 . Google Scholar Crossref Search ADS PubMed WorldCat 15. Evans J . Rocking at night . J Child Psychol Psychiatry. 1961 ; 2 : 71 – 85 . Google Scholar Crossref Search ADS PubMed WorldCat 16. Golding K . Nocturnal headbanging as a settling habit: the behavioural treatment of a 4-year-old boy . Clin Child Psychol Psychiatry . 1998 ; 3 ( 1 ): 25 – 30 . Google Scholar Crossref Search ADS WorldCat 17. Hashizume Y , et al. Case of head banging that continued to adolescence . Psychiatry Clin Neurosci. 2002 ; 56 ( 3 ): 255 – 256 . Google Scholar Crossref Search ADS PubMed WorldCat 18. Kempenaers C , et al. A rhythmic movement disorder in REM sleep: a case report . Sleep. 1994 ; 17 ( 3 ): 274 – 279 . Google Scholar Crossref Search ADS PubMed WorldCat 19. Kohyama J , et al. Rhythmic movement disorder: polysomnographic study and summary of reported cases . Brain Dev. 2002 ; 24 ( 1 ): 33 – 38 . Google Scholar Crossref Search ADS PubMed WorldCat 20. Morkous SS , et al. A Young Man Who Rocks and Rolls at Night . Ann Am Thorac Soc. 2015 ; 12 ( 9 ): 1413 – 1415 . Google Scholar Crossref Search ADS PubMed WorldCat 21. Rosenberg C . Elimination of a rhythmic movement disorder with hypnosis—a case report . Sleep. 1995 ; 18 ( 7 ): 608 – 609 . Google Scholar Crossref Search ADS PubMed WorldCat 22. Ross RR , et al. Treatment of nocturnal heandbanging by behavior modification techniques: a case report . Behav Res Therapy . 1971 ; 9 : 151 – 154 . Google Scholar Crossref Search ADS WorldCat 23. Yeh SB , et al. Atypical headbanging presentation of idiopathic sleep related rhythmic movement disorder: three cases with video-polysomnographic documentation . J Clin Sleep Med. 2012 ; 8 ( 4 ): 403 – 411 . Google Scholar Crossref Search ADS PubMed WorldCat 24. Walsh JK , et al. A case report of jactatio capitis nocturna . Am J Psychiatry. 1981 ; 138 ( 4 ): 524 – 526 . Google Scholar Crossref Search ADS PubMed WorldCat 25. Mayer G , et al. Sleep related rhythmic movement disorder revisited . J Sleep Res. 2007 ; 16 ( 1 ): 110 – 116 . Google Scholar Crossref Search ADS PubMed WorldCat 26. Xu Z , et al. Association of idiopathic rapid eye movement sleep behavior disorder in an adult with persistent, childhood onset rhythmic movement disorder . J Clin Sleep Med. 2009 ; 5 ( 4 ): 374 – 375 . Google Scholar Crossref Search ADS PubMed WorldCat 27. Mattewal A , et al. A child with REM sleep disturbance . J Clin Sleep Med. 2010 ; 6 ( 1 ): 97 – 100 . Google Scholar Crossref Search ADS PubMed WorldCat 28. Chiaro G , et al. Sleep-related rhythmic movement disorder and obstructive sleep apnea in five adult patients . J Clin Sleep Med. 2017 ; 13 ( 10 ): 1213 – 1217 . Google Scholar Crossref Search ADS PubMed WorldCat 29. Gharagozlou P , et al. Rhythmic movement disorder associated with respiratory arousals and improved by CPAP titration in a patient with restless legs syndrome and sleep apnea . Sleep Med. 2009 ; 10 ( 4 ): 501 – 503 . Google Scholar Crossref Search ADS PubMed WorldCat 30. Owens J , et al. Incidence of parasomnias in children with obstructive sleep apnea . Sleep. 1997 ; 20 ( 12 ): 1193 – 1196 . Google Scholar PubMed OpenURL Placeholder Text WorldCat 31. Urrestarazu E , et al. Body rolling in adults associated to obstructive sleep apnea . J Sleep Med Disord . 2016 ; 3 ( 3 ): 1049 . Google Scholar OpenURL Placeholder Text WorldCat 32. Walters AS , et al. Frequent occurrence of myoclonus while awake and at rest, body rocking and marching in place in a subpopulation of patients with restless legs syndrome . Acta Neurol Scand. 1988 ; 77 ( 5 ): 418 – 421 . Google Scholar Crossref Search ADS PubMed WorldCat 33. Iber C , et al. The AASM Manual for the Scoring of Sleep and Associated Events: Rules, Terminology and Technical Specifications . Vol 1 . Westchester, IL : American Academy of Sleep Medicine ; 2007 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC 34. Rechtschaffen A , et al. . eds. A Manual of Standardized Terminology, Techniques and Scoring System for Sleep Stages of Human Subjects . University of California, Los Angeles. Bethesda, MD: Brain Information Service: Neurological Information Network ; 1968 . Blindness UNIoNDa , ed. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC 35. Marcus CL , et al. Diagnosis and management of childhood obstructive sleep apnea syndrome . Pediatrics. 2012 ; 130 ( 3 ): e714 – e755 . Google Scholar Crossref Search ADS PubMed WorldCat 36. Ferri R , et al. World Association of Sleep Medicine (WASM) 2016 standards for recording and scoring leg movements in polysomnograms developed by a joint task force from the International and the European Restless Legs Syndrome Study Groups (IRLSSG and EURLSSG) . Sleep Med. 2016 ; 26 : 86 – 95 . Google Scholar Crossref Search ADS PubMed WorldCat 37. Berry RB , et al. The AASM Manual for the Scoring of Sleep and Associated Events: Rules, Terminology and Technical Specifications. Version 2.3 . Darien, IL: American Academy of Sleep Medicine ; 2016 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC 38. Gall M , et al. A novel approach to assess sleep-related rhythmic movement disorder in children using automatic 3D analysis . Front Psychiatry. 2019 ; 10 : 709 . Google Scholar Crossref Search ADS PubMed WorldCat 39. De Lissovoy V . Head banging in early childhood: a suggested cause . J Genet Psychol. 1963 ; 102 : 109 – 114 . Google Scholar Crossref Search ADS PubMed WorldCat 40. Petit D , et al. Dyssomnias and parasomnias in early childhood . Pediatrics. 2007 ; 119 ( 5 ): e1016 – e1025 . Google Scholar Crossref Search ADS PubMed WorldCat 41. Dyken ME , et al. Diagnosing rhythmic movement disorder with video-polysomnography . Pediatr Neurol. 1997 ; 16 ( 1 ): 37 – 41 . Google Scholar Crossref Search ADS PubMed WorldCat 42. Gwyther A , et al. Videosomnography assessment in childhood rhythmic movement disorder: validation of a novel scoring algorithm . J Sleep Res . 2016 ; 25 : 216 – 217 . Google Scholar Crossref Search ADS PubMed WorldCat 43. Prihodova I , et al. Sleep-related rhythmic movements and rhythmic movement disorder beyond early childhood . Sleep Med. 2019 ; 64 : 112 – 115 . Google Scholar Crossref Search ADS PubMed WorldCat 44. Sforza E , et al. Night-to-night variability in periodic leg movements in patients with restless legs syndrome . Sleep Med. 2005 ; 6 ( 3 ): 259 – 267 . Google Scholar Crossref Search ADS PubMed WorldCat 45. Manni R , et al. Rhythmic movement disorder and cyclic alternating pattern during sleep: a video‐polysomnographic study in a 9‐year‐old boy . Mov Disord . 2004 ; 19 ( 10 ): 1186 – 1190 . Google Scholar Crossref Search ADS PubMed WorldCat 46. Manni R , et al. Rhythmic movements during sleep: a physiological and pathological profile . Neurol Sci. 2005 ; 26 ( Suppl 3 ): s181 – s185 . Google Scholar Crossref Search ADS PubMed WorldCat 47. Laganière C , et al. Maternal characteristics and behavioural/emotional problems in preschoolers: how they relate to sleep rhythmic movements at sleep onset . J Sleep Res. 2019 ; 28 ( 3 ): e12707 . Google Scholar Crossref Search ADS PubMed WorldCat 48. Lombardi C , et al. Pelvic movements as rhythmic motor manifestation associated with restless legs syndrome . Mov Disord. 2003 ; 18 ( 1 ): 110 – 113 . Google Scholar Crossref Search ADS PubMed WorldCat 49. Vitello N , et al. Rhythmic movement disorder associated with restless legs syndrome . Sleep Med. 2012 ; 13 ( 10 ): 1324 – 1325 . Google Scholar Crossref Search ADS PubMed WorldCat 50. Hayward-Koennecke HK , et al. Sleep-related rhythmic movement disorder in triplets: evidence for genetic predisposition? J Clin Sleep Med. 2019 ; 15 ( 1 ): 157 – 158 . Google Scholar Crossref Search ADS PubMed WorldCat 51. Attarian H , et al. A multigenerational family with persistent sleep related rhythmic movement disorder (RMD) and insomnia . J Clin Sleep Med. 2009 ; 5 ( 6 ): 571 – 572 . Google Scholar Crossref Search ADS PubMed WorldCat © Sleep Research Society 2020. Published by Oxford University Press on behalf of the Sleep Research Society. All rights reserved. For permissions, please e-mail [email protected]. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)

Nikkonen, Sami; Korkalainen, Henri; Kainulainen, Samu; Myllymaa, Sami; Leino, Akseli; Kalevo, Laura; Oksenberg, Arie; Leppänen, Timo; Töyräs, Juha

doi: 10.1093/sleep/zsaa106pmid: 32459856

Abstract A common symptom of obstructive sleep apnea (OSA) is excessive daytime sleepiness (EDS). The gold standard test for EDS is the multiple sleep latency test (MSLT). However, due to its high cost, MSLT is not routinely conducted for OSA patients and EDS is instead evaluated using sleep questionnaires. This is problematic however, since sleep questionnaires are subjective and correlate poorly with the MSLT. Therefore, new objective tools are needed for reliable evaluation of EDS. The aim of this study was to test our hypothesis that EDS can be estimated with neural network analysis of previous night polysomnographic signals. We trained a convolutional neural network (CNN) classifier using electroencephalography, electrooculography, and chin electromyography signals from 2,014 patients with suspected OSA. The CNN was trained to classify the patients into four sleepiness categories based on their mean sleep latency (MSL); severe (MSL < 5min), moderate (5 ≤ MSL < 10), mild (10 ≤ MSL < 15), and normal (MSL ≥ 15). The CNN classified patients to the four sleepiness categories with an overall accuracy of 60.6% and Cohen’s kappa value of 0.464. In two-group classification scheme with sleepy (MSL < 10 min) and non-sleepy (MSL ≥ 10) patients, the CNN achieved an accuracy of 77.2%, with sensitivity of 76.5%, and specificity of 77.9%. Our results show that previous night’s polysomnographic signals can be used for objective estimation of EDS with at least moderate accuracy. Since the diagnosis of OSA is currently confirmed by polysomnography, the classifier could be used simultaneously to get an objective estimate of the daytime sleepiness with minimal extra workload. daytime sleepiness, obstructive sleep apnea, MSLT, EEG Statement of Significance Daytime sleepiness is a common symptom of obstructive sleep apnea (OSA), but it is somewhat ignored in sleep apnea diagnostics and treatment planning since the multiple sleep latency test is not routinely conducted for sleep apnea patients. The convolutional neural network classifier developed in this study enables the estimation of objective daytime sleepiness for OSA patients using signals recorded during polysomnography. Therefore, a reasonably accurate sleepiness estimation can be acquired without the need to conduct any additional tests. The only currently available alternatives are subjective sleep questionnaires, such as Epworth Sleepiness Scale, which the developed classifier slightly outperforms. The accuracy of the classifier could be further improved in the future with broader training material. Introduction Obstructive sleep apnea (OSA) is a common sleep disorder affecting approximately half of the adult population [1, 2]. A major symptom of OSA is excessive daytime sleepiness (EDS). Although EDS is not directly lethal, it has a significant deteriorating impact on the quality of life causing depression and cognitive impairment [3–5]. In addition, EDS is a major cause of motor vehicle accidents and sick leaves making it a substantial economic burden [6]. The gold-standard test for EDS is the multiple sleep latency test (MSLT) [7]. The MSLT is an objective, full-day trial performed in a sleep laboratory where sleep latency is measured multiple times and the average of these latencies, that is, mean sleep latency (MSL) is used to assess EDS [7]. The subjects are clinically classified into four sleepiness categories based on their MSL: severe (MSL < 5 min), moderate (5 ≤ MSL < 10), mild (10 ≤ MSL < 15), and normal (MSL ≥ 15) [8, 9]. Alternatively, a single MSL threshold of 8 or 10 min is often used to differentiate between normal patients and patients suffering from EDS [10, 11]. However, as MSLT is time-consuming and expensive, it is not routinely performed for OSA patients and EDS is instead evaluated using simpler tests, such as sleep questionnaires [12, 13]. Sleep questionnaires are problematic however, since they are dependent on the patients’ interpretation of the rating system and therefore only offer an estimation on subjective sleepiness. For example, the results of the most common subjective test, the Epworth Sleepiness Scale (ESS), do not correlate well with MSLT and have been proven to be insufficient in estimating daytime sleepiness [12, 14–16]. Due to these shortcomings, simpler and easier objective tools are needed for evaluation of EDS especially for OSA patients. Machine learning has been proven to be a powerful tool in medical signal analysis and has also shown promise in automatic diagnostics of OSA [17–19]. For example, artificial neural networks have been used for automated sleep staging using electroencephalography (EEG) [20, 21]. Based on the promising previous research, we hypothesized that previous night EEG could be used to estimate the daytime sleepiness of an OSA patient. Therefore, the aim of this study was to develop an objective, neural network method for estimation of EDS in patients with suspected OSA. We test our hypothesis by training a convolutional neural network (CNN) that estimates the results of the MSLT based on the previous night’s EEG, electrooculography (EOG), and electromyography (EMG) signals. We chose to use a convolutional neural network, which is a type of deep neural network inspired by the human visual cortex and developed specifically for visual machine learning tasks [22]. CNNs also have less parameters and are faster to train than equally sized multilayer perceptron networks which is important with large inputs such as high resolution images. Like in regular multilayer perceptron networks, layers of a CNN have neurons, which receive inputs, calculate a weighed sum from them according to the learnable weights, pass them through an activation function and generate an output. However, in CNNs, the layers are not fully connected and instead only a small part of the input layer is being operated on by the convolution kernel at a time. This kernel is then moved over the whole input layer generating the full output. Methods We developed a convolutional neural network (CNN) classifier to automatically estimate the MSLT result using EEG, EOG, and chin EMG signals recorded during in-lab polysomnography (PSG) the previous night. The CNN classifier was trained to classify the patients with suspected OSA into four sleepiness categories based on their MSL; severe (MSL < 5 min), moderate (5 ≤ MSL < 10), mild (10 ≤ MSL < 15), and normal (MSL ≥ 15). Additionally, we classified patients to EDS and normal groups using an MSL < 10 min as the threshold for EDS. Dataset The patient population consisted of 2,014 patients with suspected OSA who had undergone in-lab PSG and a next day MSLT (Table 1). The recordings were conducted during 2001–2011 in the Sleep Disorders Unit, Loewenstein Hospital—Rehabilitation Center, Raanana, Israel and analyzed using the prevailing American Academy of Sleep Medicine (AASM) guidelines [23, 24]. According to the clinical protocol at the Loewenstein Hospital, patients were referred to the MSLT because they had complained of daytime sleepiness during the clinical interview. No preliminary sleep questionnaires were performed. Ethical permission was obtained from the Ethical Committee of Loewenstein Hospital (Permission number: 0006-17-LOE). The MSLTs were conducted using four-nap protocol in uninterrupted conditions with 2 h intervals between each nap attempt [25]. The sleep onset was determined from the first stage of sleep. If no sleep occurred, the nap attempt was terminated at 20 min and the sleep latency was determined to be 20 min for that nap attempt. A total of four nap attempts were conducted and the MSL was calculated as the mean of these four readings. Table 1. Subject characteristics . Mean . Range . SD . Age (years) 50.9 18.0–88.0 13.8 BMI (kg/m2) 30.8 13.8–63.7 6.4 AHI 1/h 30.0 0.3–148.1 28.9 MSL (min) 10.2 0.5–20.0 5.1 Recording duration (h) 7.2 6.0–8.7 0.4 Number Percentage Total number of patients 2,014  Male patients 1,492 74.1  Female patients 522 25.9 EDS category  Normal 368 18.3  Mild 649 32.2  Moderate 580 28.8  Severe 417 20.7 OSA category  Normal 401 19.9  Mild 438 21.8  Moderate 422 21.0  Severe 753 37.4 . Mean . Range . SD . Age (years) 50.9 18.0–88.0 13.8 BMI (kg/m2) 30.8 13.8–63.7 6.4 AHI 1/h 30.0 0.3–148.1 28.9 MSL (min) 10.2 0.5–20.0 5.1 Recording duration (h) 7.2 6.0–8.7 0.4 Number Percentage Total number of patients 2,014  Male patients 1,492 74.1  Female patients 522 25.9 EDS category  Normal 368 18.3  Mild 649 32.2  Moderate 580 28.8  Severe 417 20.7 OSA category  Normal 401 19.9  Mild 438 21.8  Moderate 422 21.0  Severe 753 37.4 Number and percentage for categorical variables and mean, range and standard deviation for continuous variables. EDS, excessive daytime sleepiness; OSA, obstructive sleep apnea; BMI, body mass index; AHI, apnea–hypopnea index; MSL, mean sleep latency; SD, standard deviation. Open in new tab Table 1. Subject characteristics . Mean . Range . SD . Age (years) 50.9 18.0–88.0 13.8 BMI (kg/m2) 30.8 13.8–63.7 6.4 AHI 1/h 30.0 0.3–148.1 28.9 MSL (min) 10.2 0.5–20.0 5.1 Recording duration (h) 7.2 6.0–8.7 0.4 Number Percentage Total number of patients 2,014  Male patients 1,492 74.1  Female patients 522 25.9 EDS category  Normal 368 18.3  Mild 649 32.2  Moderate 580 28.8  Severe 417 20.7 OSA category  Normal 401 19.9  Mild 438 21.8  Moderate 422 21.0  Severe 753 37.4 . Mean . Range . SD . Age (years) 50.9 18.0–88.0 13.8 BMI (kg/m2) 30.8 13.8–63.7 6.4 AHI 1/h 30.0 0.3–148.1 28.9 MSL (min) 10.2 0.5–20.0 5.1 Recording duration (h) 7.2 6.0–8.7 0.4 Number Percentage Total number of patients 2,014  Male patients 1,492 74.1  Female patients 522 25.9 EDS category  Normal 368 18.3  Mild 649 32.2  Moderate 580 28.8  Severe 417 20.7 OSA category  Normal 401 19.9  Mild 438 21.8  Moderate 422 21.0  Severe 753 37.4 Number and percentage for categorical variables and mean, range and standard deviation for continuous variables. EDS, excessive daytime sleepiness; OSA, obstructive sleep apnea; BMI, body mass index; AHI, apnea–hypopnea index; MSL, mean sleep latency; SD, standard deviation. Open in new tab The EEG electrodes were placed according to the international 10–20 system [26]. Two EEG channels, C4-A1 and PZ-A1, EOG channel (ROC-A1), and chin EMG were used as an input to the CNN. AASM recommends C4-A1, F4-A1, and O2-A1 EEG channels together with EOG and EMG channels for sleep staging [24]. However, as our dataset included frontal and occipital EEG channels only for a very limited number of patients, we chose to use C4 and PZ channels because they were most frequently recorded among the patients and thus the highest possible number of patients could be included in the study. We also chose to include the EOG and chin EMG channels since according to our preliminary testing, slightly better results were obtained with all four channels compared to using EEG only. Signal processing The raw signals sampled at 256 Hz frequency were exported from REMbrandt Manager System (MedCare Co, Amsterdam, the Netherlands) and imported to MATLAB 2018b (MathWorks Inc., Natick, Massachusetts, USA), which was used to conduct all preprocessing tasks. The signals were truncated so that only the time between the lights off mark and lights on mark was included and normalized using z-score normalization, that is, subtracting the signal mean and dividing by standard deviation resulting in a signal with zero mean and a standard deviation of one. The normalization was done to unify the greatly varying signal amplitudes between different patients. The signals were divided into 512 epochs with 50% overlap. No padding at the start or end was used. As a result, each epoch length was 2/513rds of the time between lights off and lights on marks. Welch’s power spectral density (PSD) estimate [27] was then calculated for each epoch using 8 windows with 50% overlap. The PSD estimate was calculated for a frequency range of 0.3–30.3Hz using 512 data points. This frequency range was chosen since it contains the common diagnostic bands (Delta, Theta, Alpha, and Beta) and, for example, AASM recommends filtering out frequencies outside 0.3–35Hz when scoring sleep [28, 29]. The PSD estimates were converted to dB scale (xdb =10 log10x) and arranged into a 512 × 512 spectrogram image where one column corresponds to one epoch. The same procedure was repeated for all four channels and for each patient. Finally, all spectrograms were arranged into a 2,014 × 512 × 512 × 4 matrix where the first dimension represents patients, the second and third dimensions represent the spectrograms, and the fourth dimension represents the four signal channels. Example figure of the spectrograms are presented in Figure 1. Figure 1. Open in new tabDownload slide Example of the spectrograms given to the convolutional neural network as an input. Figure 1. Open in new tabDownload slide Example of the spectrograms given to the convolutional neural network as an input. Neural network The CNN was trained in Python 3.7.3 with Tensorflow 1.14.0 using Keras 2.2.4. The CNN consisted of four convolutional blocks and one fully connected block (Figure 2). Each convolutional block consisted of two 2D-convolution layers followed by a max pooling layer with a pool size of 2 × 2 and a stride of 2–2. All convolution layers used 3 × 3 convolution kernels, stride of 1–1 and a tanh activation function. The number of output filters of the convolutional layers was 12 in the first block, 18 in the second block, 24 in the third block, and 30 in the fourth block. The last block consisted of a dropout layer with a 0.3 dropout followed by a flattening layer and two fully connected layers with layer sizes of 4 and 12 and a ReLU activation. The last layer, that is, the output layer, was a fully connected layer with a size of 4 and a softmax activation. The network was trained with the Adam optimizer using a learning rate of 0.0001. Different neural network structures were also tested but they resulted in worse performance (see Table S2). The network with the lowest mean validation set loss was selected from the tested networks. We used class weighting during training to mitigate the effect of imbalanced classes. Each class weight was set to be inversely proportional to the number of patients in the class. Figure 2. Open in new tabDownload slide Structure of the convolutional neural network. Figure 2. Open in new tabDownload slide Structure of the convolutional neural network. We used 10-fold cross-validation to test the performance of the classifier. The patient population was randomly divided into 10 subpopulations, each consisting of 201 or 202 patients. The CNN was trained 10 times such that each subpopulation was used once as a test set, and 9 times in the training set. During each fold, 10% of the training set was further used as the validation set to assess the performance during training and to avoid overfitting. The training accuracy was monitored during training using sparse categorical cross-entropy as the loss function. The training was stopped after the validation set loss did not decrease for 100 continuous epochs after which the model with the lowest validation loss was selected as the model for that fold. To further interpret the model, we performed an occlusion test to estimate the relative importance of different parts of the spectrogram. A 32 × 32 mask, that sets the spectrogram values under the mask to zero, was used to occlude part of the spectrogram and these occluded spectrograms were given as an input to the trained classifier. The process was repeated by moving the mask over the whole spectrogram with no overlap resulting in a total of 256 occlusions. The accuracy of the classifier was then calculated for each occlusion. Results By using a single, 10-min, threshold for EDS classification, the classifier achieved an accuracy of 77.2% in differentiating sleepy and non-sleepy patients with suspected OSA. Sensitivity and specificity of the classifier were 76.5% and 77.9%, respectively. The receiver operating characteristic (ROC) curves for the classifier in each fold and across all folds are presented in Figure 3. The area under ROC curve (AUC) for the classifier across all folds was 0.853. The classifier achieved a positive predictive value of 78.0% and negative predictive value of 76.5%. Cohen’s kappa [30] value for the binary classification was 0.544 and F1-score was 0.772. Figure 3. Open in new tabDownload slide ROC curves for the classifier in each fold and across all folds. Figure 3. Open in new tabDownload slide ROC curves for the classifier in each fold and across all folds. When classifying patients to the four sleepiness categories, the CNN achieved an overall accuracy of 60.6%. Cohen’s kappa [30] value for the classifier was 0.464. The training, validation, and test set accuracies varied slightly between the folds. Mean training, validation, and test accuracies were 70.7%, 61.1%, and 60.6% with standard deviations of 4.5%, 7.3%, and 8.3%, respectively (see Table S1 for full cross-validation statistics). Confusion matrix showing the patient classification across all folds is presented in Figure 4. The CNN performed best in the moderate sleepiness category with an accuracy of 66.9% and worst in the normal category with an accuracy of 52.0%. Figure 4. Open in new tabDownload slide Confusion matrix showing the classification accuracy of the convolutional neural network classifier across all folds. Figure 4. Open in new tabDownload slide Confusion matrix showing the classification accuracy of the convolutional neural network classifier across all folds. To assess which group of patients is most likely to be classified correctly, the classification accuracy was compared in age, sex, BMI, and AHI subgroups (Table 1). The classification accuracy varied slightly between the subgroups. Patients with severe OSA were slightly more likely to be classified correctly than patients with lesser severity of OSA. Patients with higher BMI or age were also classified slightly more accurately than patients with low BMI or age. The results of the occlusion test when classifying the patients to the four sleepiness categories are presented in Figure 5. The accuracy varied greatly between the occlusions. Occluding the lower frequencies (0–15Hz) had slightly more detrimental effect on the accuracy of the classifier than occluding the higher frequencies. Figure 5. Open in new tabDownload slide Occlusion plots for the convolutional neural network classifier when classifying patients to the four sleepiness categories. All 32 × 32 occlusions (A) showing the difference in classification accuracy when the corresponding area of the input spectrograms are occluded. Time average of the occlusions (B) showing the average drop in accuracy for each frequency. Brighter color corresponds to a larger drop in accuracy, that is, the occluded area is more important, and darker color corresponds to a smaller drop in accuracy. Figure 5. Open in new tabDownload slide Occlusion plots for the convolutional neural network classifier when classifying patients to the four sleepiness categories. All 32 × 32 occlusions (A) showing the difference in classification accuracy when the corresponding area of the input spectrograms are occluded. Time average of the occlusions (B) showing the average drop in accuracy for each frequency. Brighter color corresponds to a larger drop in accuracy, that is, the occluded area is more important, and darker color corresponds to a smaller drop in accuracy. Discussion We developed a CNN classifier that estimates daytime sleepiness based on polysomnographic (EEG, EOG, and chin EMG) signals recorded the night before MSLT. We found that the classifier was able to estimate sleepiness with moderate accuracy. The classifier classified patients to all sleepiness categories relatively evenly with no apparent bias for any sleepiness category (Figure 3). In detecting EDS, the sensitivity (76.5%) and specificity (77.9%) were good with reasonably high positive predictive value (78.0%) and negative predictive value (76.5%). In comparison, similar sensitivities (70% and 80%) and negative predictive values (75% and 76%) have been reported with ESS using cohort-optimized cutoff values (16 and 12 points) [31, 32]. However, the specificities (55% and 69%) were considerably lower compared to our classifier along with lower positive predictive values (61% and 74%) [31, 32]. High specificity (76%) has also been reported using ESS, but with low sensitivity (64%) [33]. Simultaneous high sensitivity and specificity has been difficult to achieve with ESS even when using cohort-optimized cutoff values [31–33]. Based on the present results, the CNN classifier developed in this study seems to be able to estimate sleepiness slightly better than ESS [31–33]. However, it is important to note that ESS is better suited as a measure of chronic, long-time sleepiness rather than the acute sleepiness. Therefore, ESS is still a valuable tool in sleepiness estimation. The classifier could be used as a simple estimator of the patients’ sleepiness since it is easy to implement and does not require long and laborious full-day test (i.e. MSLT) while still providing an objective estimate of the patient’s acute daytime sleepiness. In addition, ESS could be used in conjunction with the classifier to provide information on the chronic situation. In the subgroup analysis, older patients, patients with severe OSA or patients with high BMI were classified slightly more accurately than younger patients, patients with lower severity of OSA or patients with low BMI (Table 2). Since OSA severity generally increases with age and obesity [34], it could be that the sleepiness of these patients is mainly caused by the sleep apnea, which might be more clearly detectable from the spectrograms. Table 2. Classification accuracy in subgroups across all folds when classifying patients to the four sleepiness categories Subgroup . Number of patients in subgroup . Classification accuracy (%) . Males 1,492 61.5 Females 522 57.9 AHI < 5 401 52.1 5 ≤ AHI < 15 438 59.4 15 ≤ AHI < 30 422 57.5 AHI ≥ 30 753 67.5 BMI < 25 355 56.7 25 ≤ BMI < 30 660 60.0 30 ≤ BMI < 35 598 62.7 BMI ≥ 35 421 61.5 age < 40 430 54.2 40 ≤ age < 50 406 60.1 50 ≤ age < 60 668 62.7 age ≥ 60 510 63.5 Subgroup . Number of patients in subgroup . Classification accuracy (%) . Males 1,492 61.5 Females 522 57.9 AHI < 5 401 52.1 5 ≤ AHI < 15 438 59.4 15 ≤ AHI < 30 422 57.5 AHI ≥ 30 753 67.5 BMI < 25 355 56.7 25 ≤ BMI < 30 660 60.0 30 ≤ BMI < 35 598 62.7 BMI ≥ 35 421 61.5 age < 40 430 54.2 40 ≤ age < 50 406 60.1 50 ≤ age < 60 668 62.7 age ≥ 60 510 63.5 AHI, apnea–hypopnea index; BMI, body mass index. Open in new tab Table 2. Classification accuracy in subgroups across all folds when classifying patients to the four sleepiness categories Subgroup . Number of patients in subgroup . Classification accuracy (%) . Males 1,492 61.5 Females 522 57.9 AHI < 5 401 52.1 5 ≤ AHI < 15 438 59.4 15 ≤ AHI < 30 422 57.5 AHI ≥ 30 753 67.5 BMI < 25 355 56.7 25 ≤ BMI < 30 660 60.0 30 ≤ BMI < 35 598 62.7 BMI ≥ 35 421 61.5 age < 40 430 54.2 40 ≤ age < 50 406 60.1 50 ≤ age < 60 668 62.7 age ≥ 60 510 63.5 Subgroup . Number of patients in subgroup . Classification accuracy (%) . Males 1,492 61.5 Females 522 57.9 AHI < 5 401 52.1 5 ≤ AHI < 15 438 59.4 15 ≤ AHI < 30 422 57.5 AHI ≥ 30 753 67.5 BMI < 25 355 56.7 25 ≤ BMI < 30 660 60.0 30 ≤ BMI < 35 598 62.7 BMI ≥ 35 421 61.5 age < 40 430 54.2 40 ≤ age < 50 406 60.1 50 ≤ age < 60 668 62.7 age ≥ 60 510 63.5 AHI, apnea–hypopnea index; BMI, body mass index. Open in new tab The occlusion test showed that there does not seem to be a well-defined, specific region in the spectrogram image that is most important for sleepiness classification (Figure 5, A). However, occluding the frequencies less than 15Hz seemed to have a more detrimental effect on the accuracy of the classifier (Figure 5, B). This makes sense since most of the power in the spectrogram is at the lower frequencies. In addition, delta waves, associated with deep sleep are at this low frequency range (0.5–4Hz) [28]. As the amount of slow wave sleep is important in recovery during the night and greatly affects sleepiness, it could be that the amount of slow wave sleep detected from the spectrogram is a major component of the classifier function. However, since the accuracy suffered at least slightly when any part of the image was occluded, it seems that the whole night and all frequencies are at least somewhat important for the classifier. This study has certain limitations. The use of Pz electrode might limit generalizability of the classifier since it is not a commonly used electrode placement in PSG montage. However omitting this electrode would have lowered classifier accuracy (see supplement). Although the accuracy of the classifier was moderate, it still leaves room for improvement. One complicating factor in estimating the MSLT result is that the patient’s sleepiness is not entirely dependent on the previous night. Some of the patients might have been sleep deprived for a long time while others might only be sleepy because of poor sleep during the previous night polysomnography. While they both might be classified to the severe sleepiness category, their EEG and EMG spectrograms are likely significantly different. That is, all information on the patient’s sleepiness is not available in the single night polysomnographic recording, which limits the performance of the classifier. Another limiting factor is the patient population. Although the patient population was relatively large, using even larger population would have likely improved the results. Larger population would also have allowed a larger test set and thus enabled a more robust validation of the developed classifier. Another limiting factor of the patient population is that the baseline ESS test was not conducted and therefore we could not compare classifier accuracy to ESS or assess the subjective and objective sleepiness in the same population. In addition, no information on medications or comorbidities was available for this study population. This can be considered a study limitation as both of these could have an effect on the patients’ sleepiness. It is also important to note that all of the patients in this study were suspected OSA patients complaining from daytime sleepiness during clinical interview. This results in a biased population only consisting of patients with subjective sleepiness while including no patients who were objectively sleepy but not subjectively sleepy. Thus, the network might behave differently if applied to a different population such as a healthy population or to a population with different sleep disorders. In conclusion, objective estimation of daytime sleepiness using polysomnographic signals shows promising results. The developed CNN classifier could be applied for OSA patients that undergo polysomnography to get an objective EDS evaluation with minimal workload. Disclosure Statement Financial disclosure: Nikkonen reports funding from Academy of Finland (project number 313697), Research committee of the Kuopio University Hospital Catchment Area (project number 5041781), Instrumentarium Science Foundation, Research Foundation for Pulmonary Diseases, and Orion research foundation. Korkalainen reports funding from Academy of Finland (project number 313697), The Research Committee of the Kuopio University Hospital Catchment Area for the State Research Funding (project number 5041780), Respiratory Foundation of Kuopio Region, Päivikki and Sakari Sohlberg Foundation, Research Foundation of the Pulmonary Diseases, and Foundation of the Finnish Anti-Tuberculosis Association. Kainulainen reports funding from Academy of Finland (project number 313697), The Research Committee of the Kuopio University Hospital Catchment Area for the State Research Funding (project numbers 5041779 and 5041768), Päivikki and Sakari Sohlberg Foundation, and The Research Foundation of the Pulmonary Diseases. Myllymaa reports funding from Academy of Finland (project number 313697), Business Finland, Research committee of the Kuopio University Hospital Catchment Area (project number 5041770), Paulo Foundation, Tampere Tuberculosis Foundation, Leino reports funding from Päivikki and Sakari Sohlberg Foundation, Finnish Cultural Foundation, Respiratory Foundation of the Kuopio Region, The Research Foundation of the Pulmonary Diseases, and The Research Committee of the Kuopio University Hospital Catchment Area for the State Research Funding. Kalevo reports funding from The Research Foundation of the Pulmonary Diseases. Oksenberg has no funding to disclose. Leppänen reports funding from Academy of Finland (decision number 323536), The Research Committee of the Kuopio University Hospital Catchment Area for the State Research Funding (project number 5041767), Tampere Tuberculosis Foundation, and Respiratory Foundation of Kuopio Region. Töyräs reports funding from Academy of Finland (project number 313697), Kuopio University Hospital (project number 5041768), and Business Finland. Non-financial disclosure: Myllymaa has a patent US 9,770,184 B2 with royalties paid and a patent application PCT/FI2015/050676 . Toyras has a patent US 9,770,184 B2 with royalties paid and a patent application PCT/FI2015/050676. References 1. Heinzer R , et al. Prevalence of sleep-disordered breathing in the general population: the HypnoLaus study . Lancet Respir Med. 2015 ; 3 : 310 – 318 . Google Scholar Crossref Search ADS PubMed WorldCat 2. Benjafield AV , et al. Estimation of the global prevalence and burden of obstructive sleep apnoea: a literature-based analysis . Lancet Respir Med. 2019 ; 7 : 687 – 698 . Google Scholar Crossref Search ADS PubMed WorldCat 3. Regestein Q , et al. Sleep debt and depression in female college students . Psychiatry Res. 2010 ; 176 : 34 – 39 . Google Scholar Crossref Search ADS PubMed WorldCat 4. Ohayon MM , et al. Daytime sleepiness and cognitive impairment in the elderly population . Arch Intern Med. 2002 ; 162 : 201 – 208 . Google Scholar Crossref Search ADS PubMed WorldCat 5. Slater G , et al. Excessive daytime sleepiness in sleep disorders . J Thorac Dis. 2012 ; 4 : 608 – 616 . Google Scholar PubMed OpenURL Placeholder Text WorldCat 6. Knauert M , et al. Clinical consequences and economic costs of untreated obstructive sleep apnea syndrome . World J Otorhinolaryngol Head Neck Surg. 2015 ; 1 : 17 – 27 . Google Scholar Crossref Search ADS PubMed WorldCat 7. Carskadon MA , et al. Guidelines for the multiple sleep latency test (MSLT): a standard measure of sleepiness . Sleep. 1986 ; 9 : 519 – 524 . Google Scholar Crossref Search ADS PubMed WorldCat 8. American Academy of Sleep Medicine. The International Classification of Sleep Disorders, Revised Daignostic and Coding Manual. Chicago, IL : American Academy of Sleep Medicine , 2001 . doi:10.1145/3274694.3274742 Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC 9. Thorpy MJ . The clinical use of the multiple sleep latency test. The standards of practice committee of the American sleep disorders association . Sleep. 1992 ; 15 : 268 – 276 . Google Scholar Crossref Search ADS PubMed WorldCat 10. Sateia MJ . International classification of sleep disorders-third edition: highlights and modifications . Chest. 2014 ; 146 : 1387 – 1394 . Google Scholar Crossref Search ADS PubMed WorldCat 11. Arand D , et al. The clinical use of the MSLT and MWT . Sleep. 2005 ; 28 : 123 – 144 . Google Scholar Crossref Search ADS PubMed WorldCat 12. Johns MW . Daytime sleepiness, snoring, and obstructive sleep apnea. The Epworth Sleepiness Scale . Chest. 1993 ; 103 : 30 – 36 . Google Scholar Crossref Search ADS PubMed WorldCat 13. Buysse DJ , et al. The Pittsburgh sleep quality index: a new instrument for psychiatric practice and research . Psychiatry Res. 1988 ; 28 : 193 – 213 . doi:10.1016/0165-1781(89)90047-4 Google Scholar Crossref Search ADS WorldCat 14. Fong SYY , et al. Comparing MSLT ansd ESS in the measurement of excessive daytime sleepiness in obstructive sleep apnoea syndrome . J Psychosom Res. 2005 ; 58 : 55 – 60 . Google Scholar Crossref Search ADS PubMed WorldCat 15. Park YK , et al. Clinical and polysomnographic characteristics of patients with excessive daytime sleepiness . Sleep Med Res. 2018 ; 9 : 32 – 38 . Google Scholar Crossref Search ADS WorldCat 16. Olson LG , et al. Correlations among Epworth Sleepiness Scale scores, multiple sleep latency tests and psychological symptoms . J Sleep Res. 1998 ; 7 : 248 – 253 . Google Scholar Crossref Search ADS PubMed WorldCat 17. Miotto R , et al. Deep learning for healthcare: review, opportunities and challenges . Brief Bioinform. 2017 ; 19 : 1236 – 1246 . Google Scholar Crossref Search ADS WorldCat 18. Uddin MB , et al. Classification methods to detect sleep apnea in adults based on respiratory and oximetry signals: a systematic review . Physiol Meas. 2018 ; 39 : 03TR01 . Google Scholar Crossref Search ADS PubMed WorldCat 19. Nikkonen S , et al. Artificial neural network analysis of the oxygen saturation signal enables accurate diagnostics of sleep apnea . Sci Rep. 2019 ; 9 : 13200 . Google Scholar Crossref Search ADS PubMed WorldCat 20. Boostani R , et al. A comparative review on sleep stage classification methods in patients and healthy individuals . Comput Methods Programs Biomed. 2017 ; 140 : 77 – 91 . Google Scholar Crossref Search ADS PubMed WorldCat 21. Sun H , et al. Large-scale automated sleep staging . Sleep. 2017 ; 40 . doi:10.1093/sleep/zsx139 Google Scholar OpenURL Placeholder Text WorldCat 22. Zhang W , et al. Parallel distributed processing model with local space-invariant interconnections and its optical architecture . Appl Opt. 1990 ; 29 : 4790 – 4797 . Google Scholar Crossref Search ADS PubMed WorldCat 23. American Academy of Sleep Medicine. Sleep-related breathing disorders in adults: recommendations for syndrome definition and measurement techniques in clinical research . Sleep. 1999 ; 22 : 662 – 689 . Crossref Search ADS PubMed WorldCat 24. American Academy of Sleep Medicine. The AASM Manual for the Scoring of Sleep and Associated Events: Rules, Terminology and Technical Specifications. Vol 1 . Westchester, IL : American Academy of Sleep Medicine , 2007 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC 25. Littner MR , et al. Practice parameters for clinical use of the multiple sleep latency test and the maintenance of wakefulness test . Sleep. 2005 ; 28 : 113 – 121 . Google Scholar Crossref Search ADS PubMed WorldCat 26. Jasper HH . The ten-twenty electrode system of the International Federation . Electroencephalogr Clin Neurophysiol. 1958 ; 10 : 371 – 375 . Google Scholar OpenURL Placeholder Text WorldCat 27. Welch P . The use of fast Fourier transform for the estimation of power spectra: a method based on time averaging over short, modified periodograms . IEEE Trans audio Electroacoust. 1967 ; 15 : 70 – 73 . Google Scholar Crossref Search ADS WorldCat 28. Bronzino J. Medical Devices and Systems, The Biomedical Engineering Handbook, 3rd ed. Boca Raton, FL : CRC Press , 2006 . Google Scholar Crossref Search ADS Google Preview WorldCat COPAC 29. American Academy of Sleep Medicine. AASM Manual for the Scoring of Sleep and Associated Events . Darien, IL : American Academy of Sleep Medicine , 2017 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC 30. Cohen J . A coefficient of agreement for nominal scales . Educ Psychol Meas. 1960 ; 20 : 37 – 46 . Google Scholar Crossref Search ADS WorldCat 31. Trimmel K , et al. Wanted: a better cut-off value for the Epworth Sleepiness Scale . Wien Klin Wochenschr. 2018 ; 130 : 349 – 355 . Google Scholar Crossref Search ADS PubMed WorldCat 32. Cai SJ , et al. Correlation of Epworth Sleepiness Scale with multiple sleep latency test and its diagnostic accuracy in assessing excessive daytime sleepiness in patients with obstructive sleep apnea hypopnea syndrome . Chin Med J (Engl). 2013 ; 126 : 3245 – 3250 . Google Scholar PubMed OpenURL Placeholder Text WorldCat 33. Chung KF . Use of the Epworth Sleepiness Scale in Chinese patients with obstructive sleep apnea and normal hospital employees . J Psychosom Res. 2000 ; 49 : 367 – 372 . Google Scholar Crossref Search ADS PubMed WorldCat 34. Jordan AS , et al. Adult obstructive sleep apnoea . Lancet. 2014 ; 383 : 736 – 747 . Google Scholar Crossref Search ADS PubMed WorldCat © Sleep Research Society 2020. Published by Oxford University Press on behalf of the Sleep Research Society. This is an Open Access article distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs licence (http://creativecommons.org/licenses/by-nc-nd/4.0/), which permits non-commercial reproduction and distribution of the work, in any medium, provided the original work is not altered or transformed in any way, and that the work is properly cited. For commercial re-use, please contact [email protected] © Sleep Research Society 2020. Published by Oxford University Press on behalf of the Sleep Research Society.

Wickwire, Emerson, M;Jobe, Sophia, L;Oldstone, Liesl, M;Scharf, Steven, M;Johnson, Abree, M;Albrecht, Jennifer, S

doi: 10.1093/sleep/zsaa122pmid: 32575113

Abstract Study Objectives To examine rates of adherence to continuous positive airway pressure (CPAP) therapy among a representative sample of older adult Medicare beneficiaries with obstructive sleep apnea (OSA), and to identify demographic and health-related factors associated with CPAP adherence. Methods Using a 5% sample of Medicare claims data, we utilized Medicare policy and CPAP machine charges as a proxy for CPAP adherence. A cumulative logit model was used to identify demographic, medical, and psychiatric predictors of CPAP adherence status. Results Of beneficiaries who initiated CPAP (n = 3,229), 74.9% (n = 2,417) met the so-called “90-day Medicare adherence criteria,” but only 58.8% of these individuals (n = 1,420) continued to use CPAP throughout the entire 13-month rent-to-own period. Anxiety, anemia, fibromyalgia, traumatic brain injury, and lower socioeconomic status (SES) were all associated with reduced CPAP adherence. Conclusions These results provide the first national estimates of CPAP adherence among older adult Medicare beneficiaries in the United States. In addition, findings highlight the salience of medical and psychiatric comorbidity, as well as SES, as important markers of CPAP adherence among older adults in the United States. Future studies should seek to evaluate interventions to improve CPAP adherence among older adults of lower SES. obstructive sleep apnea, continuous positive airway pressure, CPAP, treatment, adherence, Medicare, older adults Statement of Significance This is the first study to present national adherence rates for CPAP therapy among older adults in the United States. Demographic, medical, and psychiatric predictors of CPAP adherence were identified. Lower socioeconomic status was associated with reduced CPAP adherence, an important health disparities finding that warrants future research attention. Obstructive sleep apnea (OSA) is a common and costly chronic medical condition with numerous adverse health consequences. Relative to individuals without OSA, individuals with OSA are at increased risk for cardiovascular disease (CVD) [1–3], stroke [4, 5], metabolic syndromes and Type 2 diabetes [6–8], depression [9], reduced quality of life [10], and premature death [11, 12]. Further, relative to individuals without OSA, individuals with OSA incur dramatically increased health care utilization and costs borne by patients, payers, and society [13–15]. Most studies to date have evaluated OSA among middle-aged men and women. However, the prevalence of OSA increases with age and is highest among older adults [16]. The prevalence of moderate to severe OSA is approximately 14% among men and 5% among women aged 30–70 years [17], and individuals over 65 years of age at higher risk for developing OSA than any other age group [16]. Further, up to 70% of individuals in elder care facilities meet diagnostic criteria for OSA, which subsequently increases risk for dying during sleep among this population [18]. Continuous positive airway pressure (CPAP) is the most commonly prescribed treatment for OSA and is often associated with a broad range of health benefits, including reduced risk for cerebrovascular events, CVDs, diabetes, motor vehicle crashes, and overall mortality, as well as improvements in cognitive function, psychiatric symptoms, and quality of life [19]. Among older adults specifically, CPAP has been associated with improved cardiovascular outcomes [20], higher survival rates [21, 22], improved cognition [23], and a decreased risk for stroke and incident hypertension [24]. In terms of subjective outcomes, CPAP has also been associated with reductions in daytime fatigue, excessive daytime sleepiness, and depressive symptoms, as well as increased quality of life among older adults [25]. Despite these benefits, many patients struggle to adhere to the treatment, with published rates of adherence ranging from 20% to 80% [19, 26]. Somewhat surprisingly, relatively few studies have evaluated rates of CPAP usage among older adults. Twenty years ago, Parish, Lyng, and Wisbey (2000) conducted a small, prospective cohort study among older adults (n = 71) and found self-reported CPAP adherence (mean usage 6.5 hours per night on 6.5 nights per week) to be 70% after 3 months of usage [27]. The following year, Russo-Magno and colleagues conducted a retrospective chart review among older adults (N = 33) and found objective CPAP adherence (defined as average usage ≥5 hours per night) to be 60.6% over 6 months [28]. Also in 2001, Pelletier-Fleury and colleagues conducted a prospective cohort study among older adults (n = 57) and found objective CPAP adherence of at least 3 hours per day to be 80.7% at 3 months, 77.19% over 6 months, 71.93% over 1 year, and 55.92% over 3 years [29]. Within the past decade, Woehrle, Graml, and Weinreich (2011) conducted a retrospective cohort study among adults >70 years old (n = 575) and found objective CPAP adherence of >3 hours per night in 96.5% of their participants, >4 h in 92.9%, and >5 h in 82.4%, with a mean observation period of 156 ± 25 days [30]. Conversely, McMillan and colleagues [31] analyzed objective positive airway pressure (PAP) usage among older adults and found mean usage of >4 per night in 35% of participants at 3-months (n = 41/117) and 12-months (n = 36/102). Nonetheless, in aggregate these data suggest adequate CPAP usage among older adults, although most of these studies are limited by small sample sizes and varied operational definitions of CPAP usage. In light of the substantial disease burden of OSA among older adults as well as the importance of CPAP adherence in improving OSA clinical and public health outcomes, the purpose of the present study was to evaluate rates of CPAP adherence among a nationally representative sample of older adult Medicare beneficiaries newly diagnosed with OSA. In addition, we sought to identify predictors of poorer CPAP usage. Understanding predictors of CPAP usage will enable identification of individuals at risk for non-adherence and development of targeted treatment strategies to maximize CPAP adherence. Based on past literature [19, 26], we hypothesized that lower socioeconomic status (SES) and greater medical and psychiatric comorbidity would be associated with reduced CPAP adherence among older adults. Methods Data source The primary source of data for this study was a 5% random sample of Medicare administrative data for years 2008–2013 obtained from the Centers for Medicare & Medicaid Services (CMS) Chronic Condition Data Warehouse (CCW). Study population Because the purpose of this study was to examine the impact of CPAP adherence among older adults, we restricted our analysis to Medicare beneficiaries age 65 and older newly diagnosed with OSA with at least one CPAP machine charge. To ensure comprehensive data capture, we required all study participants to have continuous Medicare Parts A, B, and D, with no Medicare Part C (Medicare Advantage) coverage for the 12 months before and the 24 months after the first CPAP charge. OSA cohort We defined OSA as presence of at least one physician-assigned OSA diagnosis (i.e. inpatient or outpatient claim containing International Classification of Disease, Version 9, Clinical Modification [ICD-9-CM] codes 327.23, 780.51, 780.53, and 780.57) during the years 2009–2011. We searched durable medical equipment claims for evidence of at least one CPAP machine charge (i.e. Healthcare Common Procedure Coding System code E0601, which includes CPAP and auto-adjusting CPAP). The index date was the first date of a CPAP machine charge following a 12-month “clean” period in which no CPAP charges were observed. Other CPAP modalities were excluded. PAP adherence Beginning November 1, 2008, the Center for Medicare and Medicaid Services (CMS) implemented objective adherence criteria required for reimbursement of CPAP (i.e. >4 h of CPAP on 70% of nights or 21 days in a consecutive 30-day period within the first 90 days of CPAP initiation). Thus, because CPAP machines are billed in 13 monthly payments during a “rent-to-own” period, we utilized number of CPAP machine charges as a proxy for CPAP adherence. Beneficiaries were considered non-adherent (3 or fewer charges, demonstrating that they did not meet the 90-day Medicare criteria), partially adherent (4–12 machine charges, demonstrating that they met the 90-day criteria but subsequently returned the machine), or highly adherent (13 machine charges, demonstrating that they met the 90-day criteria and continued possession of the machine throughout the rent-to-own period). Covariates Information on beneficiary demographic characteristics was obtained from the claims files. The CCW contains information on 27 comorbid conditions, with an annual flag for each condition as well as the date of first diagnosis [32]. We used the date of first diagnosis to determine if a condition was present at the date of the first CPAP charge (i.e. index date). Other comorbidities of interest (i.e. fibromyalgia, traumatic brain injury [TBI], and the psychiatric comorbidities) were identified by searching all claim types for relevant ICD-9-CM codes during the study period. For these variables only, any diagnoses received during the year prior to the first CPAP charge were assumed to be present at the index date. A comorbidity index based on the Deyo adaptation of the Charlson comorbidity index was calculated and included in subsequent analyses [33]. We identified point of service codes for nursing homes on the claims and created a variable indicating at least one nursing home stay during the follow-up period. We also created a variable indicating that the beneficiary was eligible for Medicaid (an indicator of lower SES) at any time during the follow-up period. Original reason for entitlement is reported as age versus other (includes disability and end-stage renal disease). Data analysis Differences in the distribution and frequency of all variables across the three CPAP adherence categories were assessed using chi-square goodness of fit and one-way analysis of variance (ANOVA). To identify independent predictors of poorer CPAP adherence, a cumulative logit model was employed. First, we entered variables associated with adherence status at p < 0.1 into the model. We subsequently retained variables significant at p < 0.1 and created our final model by retaining only those variables whose p value was <0.05. Odds ratios (OR) and 95% confidence intervals (CI) are reported. All analyses were performed with SAS version 9.4 (SAS Institute, Cary, NC). This study was approved by the Institutional Review Board of the University of Maryland, Baltimore. Results We identified 29,072 older adult beneficiaries diagnosed with OSA between 2009 and 2011 and meeting continuous coverage criteria. Of these, 7,111 (24.5%) initiated CPAP and 3,229 (45.4% of CPAP initiators) were aged 65 and older. As presented in Table 1, 44% (n = 1,420) of beneficiaries who initiated CPAP subsequently met the 90-day adherence criteria and maintained possession of their CPAP machine throughout the 13-month rent-to-own period, as evidenced by the number of CPAP machine charges; these individuals were labeled “high adherers.” An additional 997 (30.9%) beneficiaries received between 4 and 12 CPAP machine charges and were labeled “partial adherers.” Finally, 812 beneficiaries (25.2%) received fewer than 4 CPAP machine charges and were labeled “non-adherers.” Relative to non-adherers, high adherers were older (mean age 79.2, standard deviation [SD] 5.5 years among high adherers vs. mean age 72.5, SD 5.8 years among non-adherers, p < 0.001). Relative to non-adherers, high adherers also demonstrated a lower comorbidity burden as evidenced by a lower frequency of individuals in the highest comorbidity score category (30.4% among high adherers vs. 35.2% among non-adherers, p = 0.002). Comorbid insomnia was highly prevalent among older adults with OSA but did not differ significantly across CPAP adherence categories. There was a clear trend toward association between Medicaid eligibility and CPAP adherence, such that Medicaid eligibility was highest among non-adherers, lower among partial adherers, and lowest among high adherers (21.9% vs. 17.2% vs. 13.6%, respectively, p < 0.001). Table 1. Characteristics of Medicare beneficiaries ≥65 years diagnosed with OSA 2009–2011 and who initiated PAP treatment, by PAP adherence category, n = 3,229 . Total sample, n = 3,229 . <4 PAP charges, n = 812 . 4–12 PAP charges, n = 997 . >12 PAP charges, n = 1,420 . P-value* . Age, mean (SD) 72.5 (5.8) 72.1 (5.3) 79.2 (5.5) <0.001 Sex, n (%) 0.06  Male 1,689 (52.3) 396 (48.8) 532 (53.3) 762 (53.7)  Female 1,540 (47.7) 416 (51.2) 466 (46.7) 658 (46.3) Race, n (%) 0.008  White 2,846 (88.1) 690 (85.0) 870 (87.3) 1,286 (90.6)  Black 238 (7.4) 78 (9.6) 77 (7.7) 83 (5.8)  Hispanic 56 (1.7) 19 (2.3) 19 (1.9) 31 (1.3)  Other 89 (2.8) 25 (3.1) 31 (3.1) 33 (2.3) Comorbid illness, n (%)  Alzheimer’s and related dementias 236 (7.3) 79 (9.7) 72 (7.2) 85 (6.0) 0.005  Anemia 1,510 (46.8) 430 (53.0) 447 (44.8) 633 (44.6) <0.001  Anxiety 462 (14.3) 137 (16.9) 162 (16.3) 163 (11.5) <0.001  Asthma 580 (18.0) 166 (20.4) 160 (16.1) 254 (17.9) 0.05  Chronic kidney disease 693 (21.5) 194 (23.9) 214 (21.5) 285 (20.1) 0.11  Chronic obstructive pulmonary disease 972 (30.1) 269 (33.1) 287 (28.8) 416 (29.3) 0.09  Depression 645 (20.0) 191 (23.5) 199 (20.0) 255 (18.0) 0.007  Diabetes 1,417 (43.9) 374 (46.1) 450 (45.1) 593 (41.8) 0.09  Fibromyalgia 817 (25.3) 236 (29.1) 255 (25.6) 326 (23.0) 0.006  Heart Failure 1,042 (32.3) 297 (36.6) 297 (29.8) 448 (31.6) 0.007  Hyperlipidemia 2,770 (85.8) 690 (85.0) 863 (86.6) 1,217 (85.7) 0.63  Hypertension 2,875 (89.0) 717 (88.3) 885 (88.8) 1,273 (89.7) 0.59  Hypothyroidism 761 (23.6) 222 (27.3) 223 (22.4) 316 (22.3) 0.01  Insomnia 2,887 (89.4) 720 (88.7) 903 (90.6) 1,264 (89.0) 0.35  Ischemic Heart Disease 1,884 (58.4) 494 (60.8) 568 (57.0) 822 (57.9) 0.23  Rheumatoid Arthritis 1,785 (55.3) 481 (59.2) 531 (53.3) 773 (54.4) 0.03  Stroke 454 (14.1) 125 (15.4) 138 (13.8) 191 (13.5) 0.43  Substance abuse 350 (10.8) 95 (11.7) 110 (11.0) 145 (10.2) 0.54  Traumatic brain injury 68 (8.4) 67 (6.7) 62 (4.4) <0.001 Comorbidity Index, n (%) 0.002  0 486 (15.1) 97 (12.0) 147 (14.7) 242 (17.0)  1 925 (28.7) 225 (27.7) 303 (30.4) 397 (28.0)  2 825 (25.6) 204 (25.1) 272 (27.3) 349 (24.6)  3 993 (30.8) 286 (35.2) 275 (27.6) 432 (30.4) At least one nursing home stay, n (%) 146 (4.5) 41 (5.1) 46 (4.6) 59 (4.2) 0.61 Medicaid eligibility, n (%) 542 (16.8) 178 (21.9) 171 (17.2) 193 (13.6) <0.001 Original reason for Medicare entitlement, n (%) 0.18 Age eligibility 2,872 (88.9) 708 (87.2) 891 (89.4) 1,273 (88.7) Other 357 (11.1) 104 (12.8) 106 (10.6) 147 (10.3) . Total sample, n = 3,229 . <4 PAP charges, n = 812 . 4–12 PAP charges, n = 997 . >12 PAP charges, n = 1,420 . P-value* . Age, mean (SD) 72.5 (5.8) 72.1 (5.3) 79.2 (5.5) <0.001 Sex, n (%) 0.06  Male 1,689 (52.3) 396 (48.8) 532 (53.3) 762 (53.7)  Female 1,540 (47.7) 416 (51.2) 466 (46.7) 658 (46.3) Race, n (%) 0.008  White 2,846 (88.1) 690 (85.0) 870 (87.3) 1,286 (90.6)  Black 238 (7.4) 78 (9.6) 77 (7.7) 83 (5.8)  Hispanic 56 (1.7) 19 (2.3) 19 (1.9) 31 (1.3)  Other 89 (2.8) 25 (3.1) 31 (3.1) 33 (2.3) Comorbid illness, n (%)  Alzheimer’s and related dementias 236 (7.3) 79 (9.7) 72 (7.2) 85 (6.0) 0.005  Anemia 1,510 (46.8) 430 (53.0) 447 (44.8) 633 (44.6) <0.001  Anxiety 462 (14.3) 137 (16.9) 162 (16.3) 163 (11.5) <0.001  Asthma 580 (18.0) 166 (20.4) 160 (16.1) 254 (17.9) 0.05  Chronic kidney disease 693 (21.5) 194 (23.9) 214 (21.5) 285 (20.1) 0.11  Chronic obstructive pulmonary disease 972 (30.1) 269 (33.1) 287 (28.8) 416 (29.3) 0.09  Depression 645 (20.0) 191 (23.5) 199 (20.0) 255 (18.0) 0.007  Diabetes 1,417 (43.9) 374 (46.1) 450 (45.1) 593 (41.8) 0.09  Fibromyalgia 817 (25.3) 236 (29.1) 255 (25.6) 326 (23.0) 0.006  Heart Failure 1,042 (32.3) 297 (36.6) 297 (29.8) 448 (31.6) 0.007  Hyperlipidemia 2,770 (85.8) 690 (85.0) 863 (86.6) 1,217 (85.7) 0.63  Hypertension 2,875 (89.0) 717 (88.3) 885 (88.8) 1,273 (89.7) 0.59  Hypothyroidism 761 (23.6) 222 (27.3) 223 (22.4) 316 (22.3) 0.01  Insomnia 2,887 (89.4) 720 (88.7) 903 (90.6) 1,264 (89.0) 0.35  Ischemic Heart Disease 1,884 (58.4) 494 (60.8) 568 (57.0) 822 (57.9) 0.23  Rheumatoid Arthritis 1,785 (55.3) 481 (59.2) 531 (53.3) 773 (54.4) 0.03  Stroke 454 (14.1) 125 (15.4) 138 (13.8) 191 (13.5) 0.43  Substance abuse 350 (10.8) 95 (11.7) 110 (11.0) 145 (10.2) 0.54  Traumatic brain injury 68 (8.4) 67 (6.7) 62 (4.4) <0.001 Comorbidity Index, n (%) 0.002  0 486 (15.1) 97 (12.0) 147 (14.7) 242 (17.0)  1 925 (28.7) 225 (27.7) 303 (30.4) 397 (28.0)  2 825 (25.6) 204 (25.1) 272 (27.3) 349 (24.6)  3 993 (30.8) 286 (35.2) 275 (27.6) 432 (30.4) At least one nursing home stay, n (%) 146 (4.5) 41 (5.1) 46 (4.6) 59 (4.2) 0.61 Medicaid eligibility, n (%) 542 (16.8) 178 (21.9) 171 (17.2) 193 (13.6) <0.001 Original reason for Medicare entitlement, n (%) 0.18 Age eligibility 2,872 (88.9) 708 (87.2) 891 (89.4) 1,273 (88.7) Other 357 (11.1) 104 (12.8) 106 (10.6) 147 (10.3) *p Value from ANOVA or chi-square goodness of fit. Open in new tab Table 1. Characteristics of Medicare beneficiaries ≥65 years diagnosed with OSA 2009–2011 and who initiated PAP treatment, by PAP adherence category, n = 3,229 . Total sample, n = 3,229 . <4 PAP charges, n = 812 . 4–12 PAP charges, n = 997 . >12 PAP charges, n = 1,420 . P-value* . Age, mean (SD) 72.5 (5.8) 72.1 (5.3) 79.2 (5.5) <0.001 Sex, n (%) 0.06  Male 1,689 (52.3) 396 (48.8) 532 (53.3) 762 (53.7)  Female 1,540 (47.7) 416 (51.2) 466 (46.7) 658 (46.3) Race, n (%) 0.008  White 2,846 (88.1) 690 (85.0) 870 (87.3) 1,286 (90.6)  Black 238 (7.4) 78 (9.6) 77 (7.7) 83 (5.8)  Hispanic 56 (1.7) 19 (2.3) 19 (1.9) 31 (1.3)  Other 89 (2.8) 25 (3.1) 31 (3.1) 33 (2.3) Comorbid illness, n (%)  Alzheimer’s and related dementias 236 (7.3) 79 (9.7) 72 (7.2) 85 (6.0) 0.005  Anemia 1,510 (46.8) 430 (53.0) 447 (44.8) 633 (44.6) <0.001  Anxiety 462 (14.3) 137 (16.9) 162 (16.3) 163 (11.5) <0.001  Asthma 580 (18.0) 166 (20.4) 160 (16.1) 254 (17.9) 0.05  Chronic kidney disease 693 (21.5) 194 (23.9) 214 (21.5) 285 (20.1) 0.11  Chronic obstructive pulmonary disease 972 (30.1) 269 (33.1) 287 (28.8) 416 (29.3) 0.09  Depression 645 (20.0) 191 (23.5) 199 (20.0) 255 (18.0) 0.007  Diabetes 1,417 (43.9) 374 (46.1) 450 (45.1) 593 (41.8) 0.09  Fibromyalgia 817 (25.3) 236 (29.1) 255 (25.6) 326 (23.0) 0.006  Heart Failure 1,042 (32.3) 297 (36.6) 297 (29.8) 448 (31.6) 0.007  Hyperlipidemia 2,770 (85.8) 690 (85.0) 863 (86.6) 1,217 (85.7) 0.63  Hypertension 2,875 (89.0) 717 (88.3) 885 (88.8) 1,273 (89.7) 0.59  Hypothyroidism 761 (23.6) 222 (27.3) 223 (22.4) 316 (22.3) 0.01  Insomnia 2,887 (89.4) 720 (88.7) 903 (90.6) 1,264 (89.0) 0.35  Ischemic Heart Disease 1,884 (58.4) 494 (60.8) 568 (57.0) 822 (57.9) 0.23  Rheumatoid Arthritis 1,785 (55.3) 481 (59.2) 531 (53.3) 773 (54.4) 0.03  Stroke 454 (14.1) 125 (15.4) 138 (13.8) 191 (13.5) 0.43  Substance abuse 350 (10.8) 95 (11.7) 110 (11.0) 145 (10.2) 0.54  Traumatic brain injury 68 (8.4) 67 (6.7) 62 (4.4) <0.001 Comorbidity Index, n (%) 0.002  0 486 (15.1) 97 (12.0) 147 (14.7) 242 (17.0)  1 925 (28.7) 225 (27.7) 303 (30.4) 397 (28.0)  2 825 (25.6) 204 (25.1) 272 (27.3) 349 (24.6)  3 993 (30.8) 286 (35.2) 275 (27.6) 432 (30.4) At least one nursing home stay, n (%) 146 (4.5) 41 (5.1) 46 (4.6) 59 (4.2) 0.61 Medicaid eligibility, n (%) 542 (16.8) 178 (21.9) 171 (17.2) 193 (13.6) <0.001 Original reason for Medicare entitlement, n (%) 0.18 Age eligibility 2,872 (88.9) 708 (87.2) 891 (89.4) 1,273 (88.7) Other 357 (11.1) 104 (12.8) 106 (10.6) 147 (10.3) . Total sample, n = 3,229 . <4 PAP charges, n = 812 . 4–12 PAP charges, n = 997 . >12 PAP charges, n = 1,420 . P-value* . Age, mean (SD) 72.5 (5.8) 72.1 (5.3) 79.2 (5.5) <0.001 Sex, n (%) 0.06  Male 1,689 (52.3) 396 (48.8) 532 (53.3) 762 (53.7)  Female 1,540 (47.7) 416 (51.2) 466 (46.7) 658 (46.3) Race, n (%) 0.008  White 2,846 (88.1) 690 (85.0) 870 (87.3) 1,286 (90.6)  Black 238 (7.4) 78 (9.6) 77 (7.7) 83 (5.8)  Hispanic 56 (1.7) 19 (2.3) 19 (1.9) 31 (1.3)  Other 89 (2.8) 25 (3.1) 31 (3.1) 33 (2.3) Comorbid illness, n (%)  Alzheimer’s and related dementias 236 (7.3) 79 (9.7) 72 (7.2) 85 (6.0) 0.005  Anemia 1,510 (46.8) 430 (53.0) 447 (44.8) 633 (44.6) <0.001  Anxiety 462 (14.3) 137 (16.9) 162 (16.3) 163 (11.5) <0.001  Asthma 580 (18.0) 166 (20.4) 160 (16.1) 254 (17.9) 0.05  Chronic kidney disease 693 (21.5) 194 (23.9) 214 (21.5) 285 (20.1) 0.11  Chronic obstructive pulmonary disease 972 (30.1) 269 (33.1) 287 (28.8) 416 (29.3) 0.09  Depression 645 (20.0) 191 (23.5) 199 (20.0) 255 (18.0) 0.007  Diabetes 1,417 (43.9) 374 (46.1) 450 (45.1) 593 (41.8) 0.09  Fibromyalgia 817 (25.3) 236 (29.1) 255 (25.6) 326 (23.0) 0.006  Heart Failure 1,042 (32.3) 297 (36.6) 297 (29.8) 448 (31.6) 0.007  Hyperlipidemia 2,770 (85.8) 690 (85.0) 863 (86.6) 1,217 (85.7) 0.63  Hypertension 2,875 (89.0) 717 (88.3) 885 (88.8) 1,273 (89.7) 0.59  Hypothyroidism 761 (23.6) 222 (27.3) 223 (22.4) 316 (22.3) 0.01  Insomnia 2,887 (89.4) 720 (88.7) 903 (90.6) 1,264 (89.0) 0.35  Ischemic Heart Disease 1,884 (58.4) 494 (60.8) 568 (57.0) 822 (57.9) 0.23  Rheumatoid Arthritis 1,785 (55.3) 481 (59.2) 531 (53.3) 773 (54.4) 0.03  Stroke 454 (14.1) 125 (15.4) 138 (13.8) 191 (13.5) 0.43  Substance abuse 350 (10.8) 95 (11.7) 110 (11.0) 145 (10.2) 0.54  Traumatic brain injury 68 (8.4) 67 (6.7) 62 (4.4) <0.001 Comorbidity Index, n (%) 0.002  0 486 (15.1) 97 (12.0) 147 (14.7) 242 (17.0)  1 925 (28.7) 225 (27.7) 303 (30.4) 397 (28.0)  2 825 (25.6) 204 (25.1) 272 (27.3) 349 (24.6)  3 993 (30.8) 286 (35.2) 275 (27.6) 432 (30.4) At least one nursing home stay, n (%) 146 (4.5) 41 (5.1) 46 (4.6) 59 (4.2) 0.61 Medicaid eligibility, n (%) 542 (16.8) 178 (21.9) 171 (17.2) 193 (13.6) <0.001 Original reason for Medicare entitlement, n (%) 0.18 Age eligibility 2,872 (88.9) 708 (87.2) 891 (89.4) 1,273 (88.7) Other 357 (11.1) 104 (12.8) 106 (10.6) 147 (10.3) *p Value from ANOVA or chi-square goodness of fit. Open in new tab Results from our cumulative logit model are presented in Table 2. We identified six independent predictors of poorer CPAP adherence. The strongest predictors were TBI (OR 1.58; 95% CI 1.21, 2.07) and Medicaid eligibility (OR 1.49; 1.26, 1.78). Table 2. Independent predictors of poorer adherence to PAP treatment among Medicare beneficiaries ≥65 years diagnosed with OSA 2009–2011 and who initiated PAP, n = 3,229 . Odds ratio (95% CI) . Anxiety 1.34 (1.12, 1.61) Anemia 1.16 (1.02, 1.32) Fibromyalgia 1.19 (1.03, 1.38) Traumatic brain injury 1.58 (1.21, 2.07) Medicaid eligibility 1.48 (1.24, 1.75) . Odds ratio (95% CI) . Anxiety 1.34 (1.12, 1.61) Anemia 1.16 (1.02, 1.32) Fibromyalgia 1.19 (1.03, 1.38) Traumatic brain injury 1.58 (1.21, 2.07) Medicaid eligibility 1.48 (1.24, 1.75) Open in new tab Table 2. Independent predictors of poorer adherence to PAP treatment among Medicare beneficiaries ≥65 years diagnosed with OSA 2009–2011 and who initiated PAP, n = 3,229 . Odds ratio (95% CI) . Anxiety 1.34 (1.12, 1.61) Anemia 1.16 (1.02, 1.32) Fibromyalgia 1.19 (1.03, 1.38) Traumatic brain injury 1.58 (1.21, 2.07) Medicaid eligibility 1.48 (1.24, 1.75) . Odds ratio (95% CI) . Anxiety 1.34 (1.12, 1.61) Anemia 1.16 (1.02, 1.32) Fibromyalgia 1.19 (1.03, 1.38) Traumatic brain injury 1.58 (1.21, 2.07) Medicaid eligibility 1.48 (1.24, 1.75) Open in new tab Discussion The current study represents the largest national analysis of CPAP adherence among older adults to date. Using a nationally representative sample of Medicare administrative claims data, rates of CPAP adherence were found to be lower than most previous findings among older adults [27–30]. Using Medicare policy and CPAP machine charges as a proxy for CPAP usage, 74.9% (n = 2,417) of older adults diagnosed with OSA and who initiated CPAP therapy (n = 3,229) met “90-day Medicare adherence criteria,” but only 43.9% (n = 1,420) of CPAP initiators subsequently maintained possession of their CPAP machine throughout the entire 13-month rent-to-own period. In addition, our hypotheses regarding predictors of CPAP usage were generally supported. Our most salient finding regarding predictors is the inverse relation between SES and CPAP adherence, such that lower SES was associated with reduced CPAP adherence in our sample. OSA incurs substantial disease burden among older adults, and CPAP adherence is an important determinant of improving OSA clinical and public health outcomes among adults of all ages [34, 35]. Thus, it is interesting to consider present findings in the context of prior research. In a comprehensive review, Wickwire and colleagues delineated patient-centered influences including medical/physiologic, behavioral/motivational, and technical/device-related domains [19]. Results from the current study, as well as prior research examining predictors of CPAP adherence among older adults, have generally supported this framework. For example, we found that medical history (anemia) was associated with reduced CPAP adherence, a result generally consistent with Baratta and colleagues [36], who in a longitudinal cohort study among older adults with OSA (N = 295) found that past medical history (cardiovascular events) and current smoking were associated with reduced CPAP adherence. Our finding regarding anxiety, fibromyalgia, and TBI are also generally consistent with findings from a retrospective cohort study (N = 315) conducted by Yang and colleagues, who found that relative to older adults with fewer medical and psychiatric comorbidities, patients with more comorbidities were less likely to try CPAP [37]. Similarly, in terms of psychiatric comorbidity Ayalon and colleagues [38] found depression to be a risk factor for poor CPAP adherence among older adults with Alzheimer’s disease (N = 30). Finally, although we were unable to identify reasons for CPAP usage, technical/device-related barriers to CPAP usage have been identified among older adults, with reduced usability of CPAP machines (i.e. the ease of setting up, using, and cleaning the equipment) being associated with reduced CPAP adherence [39]. In addition, demographic and patient-centered characteristics such as age and SES [40–43], as well as system-level variables such as provider specialty board certification status [44–46], have been implicated in CPAP adherence, although results have been somewhat inconsistent across studies [19]. In the present study, we used Medicaid eligibility as a proxy for lower SES and found that relative to Medicare-only beneficiaries, beneficiaries with dual Medicare and Medicaid eligibility were less likely to adhere to CPAP. Relatedly, others have identified higher out-of-pocket costs and current smoking as reasons for older adults’ refusal to try CPAP [37]. Clearly, social determinants of health in clinical sleep medicine warrant further research attention. Results from this study suggest four important directions for future research. Most important, because our findings suggest that 90-day and longer term CPAP adherence among older adults is lower than previously reported, greater insight is needed into barriers to CPAP adherence among this population, as are targeted interventions to increase CPAP adherence among older adults. Results from the present study also highlight the important role of social determinants of health in clinical sleep science. Relative to Medicare-only beneficiaries, those eligible for both Medicare and Medicaid were significantly less likely to adhere to CPAP. Future research should seek to develop a deeper understanding of the mechanisms through which SES and other social determinants impact patient experience throughout the OSA diagnostic and treatment process, including receiving, acclimating, and adhering to CPAP therapy. Second, further research is needed to develop interventions to increase CPAP adherence among older adults generally, as well as among older adults of lower SES specifically. Third, the impact of CPAP adherence on health and economic outcomes among older adults with OSA should be examined. Finally, it will be important to validate our proxy measure of CPAP adherence by exploring the relation between administrative-level CPAP charges and objective CPAP use. This study possesses distinct strengths. First, we employed a large, national sample of Medicare beneficiaries, resulting in the largest analysis of CPAP adherence among older adults to date. Second, our data spanned the first 5 years after the implementation of the so-called “Medicare adherence criteria.” Thus, we were able to employ a novel, policy-based operational definition for CPAP adherence not previously published in the literature. Finally, we employed sound statistical analyses including multiple sensitivity analyses, and the pattern of results remained consistent throughout. However, our administrative methodology also has limitations. Most important, although Medicare claims directly capture health, economic, and policy outcomes from the payer perspective, we were unable to assess numerous clinical variables of interest, such as specific OSA characteristics and severity, daytime symptoms, objective CPAP usage, and other clinical measures that could have enriched findings and aided interpretation of results. Further, our operational definition of SES (i.e. dual eligibility for both Medicare and Medicaid) does not permit more nuanced analysis of social determinants of health such as income, wealth, occupation, educational attainment, residential neighborhood, and race/ethnicity. Such factors are increasingly recognized as important predictors of health and wellbeing [47]. Finally, although our sample was large and derived from a national sample, it is unknown how well our findings will generalize to all older adults. For example, we were unable to compare directly demographic or other characteristics between participants in this study and all Medicare beneficiaries. Similarly, there are known differences between fee-for-service and Medicare Advantage beneficiaries, with Medicare Advantage beneficiaries generally being healthier and consuming fewer healthcare resources [48, 49]. In conclusion, present results represent the largest study to date of rates and predictors of CPAP adherence among older adults in the United States. In our national sample of Medicare beneficiaries, adherence rates were generally lower than previously reported in smaller, clinic-based studies. In addition to psychiatric and medical comorbidities, lower SES was associated with reduced adherence. Future studies should seek to advance understanding of social determinants of CPAP adherence and outcomes, and to develop interventions to increase CPAP usage among older adult populations at-risk for poor CPAP adherence. Disclosure Statement Financial disclosure: This research was supported by an investigator-initiated grant awarded from ResMed to The University of Maryland, Baltimore (PI: Wickwire). EMW, SMS, and JSA’s institution has received research funding from the AASM Foundation, Department of Defense, Merck, and ResMed. EMW has served as a scientific consultant to DayZz, Eisai, Merck, and Purdue and is an equity shareholder in WellTap. LMC is a full-time employee and equity shareholder in ResMed. JSA is supported by Agency for Healthcare Research and Quality grant K01HS024560. Non-financial disclosure: None declared. References 1. Konecny T , et al. Under-diagnosis of sleep apnea in patients after acute myocardial infarction . J Am Coll Cardiol. 2010 ; 56 ( 9 ): 742 – 743 . Google Scholar Crossref Search ADS PubMed WorldCat 2. Mehra R , et al. Association of nocturnal arrhythmias with sleep-disordered breathing: the Sleep Heart Health Study . Am J Respir Crit Care Med. 2006 ; 173 ( 8 ): 910 – 916 . Google Scholar Crossref Search ADS PubMed WorldCat 3. Peppard PE , et al. Prospective study of the association between sleep-disordered breathing and hypertension . N Engl J Med. 2000 ; 342 ( 19 ): 1378 – 1384 . Google Scholar Crossref Search ADS PubMed WorldCat 4. Yaggi HK , et al. Obstructive sleep apnea as a risk factor for stroke and death . N Engl J Med. 2005 ; 353 ( 19 ): 2034 – 2041 . Google Scholar Crossref Search ADS PubMed WorldCat 5. Chan W , et al. Sleep apnea in patients with transient ischemic attack and minor stroke: opportunity for risk reduction of recurrent stroke? Stroke. 2010 ; 41 ( 12 ): 2973 – 2975 . Google Scholar Crossref Search ADS PubMed WorldCat 6. Punjabi NM , et al. Sleep-disordered breathing, glucose intolerance, and insulin resistance: the Sleep Heart Health Study . Am J Epidemiol. 2004 ; 160 ( 6 ): 521 – 530 . Google Scholar Crossref Search ADS PubMed WorldCat 7. Reichmuth KJ , et al. Association of sleep apnea and type II diabetes: a population-based study . Am J Respir Crit Care Med. 2005 ; 172 ( 12 ): 1590 – 1595 . Google Scholar Crossref Search ADS PubMed WorldCat 8. Drager LF , et al. The impact of obstructive sleep apnea on metabolic and inflammatory markers in consecutive patients with metabolic syndrome . PLoS One. 2010 ; 5 ( 8 ): e12065 . Google Scholar Crossref Search ADS PubMed WorldCat 9. Peppard PE , et al. Longitudinal association of sleep-related breathing disorder and depression . Arch Intern Med. 2006 ; 166 ( 16 ): 1709 – 1715 . Google Scholar Crossref Search ADS PubMed WorldCat 10. Finn L , et al. Sleep-disordered breathing and self-reported general health status in the Wisconsin Sleep Cohort Study . Sleep. 1998 ; 21 ( 7 ): 701 – 706 . Google Scholar PubMed OpenURL Placeholder Text WorldCat 11. Marshall NS , et al. Sleep apnea and 20-year follow-up for all-cause mortality, stroke, and cancer incidence and mortality in the Busselton Health Study cohort . J Clin Sleep Med. 2014 ; 10 ( 4 ): 355 – 362 . Google Scholar Crossref Search ADS PubMed WorldCat 12. Punjabi NM , et al. Sleep-disordered breathing and mortality: a prospective cohort study . PLoS Med. 2009 ; 6 ( 8 ): e1000132 . Google Scholar Crossref Search ADS PubMed WorldCat 13. Kryger MH , et al. Utilization of health care services in patients with severe obstructive sleep apnea . Sleep. 1996 ; 19 ( 9 Suppl ): S111 – S116 . Google Scholar Crossref Search ADS PubMed WorldCat 14. Banno K , et al. Expenditure on health care in obese women with and without sleep apnea . Sleep. 2009 ; 32 ( 2 ): 247 – 252 . Google Scholar Crossref Search ADS PubMed WorldCat 15. American Academy of Sleep Medicine . Hidden Health Crisis Costing America Billions. Underdiagnosing and Undertreating Obstructive Sleep Apnea Draining Healthcare System . Mountain View, CA : Frost & Sullivan ; 2016 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC 16. Bixler EO , et al. Effects of age on sleep apnea in men: I. Prevalence and severity . Am J Respir Crit Care Med. 1998 ; 157 ( 1 ): 144 – 148 . Google Scholar Crossref Search ADS PubMed WorldCat 17. Peppard PE , et al. Increased prevalence of sleep-disordered breathing in adults . Am J Epidemiol. 2013 ; 177 ( 9 ): 1006 – 1014 . Google Scholar Crossref Search ADS PubMed WorldCat 18. Ancoli-Israel S , et al. Sleep apnea in female patients in a nursing home. Increased risk of mortality . Chest. 1989 ; 96 ( 5 ): 1054 – 1058 . Google Scholar Crossref Search ADS PubMed WorldCat 19. Wickwire EM , et al. Maximizing positive airway pressure adherence in adults: a common-sense approach . Chest. 2013 ; 144 ( 2 ): 680 – 693 . Google Scholar Crossref Search ADS PubMed WorldCat 20. Nishihata Y , et al. Continuous positive airway pressure treatment improves cardiovascular outcomes in elderly patients with cardiovascular disease and obstructive sleep apnea . Heart Vessels. 2015 ; 30 ( 1 ): 61 – 69 . Google Scholar Crossref Search ADS PubMed WorldCat 21. López-Padilla D , et al. Continuous positive airway pressure and survival of very elderly persons with moderate to severe obstructive sleep apnea . Sleep Med. 2016 ; 19 : 23 – 29 . Google Scholar Crossref Search ADS PubMed WorldCat 22. Ou Q , et al. Continuous positive airway pressure treatment reduces mortality in elderly patients with moderate to severe obstructive severe sleep apnea: a cohort study . PLoS One. 2015 ; 10 ( 6 ): e0127775 . Google Scholar Crossref Search ADS PubMed WorldCat 23. Richards KC , et al. CPAP adherence may slow 1-year cognitive decline in older adults with mild cognitive impairment and apnea . J Am Geriatr Soc. 2019 ; 67 ( 3 ): 558 – 564 . Google Scholar Crossref Search ADS PubMed WorldCat 24. Weaver TE , et al. Continuous positive airway pressure treatment for sleep apnea in older adults . Sleep Med Rev. 2007 ; 11 ( 2 ): 99 – 111 . Google Scholar Crossref Search ADS PubMed WorldCat 25. Pallansch J , et al. Patient-reported outcomes in older adults with obstructive sleep apnea treated with continuous positive airway pressure therapy . J Clin Sleep Med. 2018 ; 14 ( 2 ): 215 – 222 . Google Scholar Crossref Search ADS PubMed WorldCat 26. Sawyer AM , et al. A systematic review of CPAP adherence across age groups: clinical and empiric insights for developing CPAP adherence interventions . Sleep Med Rev. 2011 ; 15 ( 6 ): 343 – 356 . Google Scholar Crossref Search ADS PubMed WorldCat 27. Parish JM , et al. Compliance with CPAP in elderly patients with OSA . Sleep Med. 2000 ; 1 ( 3 ): 209 – 214 . Google Scholar Crossref Search ADS PubMed WorldCat 28. Russo-Magno P , et al. Compliance with CPAP therapy in older men with obstructive sleep apnea . J Am Geriatr Soc. 2001 ; 49 ( 9 ): 1205 – 1211 . Google Scholar Crossref Search ADS PubMed WorldCat 29. Pelletier-Fleury N , et al. The age and other factors in the evaluation of compliance with nasal continuous positive airway pressure for obstructive sleep apnea syndrome. A Cox’s proportional hazard analysis . Sleep Med. 2001 ; 2 ( 3 ): 225 – 232 . Google Scholar Crossref Search ADS PubMed WorldCat 30. Woehrle H , et al. Age- and gender-dependent adherence with continuous positive airway pressure therapy . Sleep Med. 2011 ; 12 ( 10 ): 1034 – 1036 . Google Scholar Crossref Search ADS PubMed WorldCat 31. McMillan A , et al. A multicentre randomised controlled trial and economic evaluation of continuous positive airway pressure for the treatment of obstructive sleep apnoea syndrome in older people: PREDICT . Health Technol Assess. 2015 ; 19 ( 40 ): 1 – 188 . Google Scholar Crossref Search ADS WorldCat 32. Fulda S , et al. Cognitive dysfunction in sleep disorders . Sleep Med Rev. 2001 ; 5 ( 6 ): 423 – 445 . Google Scholar Crossref Search ADS PubMed WorldCat 33. Deyo RA , et al. Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases . J Clin Epidemiol. 1992 ; 45 ( 6 ): 613 – 619 . Google Scholar Crossref Search ADS PubMed WorldCat 34. McMillan A , et al. Continuous positive airway pressure in older people with obstructive sleep apnoea syndrome (PREDICT): a 12-month, multicentre, randomised trial . Lancet Respir Med. 2014 ; 2 ( 10 ): 804 – 812 . Google Scholar Crossref Search ADS PubMed WorldCat 35. Javaheri S , et al. Sleep apnea testing and outcomes in a large cohort of Medicare beneficiaries with newly diagnosed heart failure . Am J Respir Crit Care Med. 2011 ; 183 ( 4 ): 539 – 546 . Google Scholar Crossref Search ADS PubMed WorldCat 36. Baratta F , et al. Long-term prediction of adherence to continuous positive air pressure therapy for the treatment of moderate/severe obstructive sleep apnea syndrome . Sleep Med. 2018 ; 43 : 66 – 70 . Google Scholar Crossref Search ADS PubMed WorldCat 37. Yang MC , et al. Factors affecting CPAP acceptance in elderly patients with obstructive sleep apnea in Taiwan . Respir Care. 2013 ; 58 ( 9 ): 1504 – 1513 . Google Scholar Crossref Search ADS PubMed WorldCat 38. Ayalon L , et al. Adherence to continuous positive airway pressure treatment in patients with Alzheimer disease and obstructive sleep apnea . Am J Geriatr Psychiatry . 2006 ; 14 ( 2 ): 176 – 180 . Google Scholar Crossref Search ADS PubMed WorldCat 39. Fung CH , Martin JL, Hays RD, et al. Patient-reported usability of positive airway pressure equipment is associated with adherence in older adults . Sleep . 2017 ; 40 ( 3 ). doi:10.1093/sleep/zsx007. Google Scholar OpenURL Placeholder Text WorldCat 40. Simon-Tuval T , et al. Low socioeconomic status is a risk factor for CPAP acceptance among adult OSAS patients requiring treatment . Sleep. 2009 ; 32 ( 4 ): 545 – 552 . Google Scholar Crossref Search ADS PubMed WorldCat 41. Billings ME , et al. Race and residential socioeconomics as predictors of CPAP adherence . Sleep. 2011 ; 34 ( 12 ): 1653 – 1658 . Google Scholar Crossref Search ADS PubMed WorldCat 42. Platt AB , et al. Neighborhood of residence is associated with daily adherence to CPAP therapy . Sleep. 2009 ; 32 ( 6 ): 799 – 806 . Google Scholar Crossref Search ADS PubMed WorldCat 43. Somers ML , et al. Continuous positive airway pressure adherence for obstructive sleep apnea . ISRN Otolaryngol. 2011 ; 2011 : 943586 . Google Scholar Crossref Search ADS PubMed WorldCat 44. Scharf SM , et al. Comparison of primary‐care practitioners and sleep specialists in the treatment of obstructive sleep apnea . Sleep Breath . 2004 ; 8 ( 3 ): 111 – 124 . Google Scholar PubMed OpenURL Placeholder Text WorldCat 45. Parthasarathy S , et al. A multicenter prospective comparative effectiveness study of the effect of physician certification and center accreditation on patient-centered outcomes in obstructive sleep apnea . J Clin Sleep Med. 2014 ; 10 ( 3 ): 243 – 249 . Google Scholar Crossref Search ADS PubMed WorldCat 46. Pendharkar SR , et al. A randomized controlled trial of an alternative care provider clinic for severe sleep-disordered breathing . Ann Am Thorac Soc. 2019 ; 16 ( 12 ): 1558 – 1566 . Google Scholar Crossref Search ADS PubMed WorldCat 47. Hale L , et al. Perceived neighborhood quality, sleep quality, and health status: evidence from the Survey of the Health of Wisconsin . Soc Sci Med. 2013 ; 79 : 16 – 22 . Google Scholar Crossref Search ADS PubMed WorldCat 48. Landon BE , et al. Analysis of Medicare Advantage HMOs compared with traditional Medicare shows lower use of many services during 2003-09 . Health Aff (Millwood). 2012 ; 31 ( 12 ): 2609 – 2617 . Google Scholar Crossref Search ADS PubMed WorldCat 49. Landon BE , et al. A comparison of relative resource use and quality in Medicare Advantage health plans versus traditional Medicare . Am J Manag Care. 2015 ; 21 ( 8 ): 559 – 566 . Google Scholar PubMed OpenURL Placeholder Text WorldCat © Sleep Research Society 2020. Published by Oxford University Press on behalf of the Sleep Research Society. All rights reserved. For permissions, please e-mail [email protected]. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)

Colvonen, Peter J; Almklov, Erin; Tripp, Jessica C; Ulmer, Christi S; Pittman, James O E; Afari, Niloofar

doi: 10.1093/sleep/zsaa119pmid: 32529231

Abstract Study Objectives Post-9/11 veterans are particularly vulnerable to insomnia disorder. Having accurate prevalence rates of insomnia disorder in this relatively young, diverse population, is vital to determine the resources needed to identify and treat insomnia disorder. However, there are no accurate prevalence rates for insomnia disorder in post-9/11 veterans enrolling in the VA Healthcare System (VHA). We present accurate prevalence of insomnia disorder, and correlates, in a large sample of post-9/11 veterans enrolling in a VHA. Methods This was an observational study of 5,552 post-9/11 veterans newly enrolling for health care in a VHA. Data were collected using VA eScreening. Insomnia diagnosis was determined using a clinical cutoff score of ≥ 11 on the Insomnia Severity Index. Measures also included sociodemographic, service history, posttraumatic stress disorder (PTSD), depression, suicidal ideation, alcohol misuse, military sexual trauma, traumatic brain injury (TBI), and pain intensity. Results About 57.2% of the sample population had insomnia disorder. Our sample was nationally representative for age, sex, ethnicity, branch of the military, and race. The sample also was at high-risk for a host of clinical disorders, including PTSD, TBI, and pain; all of which showed higher rates of insomnia disorder (93.3%, 77.7%, and 69.6%, respectively). Conclusions The findings suggest alarmingly high rates of insomnia disorder in this population. Examining and treating insomnia disorder, especially in the context of co-occurring disorders (e.g. PTSD), will be a necessity in the future. insomnia, prevalence rates, co-occurring disorders Statement of Significance In an observational study of 5,552 post-9/11 veterans enrolling in the VA Healthcare System (VHA), 57.2% of the veterans were at high risk for insomnia disorder. Veterans with posttraumatic stress disorder, traumatic brain injury, and pain showed highest rates of insomnia disorder (93.3%, 77.7%, and 69.6%, respectively). Given the size of our sample, its sociodemographic representativeness, and the use of an independent, validated, standardized measure of insomnia, the prevalence rates reported are the most accurate in post-9/11 veterans using VHA care. Rates of insomnia disorder in post-9/11 veterans is alarmingly high and requires increased attention and direct treatment, especially in the context of co-occurring disorders. Introduction Chronic insomnia disorder is a behavioral sleep disorder characterized as dissatisfaction with sleep quantity or quality, marked by complaints of difficulty falling or staying asleep, waking up earlier than desired, and significant sleep-related daytime impairment [1]. Insomnia is linked to reduced quality of life [2], increased risk for morbidity [3–7], premature mortality [8, 9], and elevations in health care utilization [10, 11]. Insomnia is associated with impaired functioning across multiple life areas, including cognitive, emotional, social, and physical domains [12–15], with societal cost estimated at $100 billion per year [16]. Veterans are particularly vulnerable to insomnia [17–21], with rates double and even triple those of civilian populations [22], due to the irregularity of their sleep/wake schedules during active duty, harsh living conditions, combat stress, higher rates of physical and psychological injury, and issues associated with post-deployment reintegration [22–26]. Insomnia disorder among veterans Health Administration (VHA) users is expected to continue rising as post-9/11 troops leave military service and begin accessing VHA healthcare [25, 27]. Having accurate prevalence rates of insomnia disorder in this relatively young, diverse set of veterans, along with a better understanding of the demographic and clinical correlates, is vital for VHA to determine the resources needed to identify and treat insomnia disorder using evidence-based care. Unfortunately, there are no accurate prevalence rates of insomnia disorder in post-9/11 veterans seeking services. A confounding issue of estimating insomnia prevalence is the inconsistent use of insomnia terminology throughout the literature. The gold standard for diagnosing insomnia disorder is a weekly sleep diary and an in-depth clinician interview [28]. However, due to the assessment being time consuming, diagnostic criteria for insomnia disorder are rarely assessed outside of research trials [29]. Therefore, most prior studies only examined insomnia symptoms. To ensure clarity, we use “insomnia disorder” herein when diagnosis was established, “insomnia symptoms” when diagnosis was not established, and “insomnia” when the prior literature is unclear on this point. Rates of insomnia for veterans range from 3.4% based on medical records [30] to 90% of veterans with posttraumatic stress disorder (PTSD) reporting “sleep difficulties” [18]. These rates depend on the definition of insomnia (e.g. “trouble sleeping,” insomnia disorder), data source (e.g. chart review, self-report questionnaire), sample characteristics (e.g. treatment seeking, community), objective/subjective measurement (e.g. self-report questionnaire, sleep diary), measure (e.g. sleep items from mental health [MH] questionnaires, validated sleep questionnaire), co-occurring disorders (e.g. depression, PTSD), gender, and age. Several studies have reported prevalence rates of insomnia symptoms using “trouble falling or staying sleep” assessed from other MH questionnaires. They found rates of insomnia symptoms of 24.7%–30.5% [31, 32] in a general military/veteran population to upwards of 100% in Vietnam veterans with PTSD [33]. However, a single item about sleep is not a validated measure of insomnia diagnosis, may be anchored to other items in the questionnaire, and may be overly influenced by external factors [34]. Two studies used ICD-9 diagnostic codes from large national VHA databases to estimate prevalence rates of insomnia disorder and found 2.5% of 9,786,778 US veterans seeking VA healthcare from 2000 to 2010 had an insomnia diagnosis recorded in their medical record [23]; the one-year (2010) prevalence rate of insomnia diagnosis was 3.4% in veterans [30]. PTSD was associated with the highest prevalence of insomnia (16%) [23] and deployments to Iraq or Afghanistan conflicts along with anxiety and depressive disorders were also strongly related to insomnia [30]. Unfortunately, using medical records does not provide accurate prevalence of insomnia disorder. Only 53% of VA primary care providers indicated routinely documenting insomnia disorder and 39% routinely included it in the problem list [35–37]. Given the community insomnia rates of 6%–10%, reports of shorter sleep durations, longer sleep onset latencies, more wake after sleep onset, and lower sleep efficiencies in veterans [38–40], along with the well-known problems with obtaining prevalence rates from medical records [29, 35], the rates suggested by these large studies severely underrepresent the true prevalence of insomnia disorder in veterans. There are three generalizable insomnia disorder prevalence studies that used validated measures of insomnia with clinical cutoff scores on a general sample of post-9/11 veterans. They found rates of insomnia disorder of 59.1% in 375 post-9/11 veterans [41], 53.1% of veterans without military sexual trauma (MST) and 60.8% of veterans with MST (using a subset of the current study sample) [36], and 52.3% of veterans in 660 female veterans receiving services at VA primary care facilities [42]. While these studies used validated measures of insomnia, the samples were relatively small and need follow-up with larger samples. The existing literature on the prevalence of insomnia disorder is limited by assessing insomnia using non-validated questionnaires or single items, using medical record chart review, small sample sizes, and non-generalizable samples. Accurate estimates of the insomnia disorder rates in post-9/11 veterans is critical to better target and prioritize healthcare resources. This study aimed to addresses these gaps by including an independent sleep questionnaire validated against insomnia diagnostic criteria [43] to (1) obtain prevalence estimates of insomnia disorder in a large sample of post-9/11 veterans enrolling in a VHA and (2) examine the relationship between insomnia disorder and sociodemographics and other clinical symptoms. Methods Participants and procedures The data examined was obtained through standard clinical screening processes of newly enrolling post-9/11 veterans at the VA San Diego Healthcare System between March 2012 and April 2019. Self-reported data was collected electronically using the VA eScreening program or using paper-and-pencil assessment. eScreening is a computer-based, self-screening program that efficiently and effectively collects mental and physical health information [44]. Screening was conducted in the Transition Care Management program, which coordinates healthcare for newly enrolling post-9/11 veterans. Veterans with complete data from the insomnia severity index (ISI; N = 5,552) were included in the study; 768 veterans in the database did not have an ISI. There were no differences between those with and without an ISI by gender, race/ethnicity, combat exposure, PTSD, depression, suicidal ideation, education level, marital status, or pain (p > 0.05); individuals with ISI data had higher rates of alcohol misuse. All research was approved by the Institutional Review Board and the Research and Development Committee. Measures Sociodemographic and service history Age, gender, ethnicity, race, highest level of education, and relationship status was captured by an investigator-created self-report questionnaire. History related to service era, branch of service, number of deployments, and combat exposure was also assessed. Insomnia Insomnia severity over the past 2 weeks was measured using the ISI [45]. The ISI consists of seven items, assessing severity of insomnia as well as satisfaction with sleep, effect of sleep on daytime and social functioning, and concern about current sleep. The ISI ranges from 0 to 28, with higher scores indicating more severe insomnia symptoms. Research compared insomnia symptoms assessed with the ISI to weekly sleep diaries and overnight polysomnography for insomnia disorder and found a total score of ≥11 on the ISI indicates insomnia disorder in clinical samples with 97.2% sensitivity and 100% specificity [43]. Internal consistency was excellent for community and clinical sample [43] and for the current sample (α = 0.94). Posttraumatic stress disorder The 17-item PTSD Checklist—Civilian Version (PCL-C) [46] measured the degree to which respondents were bothered by PTSD symptoms within the past month. Scores range from 17 to 85, with higher scores indicating greater severity. Cutoff scores of >44 for presumed PTSD diagnosis were chosen based on research involving military and veteran samples [47]. The PCL-C showed strong internal consistency in a military population [46] and for the current sample (α = 0.96). The 20-item PCL-5 (PTSD Checklist–DSM-5) was subsequently used following the release of the DSM-5 [47]. Items correspond to the DSM-5 diagnostic criteria, and respondents rate how much they are bothered by PTSD symptoms over the past month. Scores range from 0 to 80; cutoff scores of >33 for presumed PTSD diagnosis on the PCL-5 were used based on research involving military and veteran samples [47]. Internal consistency for the current sample was strong (α = 0.98). Depression The Patient Health Questionnaire 9-Item Depression Module (PHQ-9) [48] measured depression symptoms occurring within the past 2 weeks; with scores ranging from 0 to 27. Higher scores indicate greater severity. The PHQ-9 is a reliable and valid measure of depression with a cutoff score of ≥15 indicates moderate to severe depression [48]. Internal consistency for the current sample was strong (α = 0.93). Suicidal ideation To assess suicidal ideation, we used the PHQ-9 suicide/self-harm item, “Thoughts that you would be better off dead, or thoughts of hurting yourself in some way?,” to indicate presence/absence of suicidal ideation. “0 = Not at all” was coded as negative suicidal ideation and “1 = Several days” or greater was coded as positive. Alcohol misuse The Alcohol Use Disorders Identification Test (AUDIT-C) [49] was used to screen hazardous alcohol consumption. Scores range from 0 to 12; scores of ≥4 and ≥3 suggest alcohol misuse in men and women, respectively [50]. This instrument has high internal consistency in the literature [51] and in the current sample (α = 0.80). Traumatic brain injury (TBI) History of TBI with concurrent related symptoms was assessed using the four-item VA TBI screen. A positive TBI screen required one or more positive responses in each of the following categories: a list of events in which an injury could have occurred, immediate symptoms following the event, new or worsening symptoms, and current symptoms. The VA TBI screen has high-internal consistency and test–retest reliability, high sensitivity, and moderate specificity [52]. Military sexual trauma MST was assessed by two VA created questions: “When you were in the military, did you ever receive uninvited and unwanted sexual attention (i.e. touching, pressure for sexual favors, verbal remarks)?” and “When you were in the military, did anyone ever use force or the threat of force to have sex with you against your will?” A positive screen required an affirmative answer to either of these questions [53]. Pain Pain intensity over the past four weeks was assessed using a numerical rating scale from 0 to 10, anchored at “no pain at all” and “worst pain ever,” respectively. A rating of ≥4 was considered to be clinically significant pain [54]. Statistical analyses Data were analyzed using descriptive statistics and frequencies. Chi-squares examined insomnia by grouped categories; follow-up 2 × 2 chi-squares examined significant omnibus results with the largest group being the comparison group. Cramer’s V examined effect sizes of all chi-squares, (.1 is a small effect, .3 is a medium effect, and .5 or larger is considered a large effect) and Cohen’s d for t-tests (.2 is a small effect, .5 is a medium effect, and .8 is a large effect size) [55]. Data were analyzed using Statistical Package for Social Sciences 26 (IBM, Inc.). p < 0.05 level was used to indicate significance. Results Sociodemographic and service history Table 1 presents the means and standard deviations of the sociodemographic and service history variables for the entire sample (N = 5,552). Consistent with national cohorts of veterans’ demographics, our sample was approximately 35 years old (M = 34.8, SD = 9.10), primarily male (82.8%), with nearly a quarter (24.2%) Hispanic, and more than half (53.6%) White [56]. There was an average of 1.93 deployments (SD = 1.23) and most (52.4%) deployed to the operations in Iraq and Afghanistan. Consistent with the characteristics of the region, the sample had proportionally more veterans of the Navy (51.2%) and Marines (30.5%) compared to the Active Duty Personnel Data (Army: 36.6%; Navy: 24.8%; Air Force: 24.3%; Marines: 14.2%) [56]. Table 1. Demographic characteristics of post-9/11 veterans (N = 5,552) Demographic variable . Total %/M (SD) . Demographic variable . Total %/M (SD) . Age 34.8 (9.10) Service/branch Sex  Army 13.2%  Men 82.8%  Air Force 3.8%  Women 17.2%  Marines 30.5% Marital status  National Guard 0.9%  Single 31.8%  Navy 51.2%  Married 50.9%  Coast Guard 0.5%  Divorced/Separated 17.3% Education Service operation  High school/GED 21.4%  Bosnia <1%  Some college 43.9%  Djbouti 2.8%  College graduate 22.8%  Gulf War 3.1%  Associates Degree 11.9%  Global War on Terror 11.0% Ethnicity  Kosovo 1.1%  Hispanic 24.2%  Latin America 1.6%  Non-Hispanic 69.2%  Libya <1%  Declined Answer 6.7%  Somalia 1.8% Race  OEF 24.6%  White 53.6%  OIF 25.4%  Black 14.7%  OND 2.4%  Pacific Islander/Asian 13.3%  Other 3.7%  Bi/Multi-Racial 6.9%  None 21.5%  American Indian/Alaskan 1.5% Combat exposure 53.7%  Other 2.9% Number of Deployments 1.93 (1.23)  Declined to Answer 7.1% Demographic variable . Total %/M (SD) . Demographic variable . Total %/M (SD) . Age 34.8 (9.10) Service/branch Sex  Army 13.2%  Men 82.8%  Air Force 3.8%  Women 17.2%  Marines 30.5% Marital status  National Guard 0.9%  Single 31.8%  Navy 51.2%  Married 50.9%  Coast Guard 0.5%  Divorced/Separated 17.3% Education Service operation  High school/GED 21.4%  Bosnia <1%  Some college 43.9%  Djbouti 2.8%  College graduate 22.8%  Gulf War 3.1%  Associates Degree 11.9%  Global War on Terror 11.0% Ethnicity  Kosovo 1.1%  Hispanic 24.2%  Latin America 1.6%  Non-Hispanic 69.2%  Libya <1%  Declined Answer 6.7%  Somalia 1.8% Race  OEF 24.6%  White 53.6%  OIF 25.4%  Black 14.7%  OND 2.4%  Pacific Islander/Asian 13.3%  Other 3.7%  Bi/Multi-Racial 6.9%  None 21.5%  American Indian/Alaskan 1.5% Combat exposure 53.7%  Other 2.9% Number of Deployments 1.93 (1.23)  Declined to Answer 7.1% OEF, operation enduring freedom; OIF, operation Iraqi freedom; OND: operation new dawn. Open in new tab Table 1. Demographic characteristics of post-9/11 veterans (N = 5,552) Demographic variable . Total %/M (SD) . Demographic variable . Total %/M (SD) . Age 34.8 (9.10) Service/branch Sex  Army 13.2%  Men 82.8%  Air Force 3.8%  Women 17.2%  Marines 30.5% Marital status  National Guard 0.9%  Single 31.8%  Navy 51.2%  Married 50.9%  Coast Guard 0.5%  Divorced/Separated 17.3% Education Service operation  High school/GED 21.4%  Bosnia <1%  Some college 43.9%  Djbouti 2.8%  College graduate 22.8%  Gulf War 3.1%  Associates Degree 11.9%  Global War on Terror 11.0% Ethnicity  Kosovo 1.1%  Hispanic 24.2%  Latin America 1.6%  Non-Hispanic 69.2%  Libya <1%  Declined Answer 6.7%  Somalia 1.8% Race  OEF 24.6%  White 53.6%  OIF 25.4%  Black 14.7%  OND 2.4%  Pacific Islander/Asian 13.3%  Other 3.7%  Bi/Multi-Racial 6.9%  None 21.5%  American Indian/Alaskan 1.5% Combat exposure 53.7%  Other 2.9% Number of Deployments 1.93 (1.23)  Declined to Answer 7.1% Demographic variable . Total %/M (SD) . Demographic variable . Total %/M (SD) . Age 34.8 (9.10) Service/branch Sex  Army 13.2%  Men 82.8%  Air Force 3.8%  Women 17.2%  Marines 30.5% Marital status  National Guard 0.9%  Single 31.8%  Navy 51.2%  Married 50.9%  Coast Guard 0.5%  Divorced/Separated 17.3% Education Service operation  High school/GED 21.4%  Bosnia <1%  Some college 43.9%  Djbouti 2.8%  College graduate 22.8%  Gulf War 3.1%  Associates Degree 11.9%  Global War on Terror 11.0% Ethnicity  Kosovo 1.1%  Hispanic 24.2%  Latin America 1.6%  Non-Hispanic 69.2%  Libya <1%  Declined Answer 6.7%  Somalia 1.8% Race  OEF 24.6%  White 53.6%  OIF 25.4%  Black 14.7%  OND 2.4%  Pacific Islander/Asian 13.3%  Other 3.7%  Bi/Multi-Racial 6.9%  None 21.5%  American Indian/Alaskan 1.5% Combat exposure 53.7%  Other 2.9% Number of Deployments 1.93 (1.23)  Declined to Answer 7.1% OEF, operation enduring freedom; OIF, operation Iraqi freedom; OND: operation new dawn. Open in new tab Prevalence of insomnia disorder and clinical characteristics Table 2 shows the sample sizes and percentages of veterans screening positive for insomnia disorder, PTSD, depression, suicidal ideation, alcohol misuse, TBI, MST, and clinically significant pain. We found 57.2% of the sample had insomnia disorder using clinical cutoff scores (ISI ≥ 11) versus 40.5% who screened positive when using a more conservative cutoff ISI ≥ 15; 31.5% of the sample showed no insomnia (ISI ≤ 9). The rates in women and men were 59.6% and 56.6%, respectively. Up to 69.6% of the sample were above the cutoff for other clinically significant conditions including PTSD, depression, alcohol misuse, TBI, MST, and clinically significant pain. Table 2. Prevalence of insomnia disorder and clinical characteristics (N = 5,552) Variable (n = sample size) . Total %*/M (SD) . ISI ≥ 11 (N = 5,552) 57.2% ISI ≥ 15—Moderate to Severe 40.5% PCL-C (n = 3,918) 36.57 (19.29)  PCL-C ≥ 44 32.2% PCL-5 (n = 1,480) 25.44 (22.62)  PCL-5 ≥ 33 35.8% Combined PCL above cutoff 33.2% PHQ-9 ≥ 15 (N = 5,552) 23.3% Suicidal ideation > 0 (n = 5,532) 14.6% Audit C  Male (n = 3,982) 3.53 (2.71)   % ≥ 4 44.9%  Female (n = 827) 2.47 (2.34)   % ≥ 3 39.4% TBI positive (n = 5,366) 25.14% MST positive (n = 3,423) 11.7% Pain ≥ 4 69.6% Variable (n = sample size) . Total %*/M (SD) . ISI ≥ 11 (N = 5,552) 57.2% ISI ≥ 15—Moderate to Severe 40.5% PCL-C (n = 3,918) 36.57 (19.29)  PCL-C ≥ 44 32.2% PCL-5 (n = 1,480) 25.44 (22.62)  PCL-5 ≥ 33 35.8% Combined PCL above cutoff 33.2% PHQ-9 ≥ 15 (N = 5,552) 23.3% Suicidal ideation > 0 (n = 5,532) 14.6% Audit C  Male (n = 3,982) 3.53 (2.71)   % ≥ 4 44.9%  Female (n = 827) 2.47 (2.34)   % ≥ 3 39.4% TBI positive (n = 5,366) 25.14% MST positive (n = 3,423) 11.7% Pain ≥ 4 69.6% ISI, insomnia severity index; MST, military sexual trauma; PTSD, posttraumatic stress disorder; PCL, PTSD clinician checklist; TBI, traumatic brain injury; PHQ9, patient health questionnaire 9-Item depression module; Audit, alcohol use disorders identification test. *Percentages based on the clinical cutoffs for each of the measures. Open in new tab Table 2. Prevalence of insomnia disorder and clinical characteristics (N = 5,552) Variable (n = sample size) . Total %*/M (SD) . ISI ≥ 11 (N = 5,552) 57.2% ISI ≥ 15—Moderate to Severe 40.5% PCL-C (n = 3,918) 36.57 (19.29)  PCL-C ≥ 44 32.2% PCL-5 (n = 1,480) 25.44 (22.62)  PCL-5 ≥ 33 35.8% Combined PCL above cutoff 33.2% PHQ-9 ≥ 15 (N = 5,552) 23.3% Suicidal ideation > 0 (n = 5,532) 14.6% Audit C  Male (n = 3,982) 3.53 (2.71)   % ≥ 4 44.9%  Female (n = 827) 2.47 (2.34)   % ≥ 3 39.4% TBI positive (n = 5,366) 25.14% MST positive (n = 3,423) 11.7% Pain ≥ 4 69.6% Variable (n = sample size) . Total %*/M (SD) . ISI ≥ 11 (N = 5,552) 57.2% ISI ≥ 15—Moderate to Severe 40.5% PCL-C (n = 3,918) 36.57 (19.29)  PCL-C ≥ 44 32.2% PCL-5 (n = 1,480) 25.44 (22.62)  PCL-5 ≥ 33 35.8% Combined PCL above cutoff 33.2% PHQ-9 ≥ 15 (N = 5,552) 23.3% Suicidal ideation > 0 (n = 5,532) 14.6% Audit C  Male (n = 3,982) 3.53 (2.71)   % ≥ 4 44.9%  Female (n = 827) 2.47 (2.34)   % ≥ 3 39.4% TBI positive (n = 5,366) 25.14% MST positive (n = 3,423) 11.7% Pain ≥ 4 69.6% ISI, insomnia severity index; MST, military sexual trauma; PTSD, posttraumatic stress disorder; PCL, PTSD clinician checklist; TBI, traumatic brain injury; PHQ9, patient health questionnaire 9-Item depression module; Audit, alcohol use disorders identification test. *Percentages based on the clinical cutoffs for each of the measures. Open in new tab Demographic correlates Table 3 shows the percentages and statistics of insomnia disorder by demographic categories. The percentages of veterans screening positive for insomnia disorder did not differ by sex or ethnicity. However, there were differences by marital status, race, and branch of service. While older age showed statistically higher insomnia, the effect size was very small. Married or separated veterans had higher rates of insomnia disorder than individuals who were single, although effect sizes were small. Veterans of all other races showed higher rates of insomnia disorder compared to White veterans, with small effect sizes. Veterans of the Air Force endorsed lower levels of insomnia disorder compared with veterans of the Navy, with a very small effect size. Table 3. Demographic characteristics by insomnia disorder Characteristics . ISI < 11 . ISI ≥ 11 . Stats . Effect size Cohen’s d/Cramer’s V . Age 34.32 (9.35) 35.17 (8.84) t = −3.43 (p < 0.001)* r = 0.09 Sex Χ 2 = 2.70 (p = 0.10) V = 0.02  Men 43.4% 56.6%  Women 40.4% 59.6% Marital status Χ 2 = 94.58 (p < 0.001)* V = 0.14  Single† 51.7% 48.3%  Married 40.9% 59.1% Χ 2 = 48.65 (p < 0.001)* V = 0.11  Separated 32.8% 67.2% Χ 2 = 84.13 (p < 0.001)* V = 0.18 Ethnicity Χ 2 = 0.612 (p = 0.737) V = 0.01  Hispanic 41.7% 58.3%  Non-Hispanic 43.0% 57% Race Χ 2 = 72.26 (p < 0.001)* V = 0.12  White† 47.3% 52.7%  Black 34.2% 65.8% Χ 2 = 43.74 (p < 0.001)* V = 0.11  Pacific Islander 30.2% 69.8% Χ 2 = 6.11 (p = 0.01)* V = 0.01  Asian 42.2% 57.8% Χ 2 = 5.59 (p = 0.02)* V = 0.04  Bi/Multi-Racial 33.5% 66.5% Χ 2 = 25.30 (p < 0.001)* V = 0.09  American Indian/Alaskan 32.9% 67.1% Χ 2 = 9.21 (p = 0.002)* V = 0.06 Service/branch Χ 2 = 17.95 (p = 0.003)* V = 0.06  Navy† 43.4% 56.6%  Army 41.3% 58.7% Χ 2 = 1.05 (p = 0.31) V = 0.02  Air Force 55.0% 45.0% Χ 2 = 10.63 (p < 0.001)* V = 0.06  Marines 40.9% 59.1% Χ 2 = 2.63 (p = 0.11) V = 0.02  National Guard 35.3% 64.7% Χ 2 = 1.35 (p = 0.25) V = 0.02  Coast Guard 50.0% 50.0% Χ 2 = 0.46 (p = 0.50) V = 0.01 Characteristics . ISI < 11 . ISI ≥ 11 . Stats . Effect size Cohen’s d/Cramer’s V . Age 34.32 (9.35) 35.17 (8.84) t = −3.43 (p < 0.001)* r = 0.09 Sex Χ 2 = 2.70 (p = 0.10) V = 0.02  Men 43.4% 56.6%  Women 40.4% 59.6% Marital status Χ 2 = 94.58 (p < 0.001)* V = 0.14  Single† 51.7% 48.3%  Married 40.9% 59.1% Χ 2 = 48.65 (p < 0.001)* V = 0.11  Separated 32.8% 67.2% Χ 2 = 84.13 (p < 0.001)* V = 0.18 Ethnicity Χ 2 = 0.612 (p = 0.737) V = 0.01  Hispanic 41.7% 58.3%  Non-Hispanic 43.0% 57% Race Χ 2 = 72.26 (p < 0.001)* V = 0.12  White† 47.3% 52.7%  Black 34.2% 65.8% Χ 2 = 43.74 (p < 0.001)* V = 0.11  Pacific Islander 30.2% 69.8% Χ 2 = 6.11 (p = 0.01)* V = 0.01  Asian 42.2% 57.8% Χ 2 = 5.59 (p = 0.02)* V = 0.04  Bi/Multi-Racial 33.5% 66.5% Χ 2 = 25.30 (p < 0.001)* V = 0.09  American Indian/Alaskan 32.9% 67.1% Χ 2 = 9.21 (p = 0.002)* V = 0.06 Service/branch Χ 2 = 17.95 (p = 0.003)* V = 0.06  Navy† 43.4% 56.6%  Army 41.3% 58.7% Χ 2 = 1.05 (p = 0.31) V = 0.02  Air Force 55.0% 45.0% Χ 2 = 10.63 (p < 0.001)* V = 0.06  Marines 40.9% 59.1% Χ 2 = 2.63 (p = 0.11) V = 0.02  National Guard 35.3% 64.7% Χ 2 = 1.35 (p = 0.25) V = 0.02  Coast Guard 50.0% 50.0% Χ 2 = 0.46 (p = 0.50) V = 0.01 *Significant p < .05. †Comparison group for all analyses. Open in new tab Table 3. Demographic characteristics by insomnia disorder Characteristics . ISI < 11 . ISI ≥ 11 . Stats . Effect size Cohen’s d/Cramer’s V . Age 34.32 (9.35) 35.17 (8.84) t = −3.43 (p < 0.001)* r = 0.09 Sex Χ 2 = 2.70 (p = 0.10) V = 0.02  Men 43.4% 56.6%  Women 40.4% 59.6% Marital status Χ 2 = 94.58 (p < 0.001)* V = 0.14  Single† 51.7% 48.3%  Married 40.9% 59.1% Χ 2 = 48.65 (p < 0.001)* V = 0.11  Separated 32.8% 67.2% Χ 2 = 84.13 (p < 0.001)* V = 0.18 Ethnicity Χ 2 = 0.612 (p = 0.737) V = 0.01  Hispanic 41.7% 58.3%  Non-Hispanic 43.0% 57% Race Χ 2 = 72.26 (p < 0.001)* V = 0.12  White† 47.3% 52.7%  Black 34.2% 65.8% Χ 2 = 43.74 (p < 0.001)* V = 0.11  Pacific Islander 30.2% 69.8% Χ 2 = 6.11 (p = 0.01)* V = 0.01  Asian 42.2% 57.8% Χ 2 = 5.59 (p = 0.02)* V = 0.04  Bi/Multi-Racial 33.5% 66.5% Χ 2 = 25.30 (p < 0.001)* V = 0.09  American Indian/Alaskan 32.9% 67.1% Χ 2 = 9.21 (p = 0.002)* V = 0.06 Service/branch Χ 2 = 17.95 (p = 0.003)* V = 0.06  Navy† 43.4% 56.6%  Army 41.3% 58.7% Χ 2 = 1.05 (p = 0.31) V = 0.02  Air Force 55.0% 45.0% Χ 2 = 10.63 (p < 0.001)* V = 0.06  Marines 40.9% 59.1% Χ 2 = 2.63 (p = 0.11) V = 0.02  National Guard 35.3% 64.7% Χ 2 = 1.35 (p = 0.25) V = 0.02  Coast Guard 50.0% 50.0% Χ 2 = 0.46 (p = 0.50) V = 0.01 Characteristics . ISI < 11 . ISI ≥ 11 . Stats . Effect size Cohen’s d/Cramer’s V . Age 34.32 (9.35) 35.17 (8.84) t = −3.43 (p < 0.001)* r = 0.09 Sex Χ 2 = 2.70 (p = 0.10) V = 0.02  Men 43.4% 56.6%  Women 40.4% 59.6% Marital status Χ 2 = 94.58 (p < 0.001)* V = 0.14  Single† 51.7% 48.3%  Married 40.9% 59.1% Χ 2 = 48.65 (p < 0.001)* V = 0.11  Separated 32.8% 67.2% Χ 2 = 84.13 (p < 0.001)* V = 0.18 Ethnicity Χ 2 = 0.612 (p = 0.737) V = 0.01  Hispanic 41.7% 58.3%  Non-Hispanic 43.0% 57% Race Χ 2 = 72.26 (p < 0.001)* V = 0.12  White† 47.3% 52.7%  Black 34.2% 65.8% Χ 2 = 43.74 (p < 0.001)* V = 0.11  Pacific Islander 30.2% 69.8% Χ 2 = 6.11 (p = 0.01)* V = 0.01  Asian 42.2% 57.8% Χ 2 = 5.59 (p = 0.02)* V = 0.04  Bi/Multi-Racial 33.5% 66.5% Χ 2 = 25.30 (p < 0.001)* V = 0.09  American Indian/Alaskan 32.9% 67.1% Χ 2 = 9.21 (p = 0.002)* V = 0.06 Service/branch Χ 2 = 17.95 (p = 0.003)* V = 0.06  Navy† 43.4% 56.6%  Army 41.3% 58.7% Χ 2 = 1.05 (p = 0.31) V = 0.02  Air Force 55.0% 45.0% Χ 2 = 10.63 (p < 0.001)* V = 0.06  Marines 40.9% 59.1% Χ 2 = 2.63 (p = 0.11) V = 0.02  National Guard 35.3% 64.7% Χ 2 = 1.35 (p = 0.25) V = 0.02  Coast Guard 50.0% 50.0% Χ 2 = 0.46 (p = 0.50) V = 0.01 *Significant p < .05. †Comparison group for all analyses. Open in new tab Clinical correlates Figure 1 shows percentages of insomnia disorder by clinical correlates and Table 4 shows percentages and statistics. Neither the number of deployments, nor screening positive for alcohol misuse showed differences in screening positive for insomnia disorder. Veterans who screened positive for PTSD, depression, suicidal ideation, TBI, MST, clinically significant pain, or having exposure to combat showed higher rates of insomnia disorder. Screening positive for PTSD or MST showed large effect sizes, depression, suicidal ideation, and pain showed medium effects, while TBI and combat exposure showed small effect sizes. PTSD and depression scores were highly correlated (r (5,481) = .85, p < 0.001) suggesting they may be overlapping constructs or disorders. Figure 1. Open in new tabDownload slide Percentage of clinical correlates by insomnia disorder. Figure 1. Open in new tabDownload slide Percentage of clinical correlates by insomnia disorder. Table 4. Clinical correlates by insomnia disorder Characteristics . ISI < 11 . ISI ≥ 11 . Stats . Effect size Cohen’s d/Cramer’s V . # of Deployments 1.89 (1.21) 1.96 (1.24) t = −1.47 (p = 1.41) r = 0.06 Combat exposure Χ 2 =132.06 (p < 0.001)* V = 0.16  Yes 35.6% 64.4%  No 50.9% 49.1% PTSD Χ 2 = 1,445.16 (p < 0.001)* V = 0.52  PCL+ 6.7% 93.3%  PCL− 61.1% 38.9% Depression Χ 2 = 970.37 (p < 0.001)* V = 0.42  PHQ9+ 5.0% 95.0%  PHQ9− 54.3% 45.7% Suicidal ideation Χ 2 =356.98 (p < 0.001)* V = 0.25  Ideation+ 12.3% 87.7%  Ideation− 47.9% 52.1% Alcohol misuse Χ 2 =1.89 (p = 0.17) V = 0.04  Audit+ 48.7% 51.3%  Audit− 53.0% 47.0% TBI Χ 2 = 57.58 (p < 0.001)* V = 0.16  TBI+ 22.3% 77.7%  TBI− 39.0% 61.0% MST Χ 2 = 4.87 (p = 0.02)* V = 0.70  MST+ 39.9% 61.0%  MST− 51.7% 48.3% Pain Χ 2 = 489.70 (p < 0.001)* V = 0.35  Pain+ 29.5% 69.6%  Pain− 66.8% 33.2% Characteristics . ISI < 11 . ISI ≥ 11 . Stats . Effect size Cohen’s d/Cramer’s V . # of Deployments 1.89 (1.21) 1.96 (1.24) t = −1.47 (p = 1.41) r = 0.06 Combat exposure Χ 2 =132.06 (p < 0.001)* V = 0.16  Yes 35.6% 64.4%  No 50.9% 49.1% PTSD Χ 2 = 1,445.16 (p < 0.001)* V = 0.52  PCL+ 6.7% 93.3%  PCL− 61.1% 38.9% Depression Χ 2 = 970.37 (p < 0.001)* V = 0.42  PHQ9+ 5.0% 95.0%  PHQ9− 54.3% 45.7% Suicidal ideation Χ 2 =356.98 (p < 0.001)* V = 0.25  Ideation+ 12.3% 87.7%  Ideation− 47.9% 52.1% Alcohol misuse Χ 2 =1.89 (p = 0.17) V = 0.04  Audit+ 48.7% 51.3%  Audit− 53.0% 47.0% TBI Χ 2 = 57.58 (p < 0.001)* V = 0.16  TBI+ 22.3% 77.7%  TBI− 39.0% 61.0% MST Χ 2 = 4.87 (p = 0.02)* V = 0.70  MST+ 39.9% 61.0%  MST− 51.7% 48.3% Pain Χ 2 = 489.70 (p < 0.001)* V = 0.35  Pain+ 29.5% 69.6%  Pain− 66.8% 33.2% *Significant p < 0.05. ISI, insomnia severity index; MST, military sexual trauma; PTSD, posttraumatic stress disorder; PCL, PTSD checklist; TBI, traumatic brain injury; PHQ9, patient health questionnaire 9-Item depression module; Audit, alcohol use disorders identification test. Open in new tab Table 4. Clinical correlates by insomnia disorder Characteristics . ISI < 11 . ISI ≥ 11 . Stats . Effect size Cohen’s d/Cramer’s V . # of Deployments 1.89 (1.21) 1.96 (1.24) t = −1.47 (p = 1.41) r = 0.06 Combat exposure Χ 2 =132.06 (p < 0.001)* V = 0.16  Yes 35.6% 64.4%  No 50.9% 49.1% PTSD Χ 2 = 1,445.16 (p < 0.001)* V = 0.52  PCL+ 6.7% 93.3%  PCL− 61.1% 38.9% Depression Χ 2 = 970.37 (p < 0.001)* V = 0.42  PHQ9+ 5.0% 95.0%  PHQ9− 54.3% 45.7% Suicidal ideation Χ 2 =356.98 (p < 0.001)* V = 0.25  Ideation+ 12.3% 87.7%  Ideation− 47.9% 52.1% Alcohol misuse Χ 2 =1.89 (p = 0.17) V = 0.04  Audit+ 48.7% 51.3%  Audit− 53.0% 47.0% TBI Χ 2 = 57.58 (p < 0.001)* V = 0.16  TBI+ 22.3% 77.7%  TBI− 39.0% 61.0% MST Χ 2 = 4.87 (p = 0.02)* V = 0.70  MST+ 39.9% 61.0%  MST− 51.7% 48.3% Pain Χ 2 = 489.70 (p < 0.001)* V = 0.35  Pain+ 29.5% 69.6%  Pain− 66.8% 33.2% Characteristics . ISI < 11 . ISI ≥ 11 . Stats . Effect size Cohen’s d/Cramer’s V . # of Deployments 1.89 (1.21) 1.96 (1.24) t = −1.47 (p = 1.41) r = 0.06 Combat exposure Χ 2 =132.06 (p < 0.001)* V = 0.16  Yes 35.6% 64.4%  No 50.9% 49.1% PTSD Χ 2 = 1,445.16 (p < 0.001)* V = 0.52  PCL+ 6.7% 93.3%  PCL− 61.1% 38.9% Depression Χ 2 = 970.37 (p < 0.001)* V = 0.42  PHQ9+ 5.0% 95.0%  PHQ9− 54.3% 45.7% Suicidal ideation Χ 2 =356.98 (p < 0.001)* V = 0.25  Ideation+ 12.3% 87.7%  Ideation− 47.9% 52.1% Alcohol misuse Χ 2 =1.89 (p = 0.17) V = 0.04  Audit+ 48.7% 51.3%  Audit− 53.0% 47.0% TBI Χ 2 = 57.58 (p < 0.001)* V = 0.16  TBI+ 22.3% 77.7%  TBI− 39.0% 61.0% MST Χ 2 = 4.87 (p = 0.02)* V = 0.70  MST+ 39.9% 61.0%  MST− 51.7% 48.3% Pain Χ 2 = 489.70 (p < 0.001)* V = 0.35  Pain+ 29.5% 69.6%  Pain− 66.8% 33.2% *Significant p < 0.05. ISI, insomnia severity index; MST, military sexual trauma; PTSD, posttraumatic stress disorder; PCL, PTSD checklist; TBI, traumatic brain injury; PHQ9, patient health questionnaire 9-Item depression module; Audit, alcohol use disorders identification test. Open in new tab Discussion We found that 57.2% of a large sample of post-9/11 veterans entering the VHA screened positive for insomnia disorder, which suggests insomnia disorder is alarmingly common in this population. Given the size of our sample, its sociodemographic representativeness, and the use of an independent, validated, standardized measure of insomnia [43], the prevalence rates reported are the most accurate in post-9/11 veterans using VHA care. Confidence in our findings is further supported with comparison to previous generalizable research that found between 52.3% and 59.1% of veterans seeking care had insomnia disorder [36, 41, 42]. Insomnia is predictive of the development and recurrence of mood disorders [57, 58] and suicidality [59, 60], and is associated with the severity of psychiatric conditions (e.g. PTSD) [36]. Collectively, insomnia disorder is a risk factor for the most common medical and MH conditions seen in veterans, yet, insomnia disorder often goes unscreened and untreated in the VHA. The need to increase insomnia screening using independent measures for all veterans and offer evidence-based treatment is critical and is consistent with recommendations made by VA clinical practice guidelines [61]. The discrepancy between actual prevalence rates of insomnia disorder and what providers are documenting is larger than previously thought. Medical record codes have found rates 2.5% and 3.4% [23, 30], while we found 57.2% using an independent measure of insomnia. This confirms and expands on the finding that fewer than half of VHA providers routinely include insomnia in the encounter problem list [35]. Grandner and Chakravorty [29] suggest that many providers consider insomnia a symptom or secondary condition of a “primary disorder” rather than a comorbid diagnosis. Providers often focus on treating the “root cause” of insomnia (e.g. PTSD, pain) instead of targeting perpetuating factors (e.g. conditioned arousal), which is necessary for effective insomnia treatment. The prevalence rates in this study were also markedly different than studies that used a single questionnaire item of “trouble falling or staying asleep” [18, 31–33]. The differences could be based on different samples (e.g. military vs. veteran or Vietnam era vs. post-9/11), co-occurring disorders, and age. However, a single item about sleep is not a validated measure, misses the chronic relationship between sleep and dissatisfaction, interference with daily life, and cognitive shift or anxiety involved with insomnia disorder [34]. As such, the use of insomnia or even insomnia symptoms when using a single sleep item can be misleading when trying to assess prevalence rates in various populations. We suggest regular use of validated insomnia questionnaires, as well as clearer guidelines on insomnia language in peer-reviewed publications. Veterans who screened positive for MH disorders, increased combat exposure, positive for TBI and MST, and higher pain had considerably higher rates of insomnia disorder, consistent with previous literature [19, 41, 62, 63]. Interestingly, neither the number of deployments, nor screening positive for alcohol misuse, showed differences in screening positive for insomnia disorder. Our study also found that screening positive for suicidal ideation was strongly associated with insomnia disorder. Further, screening positive for PTSD and depression was associated with incredibly high rates of insomnia disorder. There are several possibilities for this finding. First, the higher prevalence of insomnia reported here, as compared to previous papers, may reflect high rates of combat exposure (53.7%) in our sample. Second, the co-occurrence of high PTSD and depression may have an additive effect in insomnia prevalence. Third, it is possible that while the ISI ≥ 11 is a reliable cutoff for a clinical sample, the cutoff for a more complex PTSD veteran sample needs to be higher (e.g. ≥15). Our study suggests a critical need to increase evidence-based treatments for insomnia, especially in individuals with co-occurring disorders; this may be especially pertinent to help combat suicidal ideation in veterans [64, 65]. Cognitive behavioral therapy for insomnia (CBT-I) is considered the front-line treatment for insomnia [66] and is better than pharmacological interventions [67] with longer lasting positive outcomes [68]. Despite clear recommendations in the Department of Veterans Affairs/Department of Defense (VA/DOD) Clinical Practice guidelines [61], access to CBT-I as the standard of care for insomnia disorder is limited [69]. There is even increasing evidence in support of integrated treatment, providing CBT-I plus evidence-based treatment for the primary presenting disorder [70]. Together examining and treating insomnia disorder in the context of co-occurring disorders, especially PTSD, depression, and suicidal ideation in the VHA system is a necessity. This study has several limitations. First, this sample of post-9/11 veterans may not be representative of veterans of all ages and eras using the VHA, or veterans who choose to receive their healthcare outside the VHA. Second, data were collected at time of registration for VA healthcare and not in specific clinics, limiting knowledge of what type of care veterans sought or our ability to address differences in veterans seeking mental versus physical health care. It is possible that not everyone presenting for enrollment was screened, meaning our findings may be impacted by self-selection bias. Finally, we were unable to screen for obstructive sleep apnea (OSA), which is highly co-occurring with insomnia as well as physical and MH correlates we examined, even in younger healthier veterans [71]. As such, the presence of OSA may influence insomnia prevalence rates, especially with co-occurring disorders [34]. While our sample was comprised only of post-9/11 veterans, the relevance of our cohort characteristics will only increase as they increasingly represent the veteran population using the VHA system in the future. The prevalence rates suggest that a considerable proportion of veterans enrolling in VA healthcare have clinically significant levels of insomnia. Despite clear recommendations in the VA/DOD Clinical Practice guidelines [61], access to CBT-I as the standard of care for insomnia disorder alone or with co-occurring disorders is limited [69]. Our findings, suggest heightened need for insomnia screening, documentation, and access to independent and direct interventions for insomnia in veterans entering the VHA system. Targeted efforts to increase veteran access to CBT-I, and behavioral sleep medicine, are essential in forestalling the array of adverse health consequences of chronic insomnia disorder. Acknowledgments The views expressed in this article are those of the authors only and do not reflect the official policy or position of the institutions with which the authors are affiliated, the Department of Veteran’s Affairs, nor the United States Government. Funding This material is the result of work supported by Department of Veterans Affairs (VA) Center of Excellence for Stress and Mental Health (CESAMH) and the VA Center for Innovation. This material also is the result of work supported with resources of the VA San Diego Healthcare System and the Durham Center of Innovation to Accelerate Discovery and Practice Transformation (ADAPT), (CIN 13–410) at the Durham VA Health Care System. Dr. Peter Colvonen is partly funded by VA RR&D CDA Grant #1lK2Rx002120-01. Dr Jessica Tripp is funded by a fellowship in VA Interprofessional Advanced Fellowship in Addiction Treatment supported by the office of academic affiliations. Disclosure Statement Financial disclosures: None declared. Non-Financial disclosure: None of the authors have any competing financial interests to disclose. References 1. American Psychiatric Association . Diagnostic and Statistical Manual of Mental Disorders. 5th ed. Arlington, VA : American Psychiatric Publishing ; 2013 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC 2. Katz DA , et al. The relationship between insomnia and health-related quality of life in patients with chronic illness . J Fam Pract. 2002 ; 51 ( 3 ): 229 – 235 . Google Scholar PubMed OpenURL Placeholder Text WorldCat 3. Knutson KL , et al. Associations between sleep loss and increased risk of obesity and diabetes . Ann N Y Acad Sci. 2008 ; 1129 : 287 – 304 . Google Scholar Crossref Search ADS PubMed WorldCat 4. Buxton OM , et al. Short and long sleep are positively associated with obesity, diabetes, hypertension, and cardiovascular disease among adults in the United States . Soc Sci Med. 2010 ; 71 ( 5 ): 1027 – 1036 . Google Scholar Crossref Search ADS PubMed WorldCat 5. Xi B , et al. Short sleep duration predicts risk of metabolic syndrome: a systematic review and meta-analysis . Sleep Med Rev. 2014 ; 18 ( 4 ): 293 – 297 . Google Scholar Crossref Search ADS PubMed WorldCat 6. Calhoun DA , Harding SM. Sleep and hypertension . Chest. 2010 ; 138 ( 2 ): 434 – 443 . Google Scholar Crossref Search ADS Google Preview WorldCat COPAC 7. Li M , et al. Insomnia and risk of cardiovascular disease: a meta-analysis of cohort studies . Int J Cardiol. 2014 ; 176 ( 3 ): 1044 – 1047 . Google Scholar Crossref Search ADS PubMed WorldCat 8. Dew MA , et al. Healthy older adults’ sleep predicts all-cause mortality at 4 to 19 years of follow-up . Psychosom Med. 2003 ; 65 ( 1 ): 63 – 73 . Google Scholar Crossref Search ADS PubMed WorldCat 9. Kripke DF , et al. Mortality associated with sleep duration and insomnia . Arch Gen Psychiatry. 2002 ; 59 ( 2 ): 131 – 136 . Google Scholar Crossref Search ADS PubMed WorldCat 10. Vgontzas AN , et al. Insomnia with objective short sleep duration: the most biologically severe phenotype of the disorder . Sleep Med Rev. 2013 ; 17 ( 4 ): 241 – 254 . Google Scholar Crossref Search ADS PubMed WorldCat 11. Fernandez-Mendoza J , et al. Insomnia and its impact on physical and mental health . Curr Psychiatry Rep. 2013 ; 15 ( 12 ): 418 . Google Scholar Crossref Search ADS PubMed WorldCat 12. Killgore WD , et al. Impaired decision making following 49 h of sleep deprivation . J Sleep Res. 2006 ; 15 ( 1 ): 7 – 13 . Google Scholar Crossref Search ADS PubMed WorldCat 13. Killgore WD , et al. Sleep deprivation reduces perceived emotional intelligence and constructive thinking skills . Sleep Med. 2008 ; 9 ( 5 ): 517 – 526 . Google Scholar Crossref Search ADS PubMed WorldCat 14. Pilcher JJ , et al. Effects of sleep deprivation on performance: a meta-analysis . Sleep. 1996 ; 19 ( 4 ): 318 – 326 . Google Scholar Crossref Search ADS PubMed WorldCat 15. Szentkirályi A , et al. Sleep disorders: impact on daytime functioning and quality of life . Expert Rev Pharmacoecon Outcomes Res. 2009 ; 9 ( 1 ): 49 – 64 . Google Scholar Crossref Search ADS PubMed WorldCat 16. Léger D , et al. Societal costs of insomnia . Sleep Med Rev. 2010 ; 14 ( 6 ): 379 – 389 . Google Scholar Crossref Search ADS PubMed WorldCat 17. Bramoweth AD , et al. Deployment-related insomnia in military personnel and veterans . Curr Psychiatry Rep. 2013 ; 15 ( 10 ): 401 . Google Scholar Crossref Search ADS PubMed WorldCat 18. Seelig AD , et al. Sleep patterns before, during, and after deployment to Iraq and Afghanistan . Sleep. 2010 ; 33 ( 12 ): 1615 – 1622 . Google Scholar Crossref Search ADS PubMed WorldCat 19. McLay RN , et al. Insomnia is the most commonly reported symptom and predicts other symptoms of post-traumatic stress disorder in US service members returning from military deployments . Mil Med. 2010 ; 175 ( 10 ): 759 – 762 . Google Scholar Crossref Search ADS PubMed WorldCat 20. Hughes J , et al. Insomnia and symptoms of post-traumatic stress disorder among women veterans . Behav Sleep Med. 2013 ; 11 ( 4 ): 258 – 274 . Google Scholar Crossref Search ADS PubMed WorldCat 21. Fung CH , et al. Prevalence and symptoms of occult sleep disordered breathing among older veterans with insomnia . J Clin Sleep Med. 2013 ; 9 ( 11 ): 1173 – 1178 . Google Scholar Crossref Search ADS PubMed WorldCat 22. Morin CM , Jarrin DC. Epidemiology of insomnia: prevalence, course, risk factors, and public health burden . Sleep Med. Clin. 2013 ; 8 ( 3 ): 281 – 297 . Google Scholar Crossref Search ADS WorldCat 23. Alexander M , et al. The national veteran sleep disorder study: descriptive epidemiology and secular trends, 2000–2010 . Sleep . 2016 ; 39 ( 7 ): 1399 – 1410 . Google Scholar Crossref Search ADS PubMed WorldCat 24. Ford ES , et al. Trends in insomnia and excessive daytime sleepiness among U.S. adults from 2002 to 2012 . Sleep Med. 2015 ; 16 ( 3 ): 372 – 378 . Google Scholar Crossref Search ADS PubMed WorldCat 25. Hughes JM , et al. Insomnia in United States military veterans: an integrated theoretical model . Clin Psychol Rev. 2018 ; 59 : 118 – 125 . Google Scholar Crossref Search ADS PubMed WorldCat 26. Roth T . Insomnia: definition, prevalence, etiology, and consequences . J Clin Sleep Med. 2007 ; 3 ( 5 Suppl ): S7 – 10 . Google Scholar PubMed OpenURL Placeholder Text WorldCat 27. Mysliwiec V , et al. Sleep disorders in US military personnel: a high rate of comorbid insomnia and obstructive sleep apnea . Chest. 2013 ; 144 ( 2 ): 549 – 557 . Google Scholar Crossref Search ADS PubMed WorldCat 28. Schutte-Rodin S , et al. Clinical guideline for the evaluation and management of chronic insomnia in adults . J Clin Sleep Med. 2008 ; 4 ( 5 ): 487 – 504 . Google Scholar Crossref Search ADS PubMed WorldCat 29. Grandner MA , et al. Insomnia in primary care: misreported, mishandled, and just plain missed . J Clin Sleep Med. 2017 ; 13 ( 8 ): 937 – 939 . Google Scholar Crossref Search ADS PubMed WorldCat 30. Hermes E , et al. Prevalence, pharmacotherapy and clinical correlates of diagnosed insomnia among Veterans Health Administration service users nationally . Sleep Med. 2014 ; 15 ( 5 ): 508 – 514 . Google Scholar Crossref Search ADS PubMed WorldCat 31. Hoge CW , et al. Mild traumatic brain injury in US soldiers returning from Iraq . N Engl J Med. 2008 ; 358 ( 5 ): 453 – 463 . Google Scholar Crossref Search ADS PubMed WorldCat 32. Seelig AD , et al. Sleep and health resilience metrics in a large military cohort . Sleep. 2016 ; 39 ( 5 ): 1111 – 1120 . Google Scholar Crossref Search ADS PubMed WorldCat 33. Neylan TC , et al. Sleep disturbances in the Vietnam generation: findings from a nationally representative sample of male Vietnam veterans . Am J Psychiatry. 1998 ; 155 ( 7 ): 929 – 933 . Google Scholar Crossref Search ADS PubMed WorldCat 34. Colvonen PJ , et al. Recent advancements in treating sleep disorders in co-occurring PTSD . Curr Psychiatry Rep. 2018 ; 20 ( 7 ): 48 . Google Scholar Crossref Search ADS PubMed WorldCat 35. Ulmer CS , et al. Veterans affairs primary care provider perceptions of insomnia treatment . J Clin Sleep Med. 2017 ; 13 ( 8 ): 991 – 999 . Google Scholar Crossref Search ADS PubMed WorldCat 36. Jenkins MM , et al. Prevalence and mental health correlates of insomnia in first-encounter veterans with and without military sexual trauma . Sleep. 2015 ; 38 ( 10 ): 1547 – 1554 . Google Scholar Crossref Search ADS PubMed WorldCat 37. Araújo T , et al. Qualitative studies of insomnia: current state of knowledge in the field . Sleep Med Rev. 2017 ; 31 : 58 – 69 . Google Scholar Crossref Search ADS PubMed WorldCat 38. Faestel PM , et al. Perceived insufficient rest or sleep among veterans: behavioral Risk Factor Surveillance System 2009 . J Clin Sleep Med. 2013 ; 9 ( 6 ): 577 – 584 . Google Scholar Crossref Search ADS PubMed WorldCat 39. Cepeda MS , et al. Clinical relevance of sleep duration: results from a cross-sectional analysis using NHANES . J Clin Sleep Med. 2016 ; 12 ( 6 ): 813 – 819 . Google Scholar Crossref Search ADS PubMed WorldCat 40. Ulmer CS , et al. A comparison of sleep difficulties among Iraq/Afghanistan theater veterans with and without mental health diagnoses . J Clin Sleep Med. 2015 ; 11 ( 9 ): 995 – 1005 . Google Scholar Crossref Search ADS PubMed WorldCat 41. Plumb TR , et al. Sleep disturbance is common among servicemembers and veterans of operations enduring Freedom and Iraqi Freedom . Psychol Serv. 2014 ; 11 ( 2 ): 209 – 219 . Google Scholar Crossref Search ADS PubMed WorldCat 42. Martin JL , et al. Estimated prevalence of insomnia among women veterans: results of a postal survey . Womens Health Issues. 2017 ; 27 ( 3 ): 366 – 373 . Google Scholar Crossref Search ADS PubMed WorldCat 43. Morin CM , et al. The insomnia severity index: psychometric indicators to detect insomnia cases and evaluate treatment response . Sleep. 2011 ; 34 ( 5 ): 601 – 608 . Google Scholar Crossref Search ADS PubMed WorldCat 44. Pittman JOE , et al. VA eScreening program: technology to improve care for post-9/11 veterans . Psychol Serv. 2017 ; 14 ( 1 ): 23 – 33 . Google Scholar Crossref Search ADS PubMed WorldCat 45. Bastien CH , et al. Validation of the insomnia severity index as an outcome measure for insomnia research . Sleep Med. 2001 ; 2 ( 4 ): 297 – 307 . Google Scholar Crossref Search ADS PubMed WorldCat 46. Weathers FW , et al. The PTSD Checklist (PCL): Reliability, validity, and diagnostic utility . Paper presented at Annual Convention of the International Society for Traumatic Stress Studies ; 1993 ; San Antonio, TX . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC 47. Weathers FW , et al. The PTSD Checklist for DSM-5 (PCL-5) . 2013 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC 48. Kroenke K , et al. The PHQ-9: validity of a brief depression severity measure . J Gen Intern Med. 2001 ; 16 ( 9 ): 606 – 613 . Google Scholar Crossref Search ADS PubMed WorldCat 49. Saunders JB , et al. Development of the alcohol use disorders identification test (AUDIT): WHO collaborative project on early detection of persons with harmful alcohol consumption‐II . Addiction . 1993 ; 88 ( 6 ): 791 – 804 . Google Scholar Crossref Search ADS PubMed WorldCat 50. Meneses-Gaya C , et al. The fast alcohol screening test (FAST) is as good as the AUDIT to screen alcohol use disorders . Subst Use Misuse. 2010 ; 45 ( 10 ): 1542 – 1557 . Google Scholar Crossref Search ADS PubMed WorldCat 51. Reinert DF , et al. The alcohol use disorders identification test: an update of research findings . Alcohol Clin Exp Res. 2007 ; 31 ( 2 ): 185 – 199 . Google Scholar Crossref Search ADS PubMed WorldCat 52. Donnelly KT , et al. Reliability, sensitivity, and specificity of the VA traumatic brain injury screening tool . J Head Trauma Rehabil. 2011 ; 26 ( 6 ): 439 – 453 . Google Scholar Crossref Search ADS PubMed WorldCat 53. Kimerling R , et al. The Veterans Health Administration and military sexual trauma . Am J Public Health. 2007 ; 97 ( 12 ): 2160 – 2166 . Google Scholar Crossref Search ADS PubMed WorldCat 54. Haskell SG , et al. Gender differences in rates of depression, PTSD, pain, obesity, and military sexual trauma among Connecticut War Veterans of Iraq and Afghanistan . J Womens Health (Larchmt). 2010 ; 19 ( 2 ): 267 – 271 . Google Scholar Crossref Search ADS PubMed WorldCat 55. Cohen J. Statistical Power for the Social Sciences . Hillsdale, NJ : Laurence Erlbaum and Associates ; 1988 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC 56. Reynolds G , Shendruk A. Demographics of the U.S. Military . 2018, 2019 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC 57. Armitage R . Sleep and circadian rhythms in mood disorders . Acta Psychiatr. Scand . 2007 ; 115 : 104 – 115 . Google Scholar Crossref Search ADS WorldCat 58. Manber R , et al. Insomnia and depression: a multifaceted interplay . Curr Psychiatry Rep. 2009 ; 11 ( 6 ): 437 – 442 . Google Scholar Crossref Search ADS PubMed WorldCat 59. Taylor DJ , et al. Insomnia as a health risk factor . Behav Sleep Med. 2003 ; 1 ( 4 ): 227 – 247 . Google Scholar Crossref Search ADS PubMed WorldCat 60. Sjöström N , et al. Nightmares and sleep disturbances in relation to suicidality in suicide attempters . Sleep. 2007 ; 30 ( 1 ): 91 – 95 . Google Scholar Crossref Search ADS PubMed WorldCat 61. VA/DOD . Clinical Practice Guideline for the Management of Chronic Insomnia Disorder and Obstructive Sleep Apnea . 2019 . https://www.healthquality.va.gov/guidelines/CD/insomnia/VADoDSleepCPGFinal508.pdf. 62. Ouellet MC , et al. Insomnia following traumatic brain injury: a review . Neurorehabil Neural Repair. 2004 ; 18 ( 4 ): 187 – 198 . Google Scholar Crossref Search ADS PubMed WorldCat 63. Mahmood O , et al. Neuropsychological performance and sleep disturbance following traumatic brain injury . J Head Trauma Rehabil. 2004 ; 19 ( 5 ): 378 – 390 . Google Scholar Crossref Search ADS PubMed WorldCat 64. Trockel M , et al. Effects of cognitive behavioral therapy for insomnia on suicidal ideation in veterans . Sleep. 2015 ; 38 ( 2 ): 259 – 265 . Google Scholar Crossref Search ADS PubMed WorldCat 65. Manber R , et al. Dissemination of CBTI to the non-sleep specialist: protocol development and training issues . J Clin Sleep Med. 2012 ; 8 ( 2 ): 209 – 218 . Google Scholar Crossref Search ADS PubMed WorldCat 66. Qaseem A , et al. Management of chronic insomnia disorder in adults: a clinical practice guideline from the American College of Physicians . Ann Intern Med. 2016 ; 165 ( 2 ): 125 – 133 . Google Scholar Crossref Search ADS PubMed WorldCat 67. Sateia MJ , et al. Clinical practice guideline for the pharmacologic treatment of chronic insomnia in adults: an American Academy of Sleep Medicine Clinical Practice Guideline . J Clin Sleep Med. 2017 ; 13 ( 2 ): 307 – 349 . Google Scholar Crossref Search ADS PubMed WorldCat 68. Morin CM , et al. Nonpharmacologic treatment of chronic insomnia. An American Academy of Sleep Medicine review . Sleep. 1999 ; 22 ( 8 ): 1134 – 1156 . Google Scholar Crossref Search ADS PubMed WorldCat 69. Koffel E , et al. Increasing access to and utilization of cognitive behavioral therapy for insomnia (CBT-I): a narrative review . J Gen Intern Med. 2018 ; 33 ( 6 ): 955 – 962 . Google Scholar Crossref Search ADS PubMed WorldCat 70. Colvonen PJ , et al. Piloting cognitive behavioral therapy for insomnia integrated with prolonged exposure . Psychol Trauma . 2019 ; 11 ( 1 ): 107 – 113 . Google Scholar Crossref Search ADS PubMed WorldCat 71. Colvonen PJ , et al. Obstructive sleep apnea and posttraumatic stress disorder among OEF/OIF/OND veterans . J Clin Sleep Med. 2015 ; 11 ( 5 ): 513 – 518 . Google Scholar Crossref Search ADS PubMed WorldCat Published by Oxford University Press on behalf of Sleep Research Society (SRS) 2020. This work is written by (a) US Government employee(s) and is in the public domain in the US. Published by Oxford University Press on behalf of Sleep Research Society (SRS) 2020.

Vallat,, Raphael;Nicolas,, Alain;Ruby,, Perrine

doi: 10.1093/sleep/zsaa116pmid: 32597973

Abstract Why do some individuals recall dreams every day while others hardly ever recall one? We hypothesized that sleep inertia—the transient period following awakening associated with brain and cognitive alterations—could be a key mechanism to explain interindividual differences in dream recall at awakening. To test this hypothesis, we measured the brain functional connectivity (combined electroencephalography–functional magnetic resonance imaging) and cognition (memory and mental calculation) of high dream recallers (HR, n = 20) and low dream recallers (LR, n = 18) in the minutes following awakening from an early-afternoon nap. Resting-state scans were acquired just after or before a 2 min mental calculation task, before the nap, 5 min after awakening from the nap, and 25 min after awakening. A comic was presented to the participants before the nap with no explicit instructions to memorize it. Dream(s) and comic recall were collected after the first post-awakening scan. As expected, between-group contrasts of the functional connectivity at 5 min post-awakening revealed a pattern of enhanced connectivity in HR within the default mode network (DMN) and between regions of the DMN and regions involved in memory processes. At the behavioral level, a between-group difference was observed in dream recall, but not comic recall. Our results provide the first evidence that brain functional connectivity right after awakening is associated with interindividual trait differences in dream recall and suggest that the brain connectivity of HR at awakening facilitates the maintenance of the short-term memory of the dream during the sleep–wake transition. dream recall, dreaming, sleep inertia, awakening, EEG–fMRI, functional connectivity, default mode network Statement of Significance Why do some individuals recall their dreams every day while others hardly ever recall one? In this article, we present evidence that the answer to this question lies in part in the functional state of the brain immediately upon awakening from sleep. Using a combined electroencephalography–functional magnetic resonance imaging design, we compared the brain functional connectivity of individuals with a high and low dream recall frequency (HR and LR, respectively) before and after awakening from an early afternoon nap. Our results show a pattern of enhanced connectivity in HR within the default mode network (DMN) and between the DMN and regions associated with memory processes, therefore suggesting that trait differences in dream recall are associated with a specific brain functioning during the sleep–wake transition. Introduction Dreaming is a universal and still intriguing experience that one may (or may not) recall upon awakening from sleep. Several aspects of dreaming are still poorly understood, one of which is the interindividual differences in dream recall frequency (DRF), that is, why some individuals recall dreams every morning while some hardly ever recall one. Previous research identified several state and trait factors that participate in intra- and interindividual variability in dream recall, respectively [1, 2]. The main state factors identified are the sleep stage and the amount of sleep preceding awakening [3], circadian and ultradian rhythms [4], psychotropic drug intake [5], and the method of awakening [6]. Regarding trait factors, some personality dimensions such as creativity and openness to experience as well as anxiety and interest in dreams have been consistently correlated with interindividual variability in DRF [7]. Intra-sleep awakening—which is both a state and a trait factor—was also hypothesized to promote dream recall [7, 8], a hypothesis that was recently confirmed using objective sleep measurements. In two studies [9, 10], we used full-night polysomnography to compare the sleep architecture and microstructure of high dream recallers (HR, ≥5 dream recalls per week) and low dream recallers (LR, ≤2 dream recalls per month). Results showed more intra-sleep wakefulness in HR than in LR (with no differences in other sleep parameters), which is coherent with the idea that periods of wakefulness during sleep are necessary to encode dream content into long-term memory [8]. Remarkably, LR still recall significantly fewer dreams than HR regardless of the sleep stage in which they are awakened, even when the number of awakenings is increased and experimentally controlled during a night in the lab [9, 11–13], showing that state factors cannot completely compensate for trait factors. We propose here that the functional state of the brain in the first minutes after awakening from sleep could explain, at least in part, the trait variability in DRF. The wake period immediately following awakening from sleep is indeed marked by sleep inertia, a transient state of impaired vigilance and cognition that usually dissipates within the first 30 min after awakening [14–16]. Because sleep inertia can largely vary both within and between individuals, it represents a good candidate to explain both state and trait variability in dream recall [7, 17, 18]. Previous observations argue in favor of this hypothesis. For instance, Morel and colleagues [18] reported that “the findings do not support the hypothesis linking increased recall ability to increases in cortical activation prior to awakening. However, the recall groups depicted a different pattern of arousal in their transition from sleep to wakefulness.” Similarly, Rochlen et al. [19] noted that “healthy controls showed a large shift in delta amplitude in the sleep–wake transition during successful recall but not during recall failure.” According to this hypothesis, one could expect that LR would show more sleep inertia (as measured by cognitive and brain functional connectivity disturbances [16]) after awakening than HR. This hypothesis is supported by previous results showing trait brain differences between HR and LR during both wakefulness and sleep, either while performing a task, or at rest (task: auditory oddball during wake and sleep, method: electroencephalography (EEG) [9]; task: rest during wake and sleep, method: positron emission tomography [PET] [13]). Noteworthy, we found that HR have an increased regional cerebral blood flow in the medial prefrontal cortex (MPFC) and temporoparietal junction (TPJ) at rest (during wake and sleep), as well as an increased white matter density in the MPFC [20]. The MPFC and TPJ are known to be involved in dream production and/or recall. Indeed, lesions of one or both regions induce cessation of dream reports [21, 22], and therefore, the proper functioning of these regions upon awakening could mediate the retrieval of dream content upon awakening. Specifically, the proper functional connectivity of the network that encompasses these two regions, the so-called default mode network (DMN), could be a necessary condition for successful dream recall upon awakening. Indeed, the DMN is a set of functionally coupled brain areas [23, 24] involved in mind-wandering and episodic memory retrieval [25], and several authors have postulated that the DMN might be the neural correlate of dreaming [26, 27]. Coherent with this, it is noteworthy that the DMN is typically hyperconnected during rapid eye movement (REM) sleep, the sleep stage associated with the highest rate of successful dream recall, and hypoconnected during N3 sleep, the sleep stage associated with the lowest rate of successful dream recall [28–31]. However, it is still unknown if the DMN is implicated in the production of dream experience while asleep, the recall of dream content while awake, or both. The present study aimed therefore at comparing the brain and cognitive functioning of HR and LR during the sleep inertia period following awakening from sleep. This has surprisingly never been experimentally tested before, despite several studies mentioning the brain and cognitive functioning during sleep inertia as a potential explaining factor of DRF variability [7, 32]. We designed a functional magnetic resonance imaging (fMRI) study to compare the brain functional connectivity of HR and LR in the minutes following awakening from a 45 min early afternoon nap. Resting-state scans were acquired before the nap, 5 min and 25 min after awakening to investigate the dynamics of brain functional reorganization during the first half hour following awakening, and each scan was associated with a mental calculation task to measure the cognitive impairment associated with sleep inertia. We predicted that HR would show (1) more dream recall following awakening from sleep, (2) higher functional connectivity within the DMN and between the DMN and regions involved in memory retrieval, and (3) less cognitive performance impairments, suggesting a faster recovery of normal cognitive functioning upon awakening. Methods This study is a reanalysis of previously published data [16] to specifically investigate group differences between HR and LR participants. Participants Behavioral and neurophysiological data were acquired from 55 healthy participants (28 males, mean ± SD age = 22.55 ± 2.41 years, range = 19–29 years) having normal sleep characteristics and BMI (habitual sleep time per night, 7.7 ± 0.9 h; BMI, 22.1 ± 2.6 kg/m2). The subjects were informed of the study through an announcement sent to the mailing list of Lyon University, which briefly described the study and included a link to a questionnaire concerning sleep and dream habits [33]. Subjects were selected if they reported and subsequently confirmed during a phone interview: (1) having a high or low DRF, that is, a DRF superior to 5 dream recalls per week for HR and inferior to 2 dream recalls per month for LR, as well as having (2) a regular sleep-wake schedule, no difficulty to fall asleep, being occasional or frequent nappers and having preferentially already done an MRI brain scan in the past few years (we assumed that participants who are already familiar to the MR environment might be less anxious and therefore more prone to fall asleep). Among the 55 participants, 28 of them were HR (mean DRF = 6.6 ± 0.7 dream reports per week) and 27 were LR (mean DRF = 0.2 ± 0.1 dream report per week). Apart from the DRF (p < 0.001), the two groups did not differ in other assessed parameters, including age, BMI, habitual sleep duration, or education level (all p’s ≥ 0.25). They had no history of neurological and psychiatric disorders and had no sleep disturbances. They provided written informed consent according to the Declaration of Helsinki and received monetary compensation for their participation. The study was approved by the local ethics committee (CCPPRB, Centre Leon Berard, Lyon, France). Procedure The paradigm and procedure are presented in Figure 1. Figure 1. Open in new tabDownload slide Experimental design. Figure 1. Open in new tabDownload slide Experimental design. Evening and night To facilitate sleep in the MRI environment and maximize sleep inertia upon awakening from the nap, participants underwent a partial sleep deprivation on the night before the experiment (3 h of sleep allowed). They arrived in the sleep unit of the hospital Le Vinatier (Lyon, France) at 8:00 pm. From 8:00 pm to 10:00 pm, they underwent several personality and cognitive tests administered by RV (results will be presented elsewhere). They were then instructed to stay awake until 5:00 am (the possible activities were reading, making puzzles, and watching movies), at which point they could sleep for 3 h until 8:00 am in a bed in the sleep unit. Energy drinks or physical activity were prohibited during the partial sleep deprivation, and nurses regularly checked that the subject did not fall asleep. Activity was monitored via wrist actigraphy (Actigraph, Pensacola, USA) during the whole night. At 8:00 am, participants were awakened by the experimenter, and immediately asked to report their dreams, if any. They were then offered breakfast and a shower and occupied themselves (reading or internet) under the experimenters’ supervision until the MRI session. Day After lunch at 11:30 am, participants were conducted to the neuroimaging center (CERMEP). During the first half hour, experimenters installed on the participant’s head an MRI compatible EEG cap (EASYCAP). Participants were then installed in the MRI scanner at about 1:20 pm (1:17 pm ± 13 min). They read a 5 min comic strip during the calibration of the eye-tracking camera. The subjects were led to believe that the only purpose of the comic strip was to calibrate the eye-tracking system, and therefore were not given any explicit instructions to memorize it. Next, they performed the descending subtraction task (DST) for 2 min and the first resting-state scan was subsequently acquired, with the instructions to remain awake and look at a central fixation cross on the screen. At the end of the scan, participants were informed that they could close their eyes and sleep (at 1:39 pm ± 14 min on average) during the next 45 min. At the end of the nap slot, participants were awakened, if they were sleeping, by calling their first name and the second resting-state scan was acquired as close to the awakening as possible. At the end of the scan, the second DST was performed. During the following 10 min, subjects were asked to recall the dream(s) they had during the nap if any, to comment on the quality of their sleep in the scanner and to recall the comic that they had read during the calibration of the eye-tracking camera. The third resting-state scan and DST were finally performed about 25 min after awakening, and were immediately followed by an 8-min T1 anatomical scan. Behavioral tests In the evening at the sleep unit, participants trained on the DST (six blocks of 2 min each), which was used to evaluate cognitive performances during the MRI session on the following day, to avoid a practice effect over the first trials [34]. The DST has been previously used to evidence performance deficit and normalization in the first 30 min post awakening [34–36]. Subjects were presented with a three-digit number. They were instructed to subtract 9, saying the operation and the result aloud, and then continue by subtracting 8 from the remainder, then 7, and so on until they had to subtract 1. At this point, they were to start the cycle of descending subtractions again. Participants were asked to do the task for two minutes and were instructed to be as fast and accurate as possible. In the scanner, between the second and the third scan, participants were asked to recall and describe any dream(s) they had while napping. Second, they were asked about their sleep in the scanner. Third, they were asked to recall the comic strip that they read just before the nap slot. They were first told to recall the story freely and then answered pre-determined questions about the content of the story. Data collection EEG and eye movement recordings Polysomnography data were recorded using a 15 channels MR-compatible cap (EasyCap, Brain Products GmbH, Gilching, Germany) designed for sleep studies, with a layout designed according to the American Academy of Sleep Medicine Guidelines 2007. It comprised nine EEG electrodes placed according to the international standard 10/20 system (O1, O2, C3, C4, F3, F4, M1, M2, Cz, FCz was used as reference and AFz as ground), two EOG electrodes, three EMG electrodes, and an electrocardiogram electrode placed on the back of the participant. The sampling rate was 5,000 Hz and an analog band-pass filter (0.016–250 Hz) was applied before data digitalization to prevent saturation and reduce the gradient artifact amplitude [37]. The EEG was synchronized with the MR scanner’s clock using BrainProducts’ SyncBox. A real-time pulse-artifact correction was performed using the BrainVision Recorder (Version 1.2) and BrainVision RecView (Version 1.4) software (Brain Products) to allow for online sleep scoring during the fMRI session. To ensure that the participants were not closing their eyes during the resting state scans, eye movements were monitored during the experiment using an EyeLink 1000 fMRI eye-tracking system (SR Research Ontario, Canada). The position of eyes was calibrated at the beginning of the experiment and monitored throughout the whole MRI session. MRI acquisition MRI scans were obtained from a MAGNETOM Prisma 3.0 T scanner (Siemens Healthcare, Erlangen, Germany) at the Primage neuroimaging platform (CERMEP). Structural MRI was acquired with a T1-weighted (0.9-mm isotropic resolution) MPRAGE sequence and functional MRI data with a T2*-weighted 2D gradient echo-planar imaging sequence with 180 volumes (TR/TE: 2,000/25 ms; flip angle: 80°; voxel size: 2.68 × 2.68 × 3 mm; slices: 40, duration: 6 min). Functional and anatomical scans were performed using a 20-channel head coil. The coil was foam-padded to improve subject comfort and restrict head motion. Data analysis Electroencephalography Artifacts related to gradient switching and cardiac pulse (cardio-ballistic artifact) were removed using standard routines available in BrainVision Analyzer version 2.0 software (Brain Products). Polysomnographic data were downsampled to 500 Hz and band-pass filtered between 0.1 and 40 Hz. Offline sleep stage scoring was done in epochs of 30 s following standard AASM rules [38, 39] by the first author (R.V.) using the SLEEP software [40]. Functional magnetic resonance imaging Preprocessing and quality check were performed using a standard routine in SPM12 software (Wellcome Department of Imaging Neuroscience) and the CONN toolbox, version 17f (http://www.conn-toolbox.org) [41]. Preprocessing included functional realignment, slice-time correction, coregistration to structural scan, spatial normalization, and spatial smoothing using a 6 mm full-width at half-maximum isotropic Gaussian kernel filter. No field map correction was applied. Individual T1 images were segmented into gray matter, white matter, and cerebrospinal fluid tissue maps. Functional and structural images were then normalized to MNI152 space (Montreal Neurological Institute). Functional images underwent artifact and motion regression in the Artifact Detection Toolbox (https://www.nitrc.org/projects/artifact_detect/) using the following criteria to define outliers: global signal intensity changes greater than 9 standard deviations and movement exceeding 2 mm. SPM motions parameters and outliers were subsequently included as covariates in connectivity analyses. Connectivity analyses were performed using the CONN toolbox version 17f. First, we performed a denoising step including a regression of the six motion correction parameters and their corresponding first-order temporal derivatives, as well as a component-based strategy (aCompCor [42] to identify and remove physiological confounds that are unlikely to be related to neural activity. The resulting BOLD time series were band-pass filtered (0.008–0.09 Hz) to further reduce noise and increase sensitivity [43]. Having done so, we performed seed-based analyses on the core regions of the DMN. As per our hypotheses, we focused on the default mode since it has been suggested to be involved in dream recall and/or production [13, 20–22]. Specifically, we used the MPFC (center of mass in MNI coordinates = 1, 55, −3, 1,346 voxels in 2 mm space), the posterior cingulate cortex (PCC, 1, −61, 38, 4,833 voxels) and the lateral parietal cortices (LP, right = 47, −67, 29, 1,326 voxels; left = −39, −77, 33, 1,041 voxels). Since the LPs are bilateral, functional connectivity contrasts were estimated using the main effect of both hemispheres. These anatomical regions of interest are included in the CONN toolbox default atlas and were extracted using an ICA approach on 497 subjects from the Human Connectome Project. Statistics For the DST, between-group comparisons were achieved using a two-way mixed ANOVA with a group factor (between-subject factor with two levels: HR and LR) and a time factor (within-subject factor with three levels: Pre-sleep, 5 min p-a, 25 min p-a). Post hoc two-sided t-tests were used in case of significance. All statistical tests were performed using the Pingouin package [44] for Python (https://pingouin-stats.org/). Seed-based functional connectivity analyses were performed using a cluster-defining voxel-wise height threshold of p < 0.01 (uncorrected, two-sided) and a whole-brain family-wise error (FWE) corrected extent threshold of p < 0.05. Results Sleep parameters As expected and due to the inherent discomfort of the MRI environment, several participants were not able to reach and maintain N2 and N3 sleep during the 45 min nap slot. Note that here we used a more liberal threshold than in our previous article [16] to determine whether to include or not a participant in further analysis. This decision was mainly motivated by considerations of statistical power and sample size in our HR versus LR comparison. Specifically, in our previous article, inclusion of the participants in either one of the N2 or N3 groups was based on the two following rules: (1) presence of N2/3 sleep during the nap, (2) awakening in N2/3 sleep. In the current article, inclusion of the participants in subsequent analyses was based on the three following rules: (1) presence of N2/3 sleep during the nap, (2) awakening in N1, N2, or N3 sleep, (3) more than 80% of epochs in the 10 min prior to awakening by experimenter are scored as sleep (N1 included). One subject out of the 39 remaining was discarded because of a technical failure during data acquisition, leading thus to a total of 38 participants included in the final analysis (20 HR and 18 LR). The two groups differed in DRF (HR = 6.5 ± 0.7 dream reports per week, LR = 0 ± 0 dream report per week, p < 0.001), but did not differ for all the other variables including age (HR = 22.5 ± 2.5 years, LR = 22.7 ± 2.4 years, p = 0.86), BMI (HR = 22.9 ± 2.6 kg/m2, LR = 21.8 ± 2.2 kg/m2, p = 0.15), education (HR = 3.8 ± 2.5 years of higher education, LR = 3.6 ± 1.3 years of higher education, p = 0.83), habitual sleep duration (HR = 7.7 ± 0.9 h, LR = 7.6 ± 1.0 h, p = 0.73), and sex ratio (HR = 0.81, LR = 1, p = 0.99). Means of the main sleep parameters during the nap in the two groups are presented in Table 1. Importantly, there was no significant group difference for any of the sleep parameters considered or in the latency between the awakening and the two post-awakening resting-state scans. Table 1. Mean sleep parameters of the HR (n = 20) and LR (n = 18) groups Sleep parameters . HR . LR . P-value . TST (min) 36.4 ± 7.5 38.1 ± 4.6 0.42 SE (%) 87.7 ± 7.8 87.3 ± 7.7 0.89 Wake (min) 8.4 ± 3.3 9.8 ± 4.2 0.32 N1 (min) 13 ± 7.9 10.3 ± 6.4 0.28 N2 (min) 17.2 ± 5.3 20.4 ± 6.4 0.11 N3 (min) 6.6 ± 6.3 7.7 ± 5.7 0.59 REM (min) 0 ± 0 0 ± 0 – Longest N2 (min) 12.1 ± 5.1 13.6 ± 4.4 0.34 Longest N3 (min) 6.2 ± 6.1 7.3 ± 5.3 0.57 Stage prior to awakening 2.3 ± 0.7 2.6 ± 0.6 0.25 LAS1 (min) 3.4 ± 1 5.2 ± 4.1 0.10 LAS2 (min) 24.4 ± 4.1 24 ± 3.8 0.73 Sleep parameters . HR . LR . P-value . TST (min) 36.4 ± 7.5 38.1 ± 4.6 0.42 SE (%) 87.7 ± 7.8 87.3 ± 7.7 0.89 Wake (min) 8.4 ± 3.3 9.8 ± 4.2 0.32 N1 (min) 13 ± 7.9 10.3 ± 6.4 0.28 N2 (min) 17.2 ± 5.3 20.4 ± 6.4 0.11 N3 (min) 6.6 ± 6.3 7.7 ± 5.7 0.59 REM (min) 0 ± 0 0 ± 0 – Longest N2 (min) 12.1 ± 5.1 13.6 ± 4.4 0.34 Longest N3 (min) 6.2 ± 6.1 7.3 ± 5.3 0.57 Stage prior to awakening 2.3 ± 0.7 2.6 ± 0.6 0.25 LAS1 (min) 3.4 ± 1 5.2 ± 4.1 0.10 LAS2 (min) 24.4 ± 4.1 24 ± 3.8 0.73 TST, total sleep time; SE, sleep efficiency in percentage; Wake (W), N1, N2, and N3, total duration of each sleep stage in minutes. Stage of awakening, sleep stage prior to awakening, where Wake was encoded as 0, N1 as 1, N2 as 2, and N3 as 3. Therefore, higher value represents higher sleep depth prior to awakening. Longest N2, longest period of uninterrupted N2 sleep, in minutes. Longest N3, longest period of uninterrupted N3-sleep, in minutes. LAS1, latency between the awakening and the start of the first post-awakening resting-state scan, in minutes. LAS2, latency between the awakening and the start of the second post-awakening resting-state scan, in minutes. p-Values were obtained using two-sided independent t-tests. Open in new tab Table 1. Mean sleep parameters of the HR (n = 20) and LR (n = 18) groups Sleep parameters . HR . LR . P-value . TST (min) 36.4 ± 7.5 38.1 ± 4.6 0.42 SE (%) 87.7 ± 7.8 87.3 ± 7.7 0.89 Wake (min) 8.4 ± 3.3 9.8 ± 4.2 0.32 N1 (min) 13 ± 7.9 10.3 ± 6.4 0.28 N2 (min) 17.2 ± 5.3 20.4 ± 6.4 0.11 N3 (min) 6.6 ± 6.3 7.7 ± 5.7 0.59 REM (min) 0 ± 0 0 ± 0 – Longest N2 (min) 12.1 ± 5.1 13.6 ± 4.4 0.34 Longest N3 (min) 6.2 ± 6.1 7.3 ± 5.3 0.57 Stage prior to awakening 2.3 ± 0.7 2.6 ± 0.6 0.25 LAS1 (min) 3.4 ± 1 5.2 ± 4.1 0.10 LAS2 (min) 24.4 ± 4.1 24 ± 3.8 0.73 Sleep parameters . HR . LR . P-value . TST (min) 36.4 ± 7.5 38.1 ± 4.6 0.42 SE (%) 87.7 ± 7.8 87.3 ± 7.7 0.89 Wake (min) 8.4 ± 3.3 9.8 ± 4.2 0.32 N1 (min) 13 ± 7.9 10.3 ± 6.4 0.28 N2 (min) 17.2 ± 5.3 20.4 ± 6.4 0.11 N3 (min) 6.6 ± 6.3 7.7 ± 5.7 0.59 REM (min) 0 ± 0 0 ± 0 – Longest N2 (min) 12.1 ± 5.1 13.6 ± 4.4 0.34 Longest N3 (min) 6.2 ± 6.1 7.3 ± 5.3 0.57 Stage prior to awakening 2.3 ± 0.7 2.6 ± 0.6 0.25 LAS1 (min) 3.4 ± 1 5.2 ± 4.1 0.10 LAS2 (min) 24.4 ± 4.1 24 ± 3.8 0.73 TST, total sleep time; SE, sleep efficiency in percentage; Wake (W), N1, N2, and N3, total duration of each sleep stage in minutes. Stage of awakening, sleep stage prior to awakening, where Wake was encoded as 0, N1 as 1, N2 as 2, and N3 as 3. Therefore, higher value represents higher sleep depth prior to awakening. Longest N2, longest period of uninterrupted N2 sleep, in minutes. Longest N3, longest period of uninterrupted N3-sleep, in minutes. LAS1, latency between the awakening and the start of the first post-awakening resting-state scan, in minutes. LAS2, latency between the awakening and the start of the second post-awakening resting-state scan, in minutes. p-Values were obtained using two-sided independent t-tests. Open in new tab Behavioral results Descending subtraction task DST performances are reported in Figure 2. A two-way ANOVA revealed a significant effect of time in the number of responses (F(2, 72) = 6.04, p = 0.004) and percentage of correct responses (F(2, 72) = 4.29, p = 0.017). Specifically, the total number of responses and the percentage of correct responses were reduced at 5 min p-a compared to pre-nap and 25 min p-a (all p’s < 0.05). There was no main effect of time in the percentage of mistakes, a finding in line with the generally held view that speed is more impaired than accuracy during sleep inertia [15]. Contrary to our hypothesis, there was no significant main effect of group or interaction between group and time for the DST performance. Figure 2. Open in new tabDownload slide Group performances at the DST. Red lines, HR (n = 20), black lines, LR (n = 18). Left, total number of responses (index of speed). Middle, percentage of mistakes (marker of accuracy). Right, percentage of correct responses relative to pre-nap performances (marker of both speed and accuracy). Error bars represent the 95% confidence intervals. A significant main effect of time was found for the number of responses and the percentage of correct responses, but no significant group differences or interactions between group and time were observed for any of the three outcomes. Figure 2. Open in new tabDownload slide Group performances at the DST. Red lines, HR (n = 20), black lines, LR (n = 18). Left, total number of responses (index of speed). Middle, percentage of mistakes (marker of accuracy). Right, percentage of correct responses relative to pre-nap performances (marker of both speed and accuracy). Error bars represent the 95% confidence intervals. A significant main effect of time was found for the number of responses and the percentage of correct responses, but no significant group differences or interactions between group and time were observed for any of the three outcomes. Dream recall After awakening from the partial sleep deprivation in the sleep unit, significantly more HR reported dreams than did LR (65% of HR and 22% of LR reported a full or a white [contentless] dream; X2 = 7.0, p = 0.008). A similar effect was observed after awakening from the 45 min nap inside the MRI (75% of HR vs. 33% of LR, X2 = 6.7, p = 0.010; see Table 2). The between-group difference in dream recall after the nap was significant for participants that were awakened in N2 sleep (75% of HR and 17% of LR, X2 = 4.6, p = 0.03), but not for participants that were awakened in N3 sleep (Table 2). Lastly, when combining participants from both groups, we did not find a significant between-stage difference in dream recall after awakening from the nap. That is, dream recall was not significantly different after awakening from N2 or N3 sleep (X2 = 0.08, p = 0.78). Table 2. Dream recall after the nap in the MR scanner Stage prior to awakening . HR . LR . All . P-value . N1 sleep 3/3 (100%) 0/1 (0%) 3/4 (75%) – N2 sleep 6/8 (75%) 1/6 (17%) 7/14 (50%) 0.031 N3 sleep 6/9 (67%) 5/11 (45%) 11/20 (55%) 0.343 All 15/20 (75%) 6/18 (33%) 21/38 (55%) 0.010 Stage prior to awakening . HR . LR . All . P-value . N1 sleep 3/3 (100%) 0/1 (0%) 3/4 (75%) – N2 sleep 6/8 (75%) 1/6 (17%) 7/14 (50%) 0.031 N3 sleep 6/9 (67%) 5/11 (45%) 11/20 (55%) 0.343 All 15/20 (75%) 6/18 (33%) 21/38 (55%) 0.010 Count and percentage of successful dream recall in HR and LR after awakening from the nap in the MR scanner, stratified by the sleep stage prior to awakening. p-values were calculated using a Pearson chi-squared test, with the exception of N1 sleep for which the observed frequencies were too small (< 5) to estimate reliable p-values. Open in new tab Table 2. Dream recall after the nap in the MR scanner Stage prior to awakening . HR . LR . All . P-value . N1 sleep 3/3 (100%) 0/1 (0%) 3/4 (75%) – N2 sleep 6/8 (75%) 1/6 (17%) 7/14 (50%) 0.031 N3 sleep 6/9 (67%) 5/11 (45%) 11/20 (55%) 0.343 All 15/20 (75%) 6/18 (33%) 21/38 (55%) 0.010 Stage prior to awakening . HR . LR . All . P-value . N1 sleep 3/3 (100%) 0/1 (0%) 3/4 (75%) – N2 sleep 6/8 (75%) 1/6 (17%) 7/14 (50%) 0.031 N3 sleep 6/9 (67%) 5/11 (45%) 11/20 (55%) 0.343 All 15/20 (75%) 6/18 (33%) 21/38 (55%) 0.010 Count and percentage of successful dream recall in HR and LR after awakening from the nap in the MR scanner, stratified by the sleep stage prior to awakening. p-values were calculated using a Pearson chi-squared test, with the exception of N1 sleep for which the observed frequencies were too small (< 5) to estimate reliable p-values. Open in new tab Memory recall For the recall of the comic, no significant group difference was found in the final memory score to the questionnaire about the story content (on a scale from 0 to 10: HR = 5.6 ± 2.0; LR = 4.7 ± 2.7; p = 0.26). However, there was a tendency for a significantly higher number of words during free recall in HR than LR (HR = 223.8 ± 113 words; LR = 156.1 ± 98 words; p = 0.056). Functional connectivity We found significant interaction effects between group and time for all three seeds, which are reported in Supplementary Table S1. Planned contrasts evaluating group differences at each of the three resting-state scans are presented in Table 3 and Figure 3, and summarized in the paragraph below. Importantly, there was no group difference in motion parameters at any of the three resting-state scans, as indexed by the framewise displacement [45] and the number of scans that were defined as outliers by the Artifact Detection Toolbox (all p’s > 0.4). Table 3. Seed-based functional connectivity results for the HR > LR contrast . . . MNI coordinates . . . Seed . Scan . Brain region . X . Y . Z . Peak T-value . Cluster size . MPFC Pre-nap N.S. 5 min Precuneus (L) −4 −76 52 4.75 487 25 min Putamen (R) 14 14 −8 5.15 361 PCC Pre-nap N.S. 5 min PFC (R) 36 56 14 4.22 505 SFG (R) 20 28 46 3.90 378 25 min N.S. LP Pre-nap N.S. 5 min SFG (R) 22 26 48 4.26 372 AG (R) 36 −54 42 4.05 309 25 min N.S. . . . MNI coordinates . . . Seed . Scan . Brain region . X . Y . Z . Peak T-value . Cluster size . MPFC Pre-nap N.S. 5 min Precuneus (L) −4 −76 52 4.75 487 25 min Putamen (R) 14 14 −8 5.15 361 PCC Pre-nap N.S. 5 min PFC (R) 36 56 14 4.22 505 SFG (R) 20 28 46 3.90 378 25 min N.S. LP Pre-nap N.S. 5 min SFG (R) 22 26 48 4.26 372 AG (R) 36 −54 42 4.05 309 25 min N.S. The seed regions are the MPFC (center of mass in MNI coordinates = 1, 55, −3), the PCC (1, −61, 38), and the lateral parietal cortices (LP, right = 47, −67, 29, left = −39, −77,33; the functional connectivity contrast was done using the main effect of both hemispheres). These three seeds are part of the DMN, a network that has been suggested to be involved in dream recall and/or production. The functional connectivity between the seed and the regions reported was higher in HR than LR (see also Figure 3). Statistical analyses were performed using a cluster-defining voxel-wise height threshold of p < 0.01 (uncorrected, two-sided) and a whole-brain FWE corrected extent threshold of p < 0.05. PFC, prefrontal cortex; SFG, superior frontal gyrus; AG, angular gyrus; (L), left hemisphere; (R), right hemisphere; N.S, not significant. Open in new tab Table 3. Seed-based functional connectivity results for the HR > LR contrast . . . MNI coordinates . . . Seed . Scan . Brain region . X . Y . Z . Peak T-value . Cluster size . MPFC Pre-nap N.S. 5 min Precuneus (L) −4 −76 52 4.75 487 25 min Putamen (R) 14 14 −8 5.15 361 PCC Pre-nap N.S. 5 min PFC (R) 36 56 14 4.22 505 SFG (R) 20 28 46 3.90 378 25 min N.S. LP Pre-nap N.S. 5 min SFG (R) 22 26 48 4.26 372 AG (R) 36 −54 42 4.05 309 25 min N.S. . . . MNI coordinates . . . Seed . Scan . Brain region . X . Y . Z . Peak T-value . Cluster size . MPFC Pre-nap N.S. 5 min Precuneus (L) −4 −76 52 4.75 487 25 min Putamen (R) 14 14 −8 5.15 361 PCC Pre-nap N.S. 5 min PFC (R) 36 56 14 4.22 505 SFG (R) 20 28 46 3.90 378 25 min N.S. LP Pre-nap N.S. 5 min SFG (R) 22 26 48 4.26 372 AG (R) 36 −54 42 4.05 309 25 min N.S. The seed regions are the MPFC (center of mass in MNI coordinates = 1, 55, −3), the PCC (1, −61, 38), and the lateral parietal cortices (LP, right = 47, −67, 29, left = −39, −77,33; the functional connectivity contrast was done using the main effect of both hemispheres). These three seeds are part of the DMN, a network that has been suggested to be involved in dream recall and/or production. The functional connectivity between the seed and the regions reported was higher in HR than LR (see also Figure 3). Statistical analyses were performed using a cluster-defining voxel-wise height threshold of p < 0.01 (uncorrected, two-sided) and a whole-brain FWE corrected extent threshold of p < 0.05. PFC, prefrontal cortex; SFG, superior frontal gyrus; AG, angular gyrus; (L), left hemisphere; (R), right hemisphere; N.S, not significant. Open in new tab Figure 3. Open in new tabDownload slide Functional connectivity differences between HR and LR during pre-sleep scan, 5 min post-awakening scan and 25 min post-awakening scan. The seed regions are the MPFC (center of mass in MNI coordinates = 1, 55, −3), the PCC (1, −61, 38), and the lateral parietal cortices (LP, right = 47, −67, 29, left = −39, −77, 33; the functional connectivity contrast was done using the main effect of both hemispheres combined). These three seeds are part of the DMN, a network that has been suggested to be involved in dream recall and/or production. Statistical parametric maps are superimposed on a glass brain using an uncorrected two-sided cluster-defining voxel-wise height threshold of p < 0.01 and a whole-brain FWE-corrected extent threshold of p < 0.05. Faded brain denotes an absence of significant differences between the two groups for this specific seed and time point (see also Table 3). Figure 3. Open in new tabDownload slide Functional connectivity differences between HR and LR during pre-sleep scan, 5 min post-awakening scan and 25 min post-awakening scan. The seed regions are the MPFC (center of mass in MNI coordinates = 1, 55, −3), the PCC (1, −61, 38), and the lateral parietal cortices (LP, right = 47, −67, 29, left = −39, −77, 33; the functional connectivity contrast was done using the main effect of both hemispheres combined). These three seeds are part of the DMN, a network that has been suggested to be involved in dream recall and/or production. Statistical parametric maps are superimposed on a glass brain using an uncorrected two-sided cluster-defining voxel-wise height threshold of p < 0.01 and a whole-brain FWE-corrected extent threshold of p < 0.05. Faded brain denotes an absence of significant differences between the two groups for this specific seed and time point (see also Table 3). First, we did not find any significant group differences in the functional connectivity of any of the three seeds during the pre-nap resting-state scan. In other words, the two groups did not differ in the DMN functional connectivity before the nap. Similarly, there were almost no group differences during the 25-min post-awakening resting-state scan, except increased connectivity between the MPFC and the putamen in HR. The larger group effect was observed at 5 min post-awakening. The functional connectivity of MPFC, PCC, and LP was increased in HR as compared to LR, specifically between the MPFC and the precuneus, the PCC and the right PFC, the PCC and the superior frontal gyrus (SFG), the LP and the SFG, and the LP and the right angular gyrus (AG). Discussion This study is the first one to compare the brain functional connectivity and cognitive performance of high and LR in the minutes following awakening from sleep, in an effort to understand the basis for interindividual differences in dream recall. In line with our predictions, we have found a group difference in dream recall after awakening from the nap between the two groups despite no significant difference in the recall of a comic strip that was presented to the participants just before the nap. The group difference in dream recall was stage-specific, such that significantly more HR than LR recalled a dream after being awakened from N2 sleep, but not from N3 sleep. Second, contrary to our hypothesis, we did not find a significant group difference in the behavioral impairment caused by sleep inertia (as measured by a mental calculation task, Figure 2). Third, we found several differences in brain functional connectivity between the two groups at 5 min post-awakening (Figure 3). Compared to LR, HR showed increased functional connectivity between regions of the DMN and several regions known to be involved in memory processes at 5 min post-awakening, and between the MPFC and the right putamen at 25 min post-awakening. Importantly, no group differences in functional connectivity were found during the pre-nap scan, therefore suggesting that these changes were specific to the post-awakening period. Between-group contrasts of the functional connectivity at 5 min post-awakening revealed a pattern of enhanced connectivity in HR within the DMN and between the DMN and the right PFC and AG. Remarkably, all these regions have been associated with memory processes. Indeed, a meta-analysis of the brain areas associated with episodic encoding and retrieval [46] found that the precuneus, PFC, SFG, and AG were among the largest clusters associated with retrieval success. These are almost the same regions found to be more functionally connected in HR than in LR during the first minutes following awakening. In sum, our findings confirm that the cerebral functioning of HR and LR is different, as previously demonstrated with EEG and PET [9, 10, 13], and show that specific group differences in memory-related brain regions are present during the sleep-wake transition. It is generally admitted that sleep inertia interferes more with memory retrieval on awakening than with memory consolidation [14, 32, 47, 48]. This fits well with our results showing a dream recall difference between groups at awakening but no difference in the memory recall of the comic strip that was read by the participants before the nap. These findings suggest that the pattern of enhanced functional connectivity within the DMN in HR at awakening facilitates the maintenance of short term memory content during the transition between sleep and wake rather than long-term memory recall. Building on this interpretation, one may expect to find a clear between-group difference in DST performance immediately after awakening, given that this task is considered to recruit both short-term and working memories. This counterintuitive result may be partly explained by the long delay between awakening and the first post-nap DST (the total delay is the duration of the scan plus the average delay between awakening and the onset of the scan, i.e. about 10 min), which is a consequence of our decision to prioritize the measure of brain functioning over the measure of performance upon awakening. One may hypothesize that the between-group difference at the DST would have been larger if this latter was performed immediately after awakening, as is the case in most studies using this specific task [14, 15, 47]. However, future studies are needed to confirm or refute this speculation. Altogether, our findings suggest that an increased DMN connectivity upon awakening in HR facilitates the maintenance of dreams in short-term memory during the sleep–wake transition. This hypothesis is coherent with the known cerebral correlates of short-term memory, which consistently involve the PFC and AG [49, 50]. In addition, the MPFC and precuneus are known to be involved in a state of mind attribution such as envisioning the future (prospection), and remembering the past (episodic memory) [25, 46, 51, 52]. A proper functional coupling between these brain areas seems thus appropriate to maintain in working memory a mental representation involving fragment of episodic memory [53, 54] and an imaginary/prospective experience with interacting characters (more than 95% of dream reports involve characters [55]. In conclusion, our results suggest that the increased functional connectivity in HR after awakening would promote successful dream recall at awakening by facilitating the maintenance of short-term memory. More broadly, we argue that brain activity at awakening is a critical parameter to explain the difference in DRF between HR and LR. Limitations and future directions Participants underwent a partial sleep deprivation on the night before the scanning session, which may have impacted functional connectivity at awakening of the early afternoon nap [56]. The rationale for doing a partial sleep deprivation was twofold. First, by increasing sleep pressure, we hoped that the subjects would have less trouble falling asleep in the uncomfortable MR environment [57]. Second, because sleep inertia is known to be higher following sleep deprivation, we hoped to maximize sleep inertia upon awakening. We believe that the partial sleep deprivation does not undermine the relevance of our findings since naps are usually taken as a strategy to compensate for sleep loss [58]. Furthermore, it is likely that the impact of the partial sleep deprivation on the functional connectivity, if any, was similar in HR and LR (though this remains to be tested). Future studies are needed to determine whether our results replicate for morning awakenings without prior sleep deprivation. Second, building on previous experimental findings from our team, we decided to focus solely on the relationship between sleep inertia and trait dream recall, that is, interindividual differences in DRF. Future studies should investigate whether state dream recall factors (such as the different sleep stages) involve the recruitment of the same functional networks that were identified in the current study. For intra-individual variability, based on our current findings, one may expect that enhanced connectivity between the DMN and memory-related brain areas would be also associated with higher dream recall, that is, that within an individual there would be a positive correlation between higher DMN connectivity at awakening and successful dream recall. It could notably explain the nonrapid eye movement (NREM) versus REM sleep difference in dream recall since the DMN is known to be more functionally connected during REM sleep than during NREM sleep [31, 59]. It is also important to stress that dream recall and content (complexity, vividness, and bizarreness) of dream reports are not only modulated by the sleep stage prior to awakening but also by a circadian-driven cortical arousability, which results in a strong decrease in the recall frequency and content differences between N2 and REM dreams at the end of the night. That is, at the end of the night, reports may be as frequent in N2 and REM sleep and some N2 dream reports are indistinguishable from REM sleep dream reports [4, 60–62]. Anecdotally, in the current study, some participants reported vivid and bizarre dreams resembling REM morning dreams, even though they were awakened from N2 or N3 sleep (e.g. one participant awakened in N3 sleep had a very vivid dream that a deer was having lunch with him in the lab and then taking him to the MRI facility). Further research is needed to understand how seemingly different brain states (NREM and REM) can lead to a similar subjective experience. A promising avenue of research to understand the relationship between the awakening brain and dream recall is the emerging concept of local sleep [63]. For instance, it would be interesting to explore more locally how memory-related brain regions show an intrusion of sleep-like features during the process of awakening, as has been done in rodents [64]. Increased (inhibitory) slow-wave activity in memory-related regions upon awakening, in particular, could explain low dream recall. Intracranial EEG would be the ideal technique to address this issue in humans (see e.g. [65] even if it presents several important limitations [66], that is, investigation of an epileptic brain, with the hippocampus being often part of the epileptic zone). The identification of local sleep inertia in humans should help to reconcile seemingly contradicting results from previous research. Lastly, while our results provide evidence that a specific pattern of functional connectivity upon awakening, involving the DMN and memory-related brain areas, is related to interindividual differences in dream recall, it is still unknown whether that same brain pattern is similarly involved in dream production while asleep, as hypothesized by some [27, 67]. A major obstacle to test this hypothesis is that there is currently no objective neurophysiological marker of dreaming during sleep [3], which means that researchers cannot study the dreaming brain while the dreamer is asleep, and must rely instead on dream reports collected after induced or spontaneous awakening from sleep (noteworthy, some correlates of dream recall based on the pre-awakening EEG power have been reported, though results are somewhat inconsistent across studies [68–72]). Still, there are reasons to believe that the DMN might be indeed involved in dream production. First, memory structures are certainly involved in the production of dreams since dream content is typically composed of episodic and autobiographical memory fragments [54, 55, 73–76]. Coherently, neuroimaging studies have shown that DMN and memory-related brain regions are active during sleep (especially REM sleep [77–79], and even more so in HR [13]. Second, lesions in core regions of the DMN lead to a complete cessation of dream reporting [21]. Third, daydreaming, which shares several phenomenological properties with night dreaming, is characterized by a recruitment of DMN regions [80]. Finally, the DMN is known to be more functionally connected during the REM sleep stage (as compared to NREM sleep) which is the stage associated with the highest DRF [31, 59]. Related to this issue of differentiating dream production and dream recall, another question pertains to the availability of dream content upon awakening from sleep. Can only the dream content that is present in short-term memory at the time of awakening be remembered? Or rather, is there some form of online consolidation of dream content while dreaming which would allow the dreamer to remember dream content beyond the limited capacity of short-term memory? If so, an ideal candidate would be the intermediate-term memory system, in which memories can be recovered after several hours even after dropping out of the attentional focus [81]. Here again, the absence of a physiological marker of dreaming during sleep makes it difficult to answer these questions. However, future studies could use lucid dreaming with eye movements signaling [82] as a proxy to assess the dream-related memorization mechanisms during sleep (see for example [83]; though we recently advised caution in translating findings from lucid dreaming to non-lucid dreaming [84]). Conclusion The present study showed that interindividual trait differences in DRF are associated with differences in brain functional connectivity in the first minutes following awakening. Specifically, HR showed a higher functional connectivity upon awakening between regions involved in short term and episodic memory, which could facilitate the maintenance in short term memory of the pre-awakening experience (i.e. the dream) and thus promote dream recall. More broadly, these results contribute to the growing body of evidence that trait differences in dream recall are supported by trait neurophysiological differences between HR and LR. Further work is needed to better delineate the links between neurophysiological factors, psychological factors, and interindividual differences in dream recall. Acknowledgments The authors would like to thank Basak Turker, Morgane Hamon, David Meunier, Franck Lamberton, and Danielle Ibarrola for substantial help in data collection and analysis, as well as Jamila Lagha for her help in administrative work. Funding This work was partly performed within the framework of the LABEX CORTEX (ANR-11-LABX-0042) of Agence Nationale de la Recherche (ANR), within the program ANR-11-IDEX-0007. Disclosure statement Financial disclosure: None declared. Non-financial disclosure: None declared. Author contribution R.V, A.N, and P.R designed the study. R.V and P.R acquired the data and wrote the first draft of the paper. A.N provided access to his sleep unit to conduct the overnight partial sleep deprivation. All authors read the final version of the manuscript. References 1. Schredl M , et al. Dream recall, dream length, and sleep duration: state or trait factor . Percept Mot Skills. 2008 ; 106 ( 2 ): 633 – 636 . Google Scholar Crossref Search ADS PubMed WorldCat 2. Schredl M . Reliability and stability of a dream recall frequency scale . Percept Mot Skills. 2004 ; 98 ( 3 Pt 2 ): 1422 – 1426 . Google Scholar Crossref Search ADS PubMed WorldCat 3. Ruby PM . Experimental research on dreaming: state of the art and neuropsychoanalytic perspectives . Front Psychol. 2011 ; 2 : 286 . Google Scholar Crossref Search ADS PubMed WorldCat 4. Chellappa SL , et al. Ultradian and circadian modulation of dream recall: EEG correlates and age effects . Int J Psychophysiol. 2013 ; 89 ( 2 ): 165 – 170 . Google Scholar Crossref Search ADS PubMed WorldCat 5. Tribl GG , et al. Dreaming under antidepressants: a systematic review on evidence in depressive patients and healthy volunteers . Sleep Med Rev. 2013 ; 17 ( 2 ): 133 – 142 . Google Scholar Crossref Search ADS PubMed WorldCat 6. Goodenough DR , et al. Dream reporting following abrupt and gradual awakenings from different types of sleep . J Pers Soc Psychol. 1965 ; 2 : 170 – 179 . Google Scholar Crossref Search ADS PubMed WorldCat 7. Schredl M , et al. Factors of home dream recall: a structural equation model . J Sleep Res. 2003 ; 12 ( 2 ): 133 – 141 . Google Scholar Crossref Search ADS PubMed WorldCat 8. Koulack D , et al. Dream recall and dream recall failure: an arousal-retrieval model . Psychol Bull . 1976 ; 83 ( 5 ): 975 – 984 . Google Scholar Crossref Search ADS WorldCat 9. Eichenlaub JB , et al. Brain reactivity differentiates subjects with high and low dream recall frequencies during both sleep and wakefulness . Cereb Cortex. 2014 ; 24 ( 5 ): 1206 – 1215 . Google Scholar Crossref Search ADS PubMed WorldCat 10. Vallat R , et al. Increased evoked potentials to arousing auditory stimuli during sleep: implication for the understanding of dream recall . Front Hum Neurosci. 2017 ; 11 : 132 . Google Scholar Crossref Search ADS PubMed WorldCat 11. Goodenough DR , et al. A comparison of “dreamers” and “nondreamers”: eye movements, electroencephalograms, and the recall of dreams . J Abnorm Soc Psychol . 1959 ; 59 ( 3 ): 295 . Google Scholar Crossref Search ADS WorldCat 12. Pagel JF . Non-dreamers . Sleep Med. 2003 ; 4 ( 3 ): 235 – 241 . Google Scholar Crossref Search ADS PubMed WorldCat 13. Eichenlaub JB , et al. Resting brain activity varies with dream recall frequency between subjects . Neuropsychopharmacology. 2014 ; 39 ( 7 ): 1594 – 1602 . Google Scholar Crossref Search ADS PubMed WorldCat 14. Tassi P , et al. Sleep inertia . Sleep Med Rev. 2000 ; 4 ( 4 ): 341 – 353 . Google Scholar Crossref Search ADS PubMed WorldCat 15. Trotti LM . Waking up is the hardest thing I do all day: sleep inertia and sleep drunkenness . Sleep Med Rev. 2017 ; 35 : 76 – 84 . Google Scholar Crossref Search ADS PubMed WorldCat 16. Vallat R , et al. Hard to wake up? The cerebral correlates of sleep inertia assessed using combined behavioral, EEG and fMRI measures . Neuroimage. 2019 ; 184 : 266 – 278 . Google Scholar Crossref Search ADS PubMed WorldCat 17. Koukkou M , et al. Dreaming: the functional state-shift hypothesis. A neuropsychophysiological model . Br J Psychiatry. 1983 ; 142 : 221 – 231 . Google Scholar Crossref Search ADS PubMed WorldCat 18. Morel CR , et al. The electrophysiological correlates of dream recall and nonrecall from stage 2 sleep . Can J Psychol. 1991 ; 45 ( 2 ): 140 – 147 . Google Scholar Crossref Search ADS PubMed WorldCat 19. Rochlen A , et al. EEG correlates of dream recall in depressed outpatients and healthy controls . Dreaming . 8 ( 2 ): 109 – 123 . Crossref Search ADS WorldCat 20. Vallat R , et al. Dream recall frequency is associated with medial prefrontal cortex white-matter density . Front Psychol. 2018 ; 9 : 1856 . Google Scholar Crossref Search ADS PubMed WorldCat 21. Solms M. Institute for Research in Behavioral Neuroscience.The neuropsychology of dreams: A clinico-anatomical study . Lawrence Erlbaum Associates Publishers ; 1997 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC 22. Solms M . Dreaming and REM sleep are controlled by different brain mechanisms . Behav Brain Sci. 2000 ; 23 ( 6 ): 843 – 850; discussion 904 . Google Scholar Crossref Search ADS PubMed WorldCat 23. Raichle ME , et al. A default mode of brain function . Proc Natl Acad Sci U S A. 2001 ; 98 ( 2 ): 676 – 682 . Google Scholar Crossref Search ADS PubMed WorldCat 24. Sestieri C , et al. Episodic memory retrieval, parietal cortex, and the default mode network: functional and topographic analyses . J Neurosci. 2011 ; 31 ( 12 ): 4407 – 4420 . Google Scholar Crossref Search ADS PubMed WorldCat 25. Legrand D , et al. What is self-specific? Theoretical investigation and critical review of neuroimaging results . Psychol Rev. 2009 ; 116 ( 1 ): 252 – 282 . Google Scholar Crossref Search ADS PubMed WorldCat 26. William Domhoff G . The neural substrate for dreaming: is it a subsystem of the default network? Conscious Cogn. 2011 ; 20 ( 4 ): 1163 – 1174 . Google Scholar Crossref Search ADS PubMed WorldCat 27. Christoff K , et al. Mind-wandering as spontaneous thought: a dynamic framework . Nat Rev Neurosci. 2016 ; 17 ( 11 ): 718 – 731 . Google Scholar Crossref Search ADS PubMed WorldCat 28. Cavallero C , et al. Slow wave sleep dreaming . Sleep. 1992 ; 15 ( 6 ): 562 – 566 . Google Scholar Crossref Search ADS PubMed WorldCat 29. Nielsen TA . A review of mentation in REM and NREM sleep: “covert” REM sleep as a possible reconciliation of two opposing models . Behav Brain Sci. 2000 ; 23 ( 6 ): 851 – 866; discussion 904 . Google Scholar Crossref Search ADS PubMed WorldCat 30. Horovitz SG , et al. Decoupling of the brain’s default mode network during deep sleep . Proc Natl Acad Sci U S A. 2009 ; 106 ( 27 ): 11376 – 11381 . Google Scholar Crossref Search ADS PubMed WorldCat 31. Wu CW , et al. Variations in connectivity in the sensorimotor and default-mode networks during the first nocturnal sleep cycle . Brain Connect. 2012 ; 2 ( 4 ): 177 – 190 . Google Scholar Crossref Search ADS PubMed WorldCat 32. Conduit R , et al. Poor recall of eye-movement signals from Stage 2 compared to REM sleep: implications for models of dreaming . Conscious Cogn. 2004 ; 13 ( 3 ): 484 – 500 . Google Scholar Crossref Search ADS PubMed WorldCat 33. Vallat R , et al. Sleep and dream habits in a sample of French college students who report no sleep disorders . J Sleep Res. 2018 ; 27 ( 5 ): e12659 . Google Scholar Crossref Search ADS PubMed WorldCat 34. Dinges DF , et al. Assessing performance upon abrupt awakening from naps during quasi-continuous operations . Behav Res Methods Instrum Comput . 1985 ; 17 ( 1 ): 37 – 45 . Google Scholar Crossref Search ADS WorldCat 35. Evans FJ , et al. Recovery from Fatigue . Philadelphia, PA: Pennsylvania Hospital ; 1975 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC 36. Stampi C , et al. Ultrashort sleep schedules: sleep architecture and the recuperative value of multiple 80- 50- and 20-min naps . In: Horne J, ed. Sleep 90’ . Bochum : Pontagenel Press ; 1990 : 71 – 74 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC 37. Mullinger KJ , et al. Reducing the gradient artefact in simultaneous EEG-fMRI by adjusting the subject’s axial position . Neuroimage. 2011 ; 54 ( 3 ): 1942 – 1950 . Google Scholar Crossref Search ADS PubMed WorldCat 38. Iber C , et al. The AASM Manual for the Scoring of Sleep and Associated Events: Rules, Terminology and Technical Specifications. Vol 1 . Westchester (IL) : American Academy of Sleep Medicine ; 2007 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC 39. Silber MH , et al. The visual scoring of sleep in adults . J Clin Sleep Med. 2007 ; 3 ( 2 ): 121 – 131 . Google Scholar Crossref Search ADS PubMed WorldCat 40. Combrisson E , et al. Sleep: an open-source Python software for visualization, analysis, and staging of sleep data . Front Neuroinform. 2017 ; 11 : 60 . Google Scholar Crossref Search ADS PubMed WorldCat 41. Whitfield-Gabrieli S , et al. Conn: a functional connectivity toolbox for correlated and anticorrelated brain networks . Brain Connect. 2012 ; 2 ( 3 ): 125 – 141 . Google Scholar Crossref Search ADS PubMed WorldCat 42. Behzadi Y , et al. A component based noise correction method (CompCor) for BOLD and perfusion based fMRI . Neuroimage. 2007 ; 37 ( 1 ): 90 – 101 . Google Scholar Crossref Search ADS PubMed WorldCat 43. Weissenbacher A , et al. Correlations and anticorrelations in resting-state functional connectivity MRI: a quantitative comparison of preprocessing strategies . Neuroimage. 2009 ; 47 ( 4 ): 1408 – 1416 . Google Scholar Crossref Search ADS PubMed WorldCat 44. Vallat R . Pingouin: statistics in Python . JOSS . 2018 ; 3 ( 31 ): 1026 . Google Scholar Crossref Search ADS WorldCat 45. Power JD , et al. Spurious but systematic correlations in functional connectivity MRI networks arise from subject motion . Neuroimage. 2012 ; 59 ( 3 ): 2142 – 2154 . Google Scholar Crossref Search ADS PubMed WorldCat 46. Spaniol J , et al. Event-related fMRI studies of episodic encoding and retrieval: meta-analyses using activation likelihood estimation . Neuropsychologia. 2009 ; 47 ( 8–9 ): 1765 – 1779 . Google Scholar Crossref Search ADS PubMed WorldCat 47. Bonnet MH . Memory for events occurring during arousal from sleep . Psychophysiology. 1983 ; 20 ( 1 ): 81 – 87 . Google Scholar Crossref Search ADS PubMed WorldCat 48. Dinges DF . Are you awake? Cognitive performance and reverie during the hypnopompic state . In: Bootzin RR, ed. Sleep and Cognition . Vol. 211 . Washington (DC) : American Psychological Association, ix ; 1990 : 159 – 175 . Google Scholar Crossref Search ADS Google Preview WorldCat COPAC 49. Christophel TB , et al. The distributed nature of working memory . Trends Cogn Sci. 2017 ; 21 ( 2 ): 111 – 124 . https://www.sciencedirect.com/science/article/pii/S1364661316302170. Google Scholar Crossref Search ADS PubMed WorldCat 50. Caclin A , et al. Musical and verbal short-term memory: insights from neurodevelopmental and neurological disorders . Ann N Y Acad Sci . 2018 ; 1423 (1): 155 – 165 . Google Scholar Crossref Search ADS WorldCat 51. Ruby P , et al. Effect of subjective perspective taking during simulation of action: a PET investigation of agency . Nat Neurosci. 2001 ; 4 ( 5 ): 546 – 550 . Google Scholar Crossref Search ADS PubMed WorldCat 52. Buckner RL , et al. Self-projection and the brain . Trends Cogn Sci. 2007 ; 11 ( 2 ): 49 – 57 . Google Scholar Crossref Search ADS PubMed WorldCat 53. Fosse MJ , et al. Dreaming and episodic memory: a functional dissociation? J Cogn Neurosci. 2003 ; 15 ( 1 ): 1 – 9 . Google Scholar Crossref Search ADS PubMed WorldCat 54. Schwartz S . Are life episodes replayed during dreaming? Trends Cogn Sci. 2003 ; 7 ( 8 ): 325 – 327 . Google Scholar Crossref Search ADS PubMed WorldCat 55. Vallat R , et al. Characteristics of the memory sources of dreams: A new version of the content-matching paradigm to take mundane and remote memories into account . PLoS One. 2017 ; 12 ( 10 ): e0185262 . Google Scholar Crossref Search ADS PubMed WorldCat 56. Sämann PG , et al. Increased sleep pressure reduces resting state functional connectivity . MAGMA. 2010 ; 23 ( 5–6 ): 375 – 389 . Google Scholar Crossref Search ADS PubMed WorldCat 57. Duyn JH . EEG-fMRI methods for the study of brain networks during sleep . Front Neurol. 2012 ; 3 : 100 . Google Scholar Crossref Search ADS PubMed WorldCat 58. Faraut B , et al. Napping: a public health issue. From epidemiological to laboratory studies . Sleep Med Rev. 2017 ; 35 : 85 – 100 . Google Scholar Crossref Search ADS PubMed WorldCat 59. Chow HM , et al. Rhythmic alternating patterns of brain activity distinguish rapid eye movement sleep from other states of consciousness . Proc Natl Acad Sci U S A. 2013 ; 110 ( 25 ): 10300 – 10305 . Google Scholar Crossref Search ADS PubMed WorldCat 60. Pivik T , et al. NREM mentation: relation to personality, orientation time, and time of night . J Consult Clin Psychol. 1968 ; 32 ( 2 ): 144 – 151 . Google Scholar Crossref Search ADS PubMed WorldCat 61. Wittmann L , et al. NREM sleep dream recall, dream report length and cortical activation . Sleep Hypn . 2004 ; 6 : 54 – 58 . http://www.sleepandhypnosis.org/ing/Pdf/122c34f0b6f04a4bb95fd16bbc519aa3.pdf. 62. Wamsley EJ , et al. Circadian and ultradian influences on dreaming: a dual rhythm model . Brain Res Bull. 2007 ; 71 ( 4 ): 347 – 354 . Google Scholar Crossref Search ADS PubMed WorldCat 63. Vyazovskiy VV , et al. Local sleep in awake rats . Nature. 2011 ; 472 ( 7344 ): 443 – 447 . Google Scholar Crossref Search ADS PubMed WorldCat 64. Vyazovskiy VV , et al. The dynamics of cortical neuronal activity in the first minutes after spontaneous awakening in rats and mice . Sleep. 2014 ; 37 ( 8 ): 1337 – 1347 . Google Scholar Crossref Search ADS PubMed WorldCat 65. Peter-Derex L , et al. Heterogeneity of arousals in human sleep: A stereo-electroencephalographic study . Neuroimage. 2015 ; 123 : 229 – 244 . Google Scholar Crossref Search ADS PubMed WorldCat 66. Parvizi J , et al. Promises and limitations of human intracranial electroencephalography . Nat Neurosci. 2018 ; 21 ( 4 ): 474 – 483 . Google Scholar Crossref Search ADS PubMed WorldCat 67. Domhoff GW , et al. Dreaming and the default network: A review, synthesis, and counterintuitive research proposal . Conscious Cogn. 2015 ; 33 : 342 – 353 . Google Scholar Crossref Search ADS PubMed WorldCat 68. Esposito MJ , et al. Reduced Alpha power associated with the recall of mentation from stage 2 and stage REM sleep . Psychophysiology. 2004 ; 41 ( 2 ): 288 – 297 . Google Scholar Crossref Search ADS PubMed WorldCat 69. Chellappa SL , et al. Cortical activation patterns herald successful dream recall after NREM and REM sleep . Biol Psychol. 2011 ; 87 ( 2 ): 251 – 256 . Google Scholar Crossref Search ADS PubMed WorldCat 70. Marzano C , et al. Recalling and forgetting dreams: theta and alpha oscillations during sleep predict subsequent dream recall . J Neurosci. 2011 ; 31 ( 18 ): 6674 – 6683 . Google Scholar Crossref Search ADS PubMed WorldCat 71. Siclari F , et al. The neural correlates of dreaming . Nat Neurosci. 2017 ; 20 ( 6 ): 872 – 878 . Google Scholar Crossref Search ADS PubMed WorldCat 72. Wong W , et al. The Dream Catcher experiment: blinded analyses disconfirm markers of dreaming consciousness in EEG spectral power . bioRxiv . 2019 : 643593 . Google Scholar OpenURL Placeholder Text WorldCat 73. Plailly J , et al. Incorporation of fragmented visuo-olfactory episodic memory into dreams and its association with memory performance . Sci Rep. 2019 ; 9 ( 1 ): 15687 . Google Scholar Crossref Search ADS PubMed WorldCat 74. Grenier J , et al. Temporal references in dreams and autobiographical memory . Mem Cognit. 2005 ; 33 ( 2 ): 280 – 288 . Google Scholar Crossref Search ADS PubMed WorldCat 75. Malinowski JE , et al. Memory sources of dreams: the incorporation of autobiographical rather than episodic experiences . J Sleep Res. 2014 ; 23 ( 4 ): 441 – 447 . Google Scholar Crossref Search ADS PubMed WorldCat 76. Baylor GW , et al. Memory sources associated with REM and NREM dream reports throughout the night: a new look at the data . Sleep. 2001 ; 24 ( 2 ): 165 – 170 . Google Scholar PubMed OpenURL Placeholder Text WorldCat 77. Maquet P , et al. Human cognition during REM sleep and the activity profile within frontal and parietal cortices: a reappraisal of functional neuroimaging data. In: Laureys S, ed. Progress in Brain Research . Vol 150 . Elsevier ; 2005 : 219 – 595 . Google Scholar Crossref Search ADS Google Preview WorldCat COPAC 78. Picchioni D , et al. Sleep and the functional connectome . Neuroimage. 2013 ; 80 : 387 – 396 . Google Scholar Crossref Search ADS PubMed WorldCat 79. Maquet P . Functional neuroimaging of normal human sleep by positron emission tomography . J Sleep Res. 2000 ; 9 ( 3 ): 207 – 231 . Google Scholar Crossref Search ADS PubMed WorldCat 80. Fox KC , et al. The wandering brain: meta-analysis of functional neuroimaging studies of mind-wandering and related spontaneous thought processes . Neuroimage. 2015 ; 111 : 611 – 621 . Google Scholar Crossref Search ADS PubMed WorldCat 81. Kamiński J . Intermediate-term memory as a bridge between working and long-term memory . J Neurosci . 2017 ; 37 ( 20 ): 5045 – 5047 . Google Scholar Crossref Search ADS PubMed WorldCat 82. La Berge SP , et al. Lucid dreaming verified by volitional communication during REM sleep . Percept Mot Skills. 1981 ; 52 ( 3 ): 727 – 732 . Google Scholar Crossref Search ADS PubMed WorldCat 83. Mallett R . Partial memory reinstatement while (lucid) dreaming to change the dream environment . Conscious Cogn . 2020 ; 83 : 102974 . Google Scholar Crossref Search ADS PubMed WorldCat 84. Vallat R , et al. Is it a good idea to cultivate lucid dreaming? Front Psychol. 2019 ; 10 : 2585 . Google Scholar Crossref Search ADS PubMed WorldCat © Sleep Research Society 2020. Published by Oxford University Press on behalf of the Sleep Research Society. All rights reserved. For permissions, please e-mail [email protected]. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)