Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Use of wearable devices for post-discharge monitoring of ICU patients: a feasibility study

Use of wearable devices for post-discharge monitoring of ICU patients: a feasibility study Background: Wearable devices generate signals detecting activity, sleep, and heart rate, all of which could enable detailed and near-continuous characterization of recovery following critical illness. Methods: To determine the feasibility of using a wrist-worn personal fitness tracker among patients recovering from critical illness, we conducted a prospective observational study of a convenience sample of 50 stable ICU patients. We assessed device wearability, the extent of data capture, sensitivity and specificity for detecting heart rate excursions, and correlations with questionnaire-derived sleep quality measures. Results: Wearable devices were worn over a 24-h period, with excellent capture of data. While specificity for the detection of tachycardia was high (98.8%), sensitivity was low to moderate (69.5%). There was a moderate correlation between wearable-derived sleep duration and questionnaire-derived sleep quality (r = 0.33, P =0.03). Devices were well-tolerated and demonstrated no degradation in quality of data acquisition over time. Conclusions: Wefoundthat wearabledevices couldbewornbypatients recovering from critical illness and could generate useful data for the majority of patients with little adverse effect. Further development and study are needed to better define and enhance the role of wearables in the monitoring of post-ICU recovery. Trial registration: Clinicaltrials.gov, NCT02527408 Keywords: Wearable devices, Medical informatics, Mobile health technologies, Validation study, Critical care, Sleep quality, Heart rate monitoring Background devices generate data that could also be useful in charac- Consumer interest in personal health tracking has recently terizing convalescence from acute illness. These include increased, leading to an industry in wearable devices now photoplethysmography (PPG) sensors to detect heart rate valued at more than $9 billion worldwide [1]. With more [6, 7], as well as accelerometers to track activity and wearables in use than ever before, there has been growing movement [3, 8, 9]. enthusiasm for their potential to improve health care de- Frequent heart rate tracking has the potential to identify livery [2]. Current clinical uses for wearable devices are episodes of clinical deterioration early. Accelerometer data mostly limited to outpatient settings, with a focus on the could potentially be used to encourage mobilization, management of chronic diseases [3–5]. Newer generation objectively measure functional status, and track progress towards rehabilitation goals. Wrist-worn accelerometers * Correspondence: [email protected] have also been used to evaluate sleep quality in healthy Equal contributors 1 subjects [10, 11]. In the inpatient and intensive care unit Department of Critical Care Medicine, Queen’s University and Kingston Health Sciences Centre, Kingston, Ontario, Canada (ICU) settings, where poor sleep has been linked with Department of Medicine, Queen’s University and Kingston Health Sciences adverse outcomes [12, 13], data describing sleep quality Centre, Kingston, Ontario, Canada Full list of author information is available at the end of the article © The Author(s). 2017 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Kroll et al. Journal of Intensive Care (2017) 5:64 Page 2 of 8 may be useful in identifying targets for sleep-promoting analgesia. We specifically chose to study patients who interventions [14]. were still in the ICU, as this was the most practical way to There is little clinical evidence to inform the practice obtain gold standard measurements of heart rate and of using wearables in health care, most of which is fo- sleep quality, which would otherwise require the use of cused on chronic conditions. Newer consumer-grade Holter monitors and complex follow-up procedures. To wearables have been evaluated in only a handful of stud- reduce the potential risk of transmitting nosocomial infec- ies examining their accuracy among healthy volunteers tions, patients under contact precautions for methicillin- [3–5]. These studies have called for evaluations of this resistant Staphylococcus aureus (MRSA) and Clostridium technology among a wider range of patient populations. difficile infections were also excluded. We also excluded In this study, we examine the feasibility of using a patients at risk of vascular compromise of the arm on common consumer-grade wearable device to monitor which the wearable device was to be placed, such as pa- patients recovering from critical illness. We enrolled pa- tients with upper extremity deep venous thrombosis, per- tients who no longer required intensive care measures ipherally inserted central catheters, radial arterial lines, but remained in the ICU prior to ward transfer, in order dialysis fistulas, and severe upper extremity trauma. As to best approximate post-ICU settings like the general this was a feasibility study, a convenience sample of 50 wards, while still collecting gold standard data to valid- participants was recruited. ate device functionality. We report on a number of prac- tical considerations that could affect the deployment of Ethics, consent, and permissions wearables including overall wearability, completeness of All participating patients, or substitute decision makers data capture, device longevity, and risk of transmitting on their behalf, provided written informed consent for nosocomial infections. We also evaluated the accuracy participation in this study. The Health Sciences Research of wearables for measuring sleep quality and identifying Ethics Board at Queen’s University reviewed and approved changes in heart rate that might be clinically relevant. the study protocol (DMED-1818-15), and the trial was We hypothesized that patients recovering from critical registered with clinicaltrials.gov (NCT02527408). illness would be able to wear wrist-worn devices and that useful data could be collected from these with a Device moderate degree of accuracy. Participating patients wore the Fitbit Charge HR device (Fitbit, San Francisco, CA, USA) for a single 24-h period Methods (Fig. 1). The Fitbit Charge HR is a commercially avail- Patients and setting able wrist-worn wearable that records heart rate, steps, This prospective observational study was conducted in a and sleep quality. The study employed three size large 33-bed general medical-surgical/trauma ICU in southeast- wearable devices (15.7 to 19.3 cm wrist circumference) ern Ontario, between August 2015 and February 2016. and three size extra-large wearable devices (19.6 to Adult patients (age > 17) were included if they were 22.6 cm wrist circumference). In an effort to reduce receiving continuous cardiac and oxygen saturation the risk of potential iatrogenic infection, we used dis- monitoring, but were otherwise receiving ward-level infectant wipes to thoroughly clean wearables between treatment. Exclusion criteria included mechanical venti- uses. All devices were applied to participants by a lation, vasopressor support, and continuous sedation or study investigator or coordinator. ab Fig. 1 The Fitbit ChargeHRdeviceusedinthe study (a). The wearable device as worn by a patient on the inpatient ward following ICU discharge (b) Kroll et al. Journal of Intensive Care (2017) 5:64 Page 3 of 8 Data monitoring and capture wrist-worn wearables for heart rate tracking in We used continuous pulse oximetry pulse rate record- healthy volunteers [5]. In addition to basic descriptive ings (SPO2-R) as a comparison measure of heart rate statistics, we calculated the sensitivity and specificity of (HR) in order to evaluate the ability of wearables to de- the wearables for detecting tachycardia and bradycardia. tect both tachycardia (HR > 100 bpm) and bradycardia Based on the PPG mechanism of heart rate sensing (HR < 50 bpm). We used SPO2-R values as a compara- employed in consumer-grade wearables, we hypothesized tor as both SPO2-R and wearable device values reflect that the accuracy of wearable device heart rate tracking the pulse rate (rather than electrical heart rate), and be- may be different in patients not in sinus rhythm and fur- cause this is a widely accepted method of heart rate ther analyzed these patients as a subgroup. We calculated measurement. The wearables recorded heart rate values Pearson correlation coefficients between the various every 5 min, while the SPO2-R recorded heart rate wearable-derived measures of sleep quality and the RCSQ values every minute. Cardiac rhythm was assessed at the measures of sleep quality. Based on the mechanism of time of device application, and again at the time of re- sleep sensing, which relies on the absence of movement, moval, at which time data regarding sleep quality was we hypothesized that the accuracy of wearables for sleep also collected using the Richards-Campbell Sleep Ques- tracking may differ in patients with delirium, and further tionnaire (RCSQ) [15]. This survey uses a visual analog analyzed these patients as a subgroup. Statistical analyses scale to assess sleep depth, latency, awakenings, percent- for this study were performed using R (v 3.2.2). age of time awake, and overall quality of sleep. The RCSQ was completed either by the patients themselves or by their designated night shift nurse, a practice previ- Results ously shown to have slight to moderate agreement with Patients and device wearability self-assessment [16]. Due to the interaction between We enrolled a total of 50 patients between August 2015 sleep and delirium in the ICU [17], patients were and January 2016 (Table 1). The median wrist circumfer- screened for delirium by a trained researcher using the ence in our cohort was 18.6 cm (SD 1.9 cm), with 6 of confusion assessment method (CAM)-ICU at the time the 50 patients enrolled having moderate or severe of device application, and again at the time of device edema of the wrist at the time of device application. The removal. size large device was used for 23 patients (46%), while Wearable-reported sleep data included time of sleep on- the size extra-large was used for 27 patients (54%). set and awakening, sleep duration, minutes asleep, mi- While there were no patients for whom the wearable de- nutes awake, restless count, and a calculated measure of vice could not be fitted, the fit was noted to be very tight sleep quality. Overall sleep quality was taken as the aver- Table 1 Characteristics of patients included in the study (n =50) age across sleep episodes, weighted by the duration of Mean heart rate (bpm) 88.3 each sleep episode. The percentage of total sleep occurring Mean age (years) 64 during nighttime hours, which we defined as 22:00 to 06:00, and the percentage of nighttime hours spent asleep Patients (n = 50) % were calculated. For participants who had no Fitbit- Male 26 52 detected sleep over the recording period, a score of 0 was Female 24 48 given for all sleep parameters. Methods for obtaining Admission diagnosis wearable and SPO2-R data are reported elsewhere [18], Respiratory 12 24 and in the Supplementary Content (see Additional file 1). Sepsis 7 14 Surgical 7 14 Microbiological assessment Neurologic 11 22 We conducted microbiologic sampling of the wearables Trauma 3 6 used from a convenience subset of patients (n =16) in order to evaluate both the risk of transmitting nosocomial Cardiovascular 6 12 pathogens from repeated application of wearables to Medical 4 8 different patients, as well as the efficacy of our disinfection Sinus rhythm practices (see Additional file 1). At start of monitoring 43 86 At end of monitoring 42 84 Statistical analysis Personal fitness tracker size used In the absence of preliminary data to inform a sample size Large 23 46 calculation, we targeted an enrollment of 50 patients, a Extra large 27 54 cohort size equal to that used in a similar study of Kroll et al. Journal of Intensive Care (2017) 5:64 Page 4 of 8 in one patient, and very loose in two patients. Devices 5.7/10.0 (IQR 2.7–8.0/10.0). There was a moderate were adjusted only once at the time of application and correlation between wearable-derived sleep duration were not re-assessed by study personnel for the duration and total RCSQ score (r = 0.33, P = 0.03, 95% confi- of the 24-h recording period. No intravenous lines were dence interval [CI] 0.04, 0.58) (Fig. 3). The correlation re-sited in order to facilitate application, although hos- between the percentage of nighttime asleep, as re- pital identification wristbands had to be relocated in ported by the wearable device, and total RCSQ score some cases. No wearables required removal during the was 0.36 (P = 0.02, 95% CI 0.07, 0.60). The correlation monitoring period as a result of patient discomfort. between the Fitbit-reported number of sleep periods The wearable device was removed prior to the com- and RCSQ-reported awakenings was 0.38 (P = 0.01, pletion of the monitoring period in two patients; one 95% CI 0.09, 0.61). There were no significant differ- patient was discharged earlier than expected from the ences in wearable-reported sleep parameters between ICU, while another developed a diffuse drug- the CAM-ICU positive (n = 8) and CAM-ICU negative associated rash. Excluding patients whose devices participants; however, 25% of CAM-ICU positive par- were removed early, the devices were unable to detect ticipants recorded no sleep over the entire 24-h mon- a heart rate reading 4% of the time. itoring period, compared to 8% of CAM-ICU negative participants. Tachycardia and bradycardia detection Device reusability We identified 13 SPO2-R-confirmed readings of brady- Wearables were not found to be a significant source cardia among four patients, all of whom were in sinus of pathogenic bacteria. Microbiologic sampling re- rhythm. Further statistical analysis was not done due to vealed bacteria consistent with commensal skin flora this small sample. The wearable had a sensitivity of (Staphylococcus epidermidis) and/or environmental or- 69.5% and specificity of 98.8% for the detection of tachy- ganisms (Bacillus species). S. epidermidis was only cardia (Table 2 and Fig. 2). Among patients not in sinus observed in samples taken prior to hydrogen peroxide rhythm (n = 8), the specificity for detecting tachycardia disinfection, while Bacillus species were found in both was similar (99.5%), although sensitivity was worse pre- and post-disinfection specimens. Individual wear- (51.6%). For faster heart rates (> 150 bpm), wearable de- able devices were used between 5 and 13 times. vice concordance with SPO2-R was poor. However, in There were no differences in wearable-SPO2-R heart many such cases, the wearable device reading showed rate correlations between the first and second half of better agreement with the true heart rate measured by the study (P = 0.18). continuous ECG, than did the SPO2-R readings, which tended to be falsely high. Discussion The long-term adverse consequences of critical illness Sleep data are increasingly being recognized as a research priority A summary of the sleep quality data collected by the in critical care [19]. A growing body of research is now wearables is shown in Table 3. Among the 47 partici- examining the determinants and potential modifiers of pants who had complete wearable sleep data re- post-ICU recovery, including at least one study that corded, the median wearable-reported sleep duration made use of a wearable device to track patient move- was 6.6 h (interquartile range [IQR] 2.7–13.5 h) and ment and activity [20]. However, post-ICU recovery re- the median number of sleep periods was 2 (IQR 1–4). search currently lacks the richness of data available to Five participants (11%) had no wearable device docu- researchers focused on the ICU stay itself since post- mented sleep for the entirety of the 24-h monitoring discharge data collection is limited to infrequent visits to period. Among the 43 participants for whom the follow-up clinics, or in many cases is nonexistent. New RCSQ was completed, the median total score was strategies are needed to collect data—ideally on a con- tinuous basis—that better describes ICU recovery on the Table 2 Test performance characteristics for personal fitness wards and in the patient’s home environment. tracker detection of tachycardia, as compared to SPO2-R To this end, we undertook an observational study to de- Sinus rhythm Atrial fibrillation termine the feasibility of using a commercial-grade wear- Sensitivity 0.695 0.516 able device to monitor recovery after critical illness. Overall, the device was well tolerated and captured the Specificity 0.988 0.995 vast majority of available data. For the detection of tachy- Positive predictive value 0.948 0.983 cardia, we found the wearable delivered high specificity Negative predictive value 0.914 0.804 and positive predictive value, but only low to moderate Accuracy 0.92 0.836 sensitivity. Much of the undercounting of fast heart rates Kroll et al. Journal of Intensive Care (2017) 5:64 Page 5 of 8 Fig. 2 Accuracy of wearable-derived heart rates for the detection of tachycardia (HR > 100) or bradycardia (HR < 50) as determined by SPO2 heart rates. The SPO2-derived values (dark gray) are shown sorted from lowest to highest heart rate. The corresponding wearable-derived heart rate is shown in either light gray (correct classification), green (false positive), or red (false negative). The majority of misclassified heart rates are false negatives for the detection of tachycardia. Some misclassification is due to wearable device readings of “0,” reflecting data not captured by the device by the wearable device was seen in patients who were not did not appear to degrade over time. The wearables stud- in sinus rhythm during at least some portion of the moni- ied did not appear to be a significant source of nosocomial toring period. Compared to a validated sleep question- pathogens, although the presence of Bacillus species even naire, the wearable device had a moderate correlation after device cleaning suggests that spore-forming organ- with several metrics of sleep quality. Device performance isms could persist on some devices. Whether or not wear- ables would have to be reused at all would depend on their costs—which currently are relatively low—compared Table 3 Summary of wearable-reported and RCSQ sleep to the potential cost savings achieved with better clinical parameters outcomes. The use of wearables to monitor convalescence Median (IQR) after ICU discharge will ultimately pertain to patients who Wearable no longer require the resources of heavily monitored set- Total sleep duration, hours 6.6 (2.7–13.5) tings. To that end, our results are generalizable to a large Asleep time, hours 6.1 (2.6–12.5) contingent of patients, including post-ICU patients cared for on the wards, as well as those who have been dis- Restless count 7 (2.5–19.0) charged home. Sleep quality A 45.8 (38.0–63.5) In addition to their potential use following an ICU ad- # Sleep periods 2 (1.0–4.0) mission, wearables may also play a role in monitoring in- 22:00–6:00 sleep as % of total 50% (15–80%) patients for signs of clinical deterioration, so as to % of 22:00–6:00 asleep 48% (3–84%) identify as soon as possible any patient needing a higher RCSQ level of care. Early Warning Systems (EWS) have been developed to address a “failure to rescue” problem, in Mean score 5.7 (2.7–8.0) which critical illness is identified too late [21]. Wearable 1. Sleep depth 5 (3.2–7.6) devices stand to enhance data collection and monitoring 2. Sleep latency 6.2 (2.7–8.9) both prior to and following an ICU admission, and as 3. Awakening 5 (2.6–8.6) such is of growing importance in critical care research. 4. Returning to sleep 6.4 (2.1–9.1) Interest in the clinical use of wearable devices and mo- 5. Sleep quality 5.7 (1.6–8.6) bile health technology is increasing [2, 22]. While clin- RCSQ Richards-Campbell Sleep Questionnaire ical evaluations of this technology remain scarce, some Kroll et al. Journal of Intensive Care (2017) 5:64 Page 6 of 8 Fig. 3 Correlation between mean score on the Richards-Campbell Sleep Questionnaire (RCSQ) and wearable-derived measure of the number of minutes asleep overnight (between 22:00 and 06:00). The Pearson correlation coefficient was 0.33 (95% CI 0.04 - 0.58) rigorous evaluations have been reported among healthy of the PPG-based sensing mechanism employed, which volunteers [4, 5] and among outpatients [23]. To our may perform poorly in patients with a pulse deficit, such knowledge, this is the first study to examine the feasibil- as those in atrial fibrillation. ity of using commercially available wearable devices Hospitalized patients often have a severely disrupted among hospital inpatients to evaluate for heart rate de- sleep, which may impair recovery [12]. Illness, medica- rangements and sleep quality. tions, around-the-clock care activities, and environmental Wearables have the potential to become a useful tool light and noise may contribute to perturbed sleep. in the early detection of critical illness. Heart rate is fac- Consumer-grade wearables with sleep monitoring capabil- tored into the majority of EWS algorithms [24–28], and ities could facilitate the routine evaluation of sleep among while the role of an EWS in reducing mortality remains inpatients and the assessment of sleep-promoting inter- unclear, there is evidence to suggest that these systems ventions. Resource-intensive polysomnography (PSG) is may be helpful [24]. Changes in heart rate may also por- impractical for routine sleep monitoring, and compliance tend changes in clinical status among ICU survivors on with sleep questionnaires and sleep diaries is poor among the wards or following hospital discharge. In this study, inpatients [29]. Continuous data collection from wearables the high specificity but low to moderate sensitivity iden- is passive and unobtrusive, and wearables are far less ex- tified for the detection of tachycardia suggests that as pensive than both PSG equipment and standard actigra- currently configured, wearable-derived heart rate track- phy devices. ing would be highly specific, thereby mitigating alarm fa- Two recent studies have compared commercial-grade tigue, but may lack sensitivity in some situations, wearables with PSG in healthy subjects [10, 11]. Mantua et resulting in missed detection of heart rate excursions. al. found a strong correlation in total sleep time between Ultimately, further confirmatory studies are required, wearable-derived data and PSG, and De Zambotti et al. which should also investigate alternate approaches to found good agreement between wearables and PSG in event detection, such as those based on proportional measuring sleep, despite slight but significant overesti- changes in heart rate. One potential limitation of mation of total sleep by the wearable devices. Altered sleep wearable-enabled heart rate monitoring is a direct result and activity patterns among inpatients may decrease the Kroll et al. Journal of Intensive Care (2017) 5:64 Page 7 of 8 accuracy of wearables, which rely on movement to deter- Abbreviations CAM: Confusion Assessment Method; ECG: Electrocardiography; EWS: Early mine wakefulness, and could overestimate sleep in inpa- WarningSystem; HR:Heart rate;ICU:Intensive care unit;MRSA: Methicillin-resistant tients, who may be awake but immobile for long periods. Staphylococcus aureus; PPG: Photoplethysmography; PSG: Polysomnography; The wearable device used in our study only counts periods RCSQ: Richards-Campbell Sleep Questionnaire; SPO2-R: Pulse oximetry pulse rate recordings (SPO2-R) of inactivity that exceed one hour as sleep, and may not capture fragmented naps, which are common in critically ill Acknowledgements patients [30, 31]. The authors wish to thank Miranda Hunt, Ilinca Georgescu, Tracy Boyd, and the nursing staff in the ICU at Kingston General Hospital. Our study has a number of limitations that should be considered in interpreting the results. Conclusions re- Funding garding the influence of non-sinus rhythm on the accur- This work was supported by an Innovation Fund award from the Southeastern Ontario Academic Medical Organization (SEAMO). Drs. Maslove acy of heart rate monitoring are limited by the relatively and Boyd are supported by New Clinician Scientists awards from SEAMO. low prevalence of this condition in the study cohort, as are the findings relating sleep with delirium, which also Availability of data and materials had a low prevalence. While we considered the absence All primary data are available upon written request. Requests are subject to ethics approval and data sharing agreements. of sleep quality measures reported to indicate an absence of sleep during the monitoring period, an alternate inter- Authors’ contributions pretation is that these conditions reflect a failure of data RRK participated in primary data collection and analysis, data interpretation, and contributed to the drafting of the manuscript. EDM contributed to the capture. It is worth noting, however, that for the cases analysis and interpretation of data and drafting of the manuscript. JGB included that recorded no sleep data, heart rate data was contributed to the study concept and design, data analysis and interpretation, successfully collected, making a failure of data capture and revising of the manuscript. PS conducted the microbiologic sampling sub-study and contributed to the interpretation of results. DH contributed to an unlikely explanation for these findings. Lastly, differ- the study concept and interpretation of results. MDW contributed to data ences between the internal clocks of the wearables and collection and analysis and interpretation of results. DMM contributed to the bedside monitors may have resulted in asynchronous study concept and design, data collection, data analysis, interpretation of results, and drafting of the manuscript. All authors read and approved the final heart rate recordings being treated as simultaneous, al- manuscript. though correction factors were used in the analysis, and the time differences observed were shorter than the Ethics approval and consent to participate All participating patients, or substitute decision makers on their behalf, 5 min sampling interval of the wearable device. provided written informed consent for participation in this study. The Health Sciences Research Ethics Board at Queen’s University reviewed and approved Conclusions the study protocol, and the trial was registered with clinicaltrials.gov (NCT02527408). In this observational study, we compared heart rate and sleep data recorded from a commercial-grade wearable Consent for publication device, with data from cardiac telemetry and sleep ques- The patient depicted in Fig. 1 provided consent for the photo to be published in this manuscript. tionnaires. Devices showed high specificity and moderate sensitivity for the detection of tachycardia, with better per- Competing interests formance in patients in sinus rhythm. Sleep quality met- The authors declare that they have no competing interests. rics were moderately correlated with questionnaire data. Future research in this area should focus on improving Publisher’sNote tachycardia detection, evaluating patients on the wards Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. and at home, integrating wearable-derived data into the study of ICU recovery, and determining the impact of Author details integrating wearable devices into hospital-wide EWS or Department of Critical Care Medicine, Queen’s University and Kingston Health Sciences Centre, Kingston, Ontario, Canada. School of Medicine, rapid response services. Patients with arrhythmias Queen’s University, Kingston, Ontario, Canada. Department of Medicine, should be studied as a subgroup in order to better define Queen’s University and Kingston Health Sciences Centre, Kingston, Ontario, the accuracy of wearable-based heart rate sensing in this Canada. Department of Pathology and Molecular Medicine, Queen’s University and Health Sciences Centre, Kingston, Ontario, Canada. population. Further validation of sleep quality accuracy Department of Emergency Medicine, Queen’s University and Kingston using other comparators such as PSG or conventional Health Sciences Centre, Kingston, Ontario, Canada. Department of actigraphy would be useful, as would assessments of the Neuroscience, Queen’s University, Kingston, Ontario, Canada. Kingston Health Sciences Centre, Kingston General Hospital, Davies 2, 76 Stuart St., accuracy of activity tracking. Kingston, Ontario K7L 2V7, Canada. Additional file Received: 19 September 2017 Accepted: 8 November 2017 Additional file 1: Details regarding microbiologic assessment of the References wearable devices as well as specifics regarding wearble device data 1. Salah H, MacIntosh E, Rajakulendran N. Wearable tech: leveraging canadian capture and analysis. (PDF 49 kb) innovation to improve health: MaRS Market Insights; 2014 Mar 26. p. 1–45. Kroll et al. Journal of Intensive Care (2017) 5:64 Page 8 of 8 Available from https://www.marsdd.com/wp-content/uploads/2015/02/ 25. Churpek MM, Yuen TC, Park SY, Meltzer DO, Hall JB, Edelson DP. Derivation MaRSReport-WearableTech.pdf of a cardiac arrest prediction model using ward vital signs. Crit Care Med. 2. Savage N. Mobile data: made to measure. Nature. 2015;527:S12–3. 2012;40(7):2102–8. 26. Churpek MM, Yuen TC, Winslow C, Meltzer DO, Kattan MW, Edelson DP. 3. El-Amrawy F, Nounou MI. Are currently available wearable devices for Multicenter comparison of machine learning methods and conventional activity tracking and heart rate monitoring accurate, precise, and medically regression for predicting clinical deterioration on the wards. Crit Care Med. beneficial? Healthc Inform Res. 2015;21(4):315–6. 2016;44(2):368–74. 4. Case MA, Burwick HA, Volpp KG, Patel MS. Accuracy of smartphone 27. Churpek MM, Yuen TC, Park SY, Gibbons R, Edelson DP. Using electronic applications and wearable devices for tracking physical activity data. JAMA-J health record data to develop and validate a prediction model for adverse Am Med Assoc. 2015;313(6):625–6. outcomes in the wards. Crit Care Med. 2014;42(4):841. 5. Wang R, Blackburn G, Desai M, Phelan D, Gillinov L, Houghtaling P, et al. 28. Churpek MM, Yuen TC, Winslow C, Hall J, Edelson DP. Differences in vital Accuracy of wrist-worn heart rate monitors. JAMA Cardiol. 2017;2(1):104–6. signs between elderly and nonelderly patients prior to ward cardiac arrest. 6. Asada HH, Shaltis P, Reisner A, Rhee S. Mobile monitoring with Crit Care Med. 2015;43(4):816–22. wearable photoplethysmographic biosensors. IEEE Eng Med Biol Mag. 29. Bano M, Chiaromanni F, Corrias M, Turco M, De Rui M, Amodio P, et al. The 2003;22(3):28–40. influence of environmental factors on sleep quality in hospitalized medical 7. Tamura T, Maeda Y, Sekine M, Yoshida M. Wearable photoplethysmographic patients. Front Neurol. 2014;5(4):267. sensors—past and present. Electronics. 2014;3(2):282–302. 30. Fitbit Inc. Fitbit help: how do I track my sleep? [internet]. help.fitbit.com. 8. Pantelopoulos A, Bourbakis NG. A survey on wearable sensor-based systems [accessed 2016 Oct 15]. Available from: https://help.fitbit.com/articles/en_ for health monitoring and prognosis. IEEE Trans Syst, Man, Cybern C. 2009; US/Help_article/1314 40(1):1–12. 31. Elliott R, McKinley S, Cistulli P, Fien M. Characterisation of sleep in intensive 9. Di Rienzo M, Rizzo F, Parati G, Brambilla G, Ferratini M, Castiglioni P. MagIC care using 24-hour polysomnography: an observational study. Crit Care. system: a new textile-based wearable device for biological signal 2013;17(2):R46. monitoring. Applicability in daily life and clinical setting. In: Engineering in medicine and biology society, 2005. IEEE-EMBS 2005. 27th annual international conference of the 2005 (pp. 7167–7169). Shanghai: IEEE. 10. de Zambotti M, Baker FC, Willoughby AR, Godino JG, Wing D, Patrick K, et al. Measures of sleep and cardiac functioning during sleep using a multi- sensory commercially-available wristband in adolescents. Physiol Behav. 2016;158:143–9. 11. Mantua J, Gravel N, Spencer R. Reliability of sleep measures from four personal health monitoring devices compared to research-based actigraphy and polysomnography. Sensors. 2016;16(5):646. 12. Young JS, Bourgeois JA, Hilty DM, Hardin KA. Sleep in hospitalized medical patients, part 1: factors affecting sleep. J Hosp Med. 2008;3(6):473–82. 13. Kamdar BB, Needham DM, Collop NA. Sleep deprivation in critical illness: its role in physical and psychological recovery. J Intensive Care Med. 2012; 27(2):97–111. 14. Litton E, Carnegie V, Elliott R, Webb SAR. The efficacy of earplugs as a sleep hygiene strategy for reducing delirium in the ICU: a systematic review and meta-analysis. Crit Care Med. 2016;44(5):992–9. 15. Richards KC, O’Sullivan PS, Phillips RL. Measurement of sleep in critically ill patients. J Nurs Meas. 2000;8(2):131–44. 16. Kamdar BB, Shah PA, King LM, Kho ME, Zhou X, Colantuoni E, et al. Patient- nurse interrater reliability and agreement of the Richards-Campbell Sleep Questionnaire. Am J Crit Care. 2012;21(4):261–9. 17. Flannery AH, Oyler DR, Weinhouse GL. The impact of interventions to improve sleep on delirium in the ICU: a systematic review and research framework. Crit Care Med. 2016;44(12):2231–40. 18. Kroll RR, Boyd JG, Maslove DM. Accuracy of a wrist-worn wearable device for monitoring heart rates in hospital inpatients: a prospective observational study. J Med Internet Res. 2016;18(9):e253. 19. Herridge MS, Chu LM, Matte A, Tomlinson G, Chan L, Thomas C, et al. The RECOVER program: disability risk groups and 1-year outcome after 7 or more days of mechanical ventilation. Am J Crit Care Med. 2016;194(7):831–44. 20. McNelly AS, Rawal J, Shrikrishna D, Hopkinson NS, Moxham J, Harridge SD, et al. An exploratory study of long-term outcome measures in critical illness survivors: construct validity of physical activity, frailty, and health-related quality of life measures. Crit Care Med. 2016;44(6):e362–9. Submit your next manuscript to BioMed Central 21. Roney JK, Whitley BE, Maples JC, Futrell LS, Stunkard KA, Long JD. Modified early warning scoring (MEWS): evaluating the evidence for tool inclusion of and we will help you at every step: sepsis screening criteria and impact on mortality and failure to rescue. J Clin • We accept pre-submission inquiries Nurs. 2015;24(23–24):3343–54. 22. Semple JL, Armstrong KA. Mobile applications for postoperative monitoring � Our selector tool helps you to find the most relevant journal after discharge. Can Med Assoc J. 2016. electronically published ahead of � We provide round the clock customer support print; https://doi.org/10.1503/cmaj.160195. � Convenient online submission 23. Jakicic JM, Davis KK, Rogers RJ, King WC, Marcus MD, Helsel D, et al. Effect of wearable technology combined with a lifestyle intervention on long-term � Thorough peer review weight loss: the IDEA randomized clinical trial. JAMA. 2016;316(11):1161–71. � Inclusion in PubMed and all major indexing services 24. Smith MEB, Chiovaro JC, O'Neil M, Kansagara D, Quiñones AR, Freeman M, � Maximum visibility for your research et al. Early warning system scores for clinical deterioration in hospitalized patients: a systematic review. Ann Am Thorac Soc. 2014;11(9):1454–65. Submit your manuscript at www.biomedcentral.com/submit http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Journal of Intensive Care Springer Journals

Use of wearable devices for post-discharge monitoring of ICU patients: a feasibility study

Loading next page...
 
/lp/springer-journals/use-of-wearable-devices-for-post-discharge-monitoring-of-icu-patients-4c0Q42r0Uv

References (32)

Publisher
Springer Journals
Copyright
Copyright © 2017 by The Author(s).
Subject
Medicine & Public Health; Intensive / Critical Care Medicine
eISSN
2052-0492
DOI
10.1186/s40560-017-0261-9
Publisher site
See Article on Publisher Site

Abstract

Background: Wearable devices generate signals detecting activity, sleep, and heart rate, all of which could enable detailed and near-continuous characterization of recovery following critical illness. Methods: To determine the feasibility of using a wrist-worn personal fitness tracker among patients recovering from critical illness, we conducted a prospective observational study of a convenience sample of 50 stable ICU patients. We assessed device wearability, the extent of data capture, sensitivity and specificity for detecting heart rate excursions, and correlations with questionnaire-derived sleep quality measures. Results: Wearable devices were worn over a 24-h period, with excellent capture of data. While specificity for the detection of tachycardia was high (98.8%), sensitivity was low to moderate (69.5%). There was a moderate correlation between wearable-derived sleep duration and questionnaire-derived sleep quality (r = 0.33, P =0.03). Devices were well-tolerated and demonstrated no degradation in quality of data acquisition over time. Conclusions: Wefoundthat wearabledevices couldbewornbypatients recovering from critical illness and could generate useful data for the majority of patients with little adverse effect. Further development and study are needed to better define and enhance the role of wearables in the monitoring of post-ICU recovery. Trial registration: Clinicaltrials.gov, NCT02527408 Keywords: Wearable devices, Medical informatics, Mobile health technologies, Validation study, Critical care, Sleep quality, Heart rate monitoring Background devices generate data that could also be useful in charac- Consumer interest in personal health tracking has recently terizing convalescence from acute illness. These include increased, leading to an industry in wearable devices now photoplethysmography (PPG) sensors to detect heart rate valued at more than $9 billion worldwide [1]. With more [6, 7], as well as accelerometers to track activity and wearables in use than ever before, there has been growing movement [3, 8, 9]. enthusiasm for their potential to improve health care de- Frequent heart rate tracking has the potential to identify livery [2]. Current clinical uses for wearable devices are episodes of clinical deterioration early. Accelerometer data mostly limited to outpatient settings, with a focus on the could potentially be used to encourage mobilization, management of chronic diseases [3–5]. Newer generation objectively measure functional status, and track progress towards rehabilitation goals. Wrist-worn accelerometers * Correspondence: [email protected] have also been used to evaluate sleep quality in healthy Equal contributors 1 subjects [10, 11]. In the inpatient and intensive care unit Department of Critical Care Medicine, Queen’s University and Kingston Health Sciences Centre, Kingston, Ontario, Canada (ICU) settings, where poor sleep has been linked with Department of Medicine, Queen’s University and Kingston Health Sciences adverse outcomes [12, 13], data describing sleep quality Centre, Kingston, Ontario, Canada Full list of author information is available at the end of the article © The Author(s). 2017 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Kroll et al. Journal of Intensive Care (2017) 5:64 Page 2 of 8 may be useful in identifying targets for sleep-promoting analgesia. We specifically chose to study patients who interventions [14]. were still in the ICU, as this was the most practical way to There is little clinical evidence to inform the practice obtain gold standard measurements of heart rate and of using wearables in health care, most of which is fo- sleep quality, which would otherwise require the use of cused on chronic conditions. Newer consumer-grade Holter monitors and complex follow-up procedures. To wearables have been evaluated in only a handful of stud- reduce the potential risk of transmitting nosocomial infec- ies examining their accuracy among healthy volunteers tions, patients under contact precautions for methicillin- [3–5]. These studies have called for evaluations of this resistant Staphylococcus aureus (MRSA) and Clostridium technology among a wider range of patient populations. difficile infections were also excluded. We also excluded In this study, we examine the feasibility of using a patients at risk of vascular compromise of the arm on common consumer-grade wearable device to monitor which the wearable device was to be placed, such as pa- patients recovering from critical illness. We enrolled pa- tients with upper extremity deep venous thrombosis, per- tients who no longer required intensive care measures ipherally inserted central catheters, radial arterial lines, but remained in the ICU prior to ward transfer, in order dialysis fistulas, and severe upper extremity trauma. As to best approximate post-ICU settings like the general this was a feasibility study, a convenience sample of 50 wards, while still collecting gold standard data to valid- participants was recruited. ate device functionality. We report on a number of prac- tical considerations that could affect the deployment of Ethics, consent, and permissions wearables including overall wearability, completeness of All participating patients, or substitute decision makers data capture, device longevity, and risk of transmitting on their behalf, provided written informed consent for nosocomial infections. We also evaluated the accuracy participation in this study. The Health Sciences Research of wearables for measuring sleep quality and identifying Ethics Board at Queen’s University reviewed and approved changes in heart rate that might be clinically relevant. the study protocol (DMED-1818-15), and the trial was We hypothesized that patients recovering from critical registered with clinicaltrials.gov (NCT02527408). illness would be able to wear wrist-worn devices and that useful data could be collected from these with a Device moderate degree of accuracy. Participating patients wore the Fitbit Charge HR device (Fitbit, San Francisco, CA, USA) for a single 24-h period Methods (Fig. 1). The Fitbit Charge HR is a commercially avail- Patients and setting able wrist-worn wearable that records heart rate, steps, This prospective observational study was conducted in a and sleep quality. The study employed three size large 33-bed general medical-surgical/trauma ICU in southeast- wearable devices (15.7 to 19.3 cm wrist circumference) ern Ontario, between August 2015 and February 2016. and three size extra-large wearable devices (19.6 to Adult patients (age > 17) were included if they were 22.6 cm wrist circumference). In an effort to reduce receiving continuous cardiac and oxygen saturation the risk of potential iatrogenic infection, we used dis- monitoring, but were otherwise receiving ward-level infectant wipes to thoroughly clean wearables between treatment. Exclusion criteria included mechanical venti- uses. All devices were applied to participants by a lation, vasopressor support, and continuous sedation or study investigator or coordinator. ab Fig. 1 The Fitbit ChargeHRdeviceusedinthe study (a). The wearable device as worn by a patient on the inpatient ward following ICU discharge (b) Kroll et al. Journal of Intensive Care (2017) 5:64 Page 3 of 8 Data monitoring and capture wrist-worn wearables for heart rate tracking in We used continuous pulse oximetry pulse rate record- healthy volunteers [5]. In addition to basic descriptive ings (SPO2-R) as a comparison measure of heart rate statistics, we calculated the sensitivity and specificity of (HR) in order to evaluate the ability of wearables to de- the wearables for detecting tachycardia and bradycardia. tect both tachycardia (HR > 100 bpm) and bradycardia Based on the PPG mechanism of heart rate sensing (HR < 50 bpm). We used SPO2-R values as a compara- employed in consumer-grade wearables, we hypothesized tor as both SPO2-R and wearable device values reflect that the accuracy of wearable device heart rate tracking the pulse rate (rather than electrical heart rate), and be- may be different in patients not in sinus rhythm and fur- cause this is a widely accepted method of heart rate ther analyzed these patients as a subgroup. We calculated measurement. The wearables recorded heart rate values Pearson correlation coefficients between the various every 5 min, while the SPO2-R recorded heart rate wearable-derived measures of sleep quality and the RCSQ values every minute. Cardiac rhythm was assessed at the measures of sleep quality. Based on the mechanism of time of device application, and again at the time of re- sleep sensing, which relies on the absence of movement, moval, at which time data regarding sleep quality was we hypothesized that the accuracy of wearables for sleep also collected using the Richards-Campbell Sleep Ques- tracking may differ in patients with delirium, and further tionnaire (RCSQ) [15]. This survey uses a visual analog analyzed these patients as a subgroup. Statistical analyses scale to assess sleep depth, latency, awakenings, percent- for this study were performed using R (v 3.2.2). age of time awake, and overall quality of sleep. The RCSQ was completed either by the patients themselves or by their designated night shift nurse, a practice previ- Results ously shown to have slight to moderate agreement with Patients and device wearability self-assessment [16]. Due to the interaction between We enrolled a total of 50 patients between August 2015 sleep and delirium in the ICU [17], patients were and January 2016 (Table 1). The median wrist circumfer- screened for delirium by a trained researcher using the ence in our cohort was 18.6 cm (SD 1.9 cm), with 6 of confusion assessment method (CAM)-ICU at the time the 50 patients enrolled having moderate or severe of device application, and again at the time of device edema of the wrist at the time of device application. The removal. size large device was used for 23 patients (46%), while Wearable-reported sleep data included time of sleep on- the size extra-large was used for 27 patients (54%). set and awakening, sleep duration, minutes asleep, mi- While there were no patients for whom the wearable de- nutes awake, restless count, and a calculated measure of vice could not be fitted, the fit was noted to be very tight sleep quality. Overall sleep quality was taken as the aver- Table 1 Characteristics of patients included in the study (n =50) age across sleep episodes, weighted by the duration of Mean heart rate (bpm) 88.3 each sleep episode. The percentage of total sleep occurring Mean age (years) 64 during nighttime hours, which we defined as 22:00 to 06:00, and the percentage of nighttime hours spent asleep Patients (n = 50) % were calculated. For participants who had no Fitbit- Male 26 52 detected sleep over the recording period, a score of 0 was Female 24 48 given for all sleep parameters. Methods for obtaining Admission diagnosis wearable and SPO2-R data are reported elsewhere [18], Respiratory 12 24 and in the Supplementary Content (see Additional file 1). Sepsis 7 14 Surgical 7 14 Microbiological assessment Neurologic 11 22 We conducted microbiologic sampling of the wearables Trauma 3 6 used from a convenience subset of patients (n =16) in order to evaluate both the risk of transmitting nosocomial Cardiovascular 6 12 pathogens from repeated application of wearables to Medical 4 8 different patients, as well as the efficacy of our disinfection Sinus rhythm practices (see Additional file 1). At start of monitoring 43 86 At end of monitoring 42 84 Statistical analysis Personal fitness tracker size used In the absence of preliminary data to inform a sample size Large 23 46 calculation, we targeted an enrollment of 50 patients, a Extra large 27 54 cohort size equal to that used in a similar study of Kroll et al. Journal of Intensive Care (2017) 5:64 Page 4 of 8 in one patient, and very loose in two patients. Devices 5.7/10.0 (IQR 2.7–8.0/10.0). There was a moderate were adjusted only once at the time of application and correlation between wearable-derived sleep duration were not re-assessed by study personnel for the duration and total RCSQ score (r = 0.33, P = 0.03, 95% confi- of the 24-h recording period. No intravenous lines were dence interval [CI] 0.04, 0.58) (Fig. 3). The correlation re-sited in order to facilitate application, although hos- between the percentage of nighttime asleep, as re- pital identification wristbands had to be relocated in ported by the wearable device, and total RCSQ score some cases. No wearables required removal during the was 0.36 (P = 0.02, 95% CI 0.07, 0.60). The correlation monitoring period as a result of patient discomfort. between the Fitbit-reported number of sleep periods The wearable device was removed prior to the com- and RCSQ-reported awakenings was 0.38 (P = 0.01, pletion of the monitoring period in two patients; one 95% CI 0.09, 0.61). There were no significant differ- patient was discharged earlier than expected from the ences in wearable-reported sleep parameters between ICU, while another developed a diffuse drug- the CAM-ICU positive (n = 8) and CAM-ICU negative associated rash. Excluding patients whose devices participants; however, 25% of CAM-ICU positive par- were removed early, the devices were unable to detect ticipants recorded no sleep over the entire 24-h mon- a heart rate reading 4% of the time. itoring period, compared to 8% of CAM-ICU negative participants. Tachycardia and bradycardia detection Device reusability We identified 13 SPO2-R-confirmed readings of brady- Wearables were not found to be a significant source cardia among four patients, all of whom were in sinus of pathogenic bacteria. Microbiologic sampling re- rhythm. Further statistical analysis was not done due to vealed bacteria consistent with commensal skin flora this small sample. The wearable had a sensitivity of (Staphylococcus epidermidis) and/or environmental or- 69.5% and specificity of 98.8% for the detection of tachy- ganisms (Bacillus species). S. epidermidis was only cardia (Table 2 and Fig. 2). Among patients not in sinus observed in samples taken prior to hydrogen peroxide rhythm (n = 8), the specificity for detecting tachycardia disinfection, while Bacillus species were found in both was similar (99.5%), although sensitivity was worse pre- and post-disinfection specimens. Individual wear- (51.6%). For faster heart rates (> 150 bpm), wearable de- able devices were used between 5 and 13 times. vice concordance with SPO2-R was poor. However, in There were no differences in wearable-SPO2-R heart many such cases, the wearable device reading showed rate correlations between the first and second half of better agreement with the true heart rate measured by the study (P = 0.18). continuous ECG, than did the SPO2-R readings, which tended to be falsely high. Discussion The long-term adverse consequences of critical illness Sleep data are increasingly being recognized as a research priority A summary of the sleep quality data collected by the in critical care [19]. A growing body of research is now wearables is shown in Table 3. Among the 47 partici- examining the determinants and potential modifiers of pants who had complete wearable sleep data re- post-ICU recovery, including at least one study that corded, the median wearable-reported sleep duration made use of a wearable device to track patient move- was 6.6 h (interquartile range [IQR] 2.7–13.5 h) and ment and activity [20]. However, post-ICU recovery re- the median number of sleep periods was 2 (IQR 1–4). search currently lacks the richness of data available to Five participants (11%) had no wearable device docu- researchers focused on the ICU stay itself since post- mented sleep for the entirety of the 24-h monitoring discharge data collection is limited to infrequent visits to period. Among the 43 participants for whom the follow-up clinics, or in many cases is nonexistent. New RCSQ was completed, the median total score was strategies are needed to collect data—ideally on a con- tinuous basis—that better describes ICU recovery on the Table 2 Test performance characteristics for personal fitness wards and in the patient’s home environment. tracker detection of tachycardia, as compared to SPO2-R To this end, we undertook an observational study to de- Sinus rhythm Atrial fibrillation termine the feasibility of using a commercial-grade wear- Sensitivity 0.695 0.516 able device to monitor recovery after critical illness. Overall, the device was well tolerated and captured the Specificity 0.988 0.995 vast majority of available data. For the detection of tachy- Positive predictive value 0.948 0.983 cardia, we found the wearable delivered high specificity Negative predictive value 0.914 0.804 and positive predictive value, but only low to moderate Accuracy 0.92 0.836 sensitivity. Much of the undercounting of fast heart rates Kroll et al. Journal of Intensive Care (2017) 5:64 Page 5 of 8 Fig. 2 Accuracy of wearable-derived heart rates for the detection of tachycardia (HR > 100) or bradycardia (HR < 50) as determined by SPO2 heart rates. The SPO2-derived values (dark gray) are shown sorted from lowest to highest heart rate. The corresponding wearable-derived heart rate is shown in either light gray (correct classification), green (false positive), or red (false negative). The majority of misclassified heart rates are false negatives for the detection of tachycardia. Some misclassification is due to wearable device readings of “0,” reflecting data not captured by the device by the wearable device was seen in patients who were not did not appear to degrade over time. The wearables stud- in sinus rhythm during at least some portion of the moni- ied did not appear to be a significant source of nosocomial toring period. Compared to a validated sleep question- pathogens, although the presence of Bacillus species even naire, the wearable device had a moderate correlation after device cleaning suggests that spore-forming organ- with several metrics of sleep quality. Device performance isms could persist on some devices. Whether or not wear- ables would have to be reused at all would depend on their costs—which currently are relatively low—compared Table 3 Summary of wearable-reported and RCSQ sleep to the potential cost savings achieved with better clinical parameters outcomes. The use of wearables to monitor convalescence Median (IQR) after ICU discharge will ultimately pertain to patients who Wearable no longer require the resources of heavily monitored set- Total sleep duration, hours 6.6 (2.7–13.5) tings. To that end, our results are generalizable to a large Asleep time, hours 6.1 (2.6–12.5) contingent of patients, including post-ICU patients cared for on the wards, as well as those who have been dis- Restless count 7 (2.5–19.0) charged home. Sleep quality A 45.8 (38.0–63.5) In addition to their potential use following an ICU ad- # Sleep periods 2 (1.0–4.0) mission, wearables may also play a role in monitoring in- 22:00–6:00 sleep as % of total 50% (15–80%) patients for signs of clinical deterioration, so as to % of 22:00–6:00 asleep 48% (3–84%) identify as soon as possible any patient needing a higher RCSQ level of care. Early Warning Systems (EWS) have been developed to address a “failure to rescue” problem, in Mean score 5.7 (2.7–8.0) which critical illness is identified too late [21]. Wearable 1. Sleep depth 5 (3.2–7.6) devices stand to enhance data collection and monitoring 2. Sleep latency 6.2 (2.7–8.9) both prior to and following an ICU admission, and as 3. Awakening 5 (2.6–8.6) such is of growing importance in critical care research. 4. Returning to sleep 6.4 (2.1–9.1) Interest in the clinical use of wearable devices and mo- 5. Sleep quality 5.7 (1.6–8.6) bile health technology is increasing [2, 22]. While clin- RCSQ Richards-Campbell Sleep Questionnaire ical evaluations of this technology remain scarce, some Kroll et al. Journal of Intensive Care (2017) 5:64 Page 6 of 8 Fig. 3 Correlation between mean score on the Richards-Campbell Sleep Questionnaire (RCSQ) and wearable-derived measure of the number of minutes asleep overnight (between 22:00 and 06:00). The Pearson correlation coefficient was 0.33 (95% CI 0.04 - 0.58) rigorous evaluations have been reported among healthy of the PPG-based sensing mechanism employed, which volunteers [4, 5] and among outpatients [23]. To our may perform poorly in patients with a pulse deficit, such knowledge, this is the first study to examine the feasibil- as those in atrial fibrillation. ity of using commercially available wearable devices Hospitalized patients often have a severely disrupted among hospital inpatients to evaluate for heart rate de- sleep, which may impair recovery [12]. Illness, medica- rangements and sleep quality. tions, around-the-clock care activities, and environmental Wearables have the potential to become a useful tool light and noise may contribute to perturbed sleep. in the early detection of critical illness. Heart rate is fac- Consumer-grade wearables with sleep monitoring capabil- tored into the majority of EWS algorithms [24–28], and ities could facilitate the routine evaluation of sleep among while the role of an EWS in reducing mortality remains inpatients and the assessment of sleep-promoting inter- unclear, there is evidence to suggest that these systems ventions. Resource-intensive polysomnography (PSG) is may be helpful [24]. Changes in heart rate may also por- impractical for routine sleep monitoring, and compliance tend changes in clinical status among ICU survivors on with sleep questionnaires and sleep diaries is poor among the wards or following hospital discharge. In this study, inpatients [29]. Continuous data collection from wearables the high specificity but low to moderate sensitivity iden- is passive and unobtrusive, and wearables are far less ex- tified for the detection of tachycardia suggests that as pensive than both PSG equipment and standard actigra- currently configured, wearable-derived heart rate track- phy devices. ing would be highly specific, thereby mitigating alarm fa- Two recent studies have compared commercial-grade tigue, but may lack sensitivity in some situations, wearables with PSG in healthy subjects [10, 11]. Mantua et resulting in missed detection of heart rate excursions. al. found a strong correlation in total sleep time between Ultimately, further confirmatory studies are required, wearable-derived data and PSG, and De Zambotti et al. which should also investigate alternate approaches to found good agreement between wearables and PSG in event detection, such as those based on proportional measuring sleep, despite slight but significant overesti- changes in heart rate. One potential limitation of mation of total sleep by the wearable devices. Altered sleep wearable-enabled heart rate monitoring is a direct result and activity patterns among inpatients may decrease the Kroll et al. Journal of Intensive Care (2017) 5:64 Page 7 of 8 accuracy of wearables, which rely on movement to deter- Abbreviations CAM: Confusion Assessment Method; ECG: Electrocardiography; EWS: Early mine wakefulness, and could overestimate sleep in inpa- WarningSystem; HR:Heart rate;ICU:Intensive care unit;MRSA: Methicillin-resistant tients, who may be awake but immobile for long periods. Staphylococcus aureus; PPG: Photoplethysmography; PSG: Polysomnography; The wearable device used in our study only counts periods RCSQ: Richards-Campbell Sleep Questionnaire; SPO2-R: Pulse oximetry pulse rate recordings (SPO2-R) of inactivity that exceed one hour as sleep, and may not capture fragmented naps, which are common in critically ill Acknowledgements patients [30, 31]. The authors wish to thank Miranda Hunt, Ilinca Georgescu, Tracy Boyd, and the nursing staff in the ICU at Kingston General Hospital. Our study has a number of limitations that should be considered in interpreting the results. Conclusions re- Funding garding the influence of non-sinus rhythm on the accur- This work was supported by an Innovation Fund award from the Southeastern Ontario Academic Medical Organization (SEAMO). Drs. Maslove acy of heart rate monitoring are limited by the relatively and Boyd are supported by New Clinician Scientists awards from SEAMO. low prevalence of this condition in the study cohort, as are the findings relating sleep with delirium, which also Availability of data and materials had a low prevalence. While we considered the absence All primary data are available upon written request. Requests are subject to ethics approval and data sharing agreements. of sleep quality measures reported to indicate an absence of sleep during the monitoring period, an alternate inter- Authors’ contributions pretation is that these conditions reflect a failure of data RRK participated in primary data collection and analysis, data interpretation, and contributed to the drafting of the manuscript. EDM contributed to the capture. It is worth noting, however, that for the cases analysis and interpretation of data and drafting of the manuscript. JGB included that recorded no sleep data, heart rate data was contributed to the study concept and design, data analysis and interpretation, successfully collected, making a failure of data capture and revising of the manuscript. PS conducted the microbiologic sampling sub-study and contributed to the interpretation of results. DH contributed to an unlikely explanation for these findings. Lastly, differ- the study concept and interpretation of results. MDW contributed to data ences between the internal clocks of the wearables and collection and analysis and interpretation of results. DMM contributed to the bedside monitors may have resulted in asynchronous study concept and design, data collection, data analysis, interpretation of results, and drafting of the manuscript. All authors read and approved the final heart rate recordings being treated as simultaneous, al- manuscript. though correction factors were used in the analysis, and the time differences observed were shorter than the Ethics approval and consent to participate All participating patients, or substitute decision makers on their behalf, 5 min sampling interval of the wearable device. provided written informed consent for participation in this study. The Health Sciences Research Ethics Board at Queen’s University reviewed and approved Conclusions the study protocol, and the trial was registered with clinicaltrials.gov (NCT02527408). In this observational study, we compared heart rate and sleep data recorded from a commercial-grade wearable Consent for publication device, with data from cardiac telemetry and sleep ques- The patient depicted in Fig. 1 provided consent for the photo to be published in this manuscript. tionnaires. Devices showed high specificity and moderate sensitivity for the detection of tachycardia, with better per- Competing interests formance in patients in sinus rhythm. Sleep quality met- The authors declare that they have no competing interests. rics were moderately correlated with questionnaire data. Future research in this area should focus on improving Publisher’sNote tachycardia detection, evaluating patients on the wards Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. and at home, integrating wearable-derived data into the study of ICU recovery, and determining the impact of Author details integrating wearable devices into hospital-wide EWS or Department of Critical Care Medicine, Queen’s University and Kingston Health Sciences Centre, Kingston, Ontario, Canada. School of Medicine, rapid response services. Patients with arrhythmias Queen’s University, Kingston, Ontario, Canada. Department of Medicine, should be studied as a subgroup in order to better define Queen’s University and Kingston Health Sciences Centre, Kingston, Ontario, the accuracy of wearable-based heart rate sensing in this Canada. Department of Pathology and Molecular Medicine, Queen’s University and Health Sciences Centre, Kingston, Ontario, Canada. population. Further validation of sleep quality accuracy Department of Emergency Medicine, Queen’s University and Kingston using other comparators such as PSG or conventional Health Sciences Centre, Kingston, Ontario, Canada. Department of actigraphy would be useful, as would assessments of the Neuroscience, Queen’s University, Kingston, Ontario, Canada. Kingston Health Sciences Centre, Kingston General Hospital, Davies 2, 76 Stuart St., accuracy of activity tracking. Kingston, Ontario K7L 2V7, Canada. Additional file Received: 19 September 2017 Accepted: 8 November 2017 Additional file 1: Details regarding microbiologic assessment of the References wearable devices as well as specifics regarding wearble device data 1. Salah H, MacIntosh E, Rajakulendran N. Wearable tech: leveraging canadian capture and analysis. (PDF 49 kb) innovation to improve health: MaRS Market Insights; 2014 Mar 26. p. 1–45. Kroll et al. Journal of Intensive Care (2017) 5:64 Page 8 of 8 Available from https://www.marsdd.com/wp-content/uploads/2015/02/ 25. Churpek MM, Yuen TC, Park SY, Meltzer DO, Hall JB, Edelson DP. Derivation MaRSReport-WearableTech.pdf of a cardiac arrest prediction model using ward vital signs. Crit Care Med. 2. Savage N. Mobile data: made to measure. Nature. 2015;527:S12–3. 2012;40(7):2102–8. 26. Churpek MM, Yuen TC, Winslow C, Meltzer DO, Kattan MW, Edelson DP. 3. El-Amrawy F, Nounou MI. Are currently available wearable devices for Multicenter comparison of machine learning methods and conventional activity tracking and heart rate monitoring accurate, precise, and medically regression for predicting clinical deterioration on the wards. Crit Care Med. beneficial? Healthc Inform Res. 2015;21(4):315–6. 2016;44(2):368–74. 4. Case MA, Burwick HA, Volpp KG, Patel MS. Accuracy of smartphone 27. Churpek MM, Yuen TC, Park SY, Gibbons R, Edelson DP. Using electronic applications and wearable devices for tracking physical activity data. JAMA-J health record data to develop and validate a prediction model for adverse Am Med Assoc. 2015;313(6):625–6. outcomes in the wards. Crit Care Med. 2014;42(4):841. 5. Wang R, Blackburn G, Desai M, Phelan D, Gillinov L, Houghtaling P, et al. 28. Churpek MM, Yuen TC, Winslow C, Hall J, Edelson DP. Differences in vital Accuracy of wrist-worn heart rate monitors. JAMA Cardiol. 2017;2(1):104–6. signs between elderly and nonelderly patients prior to ward cardiac arrest. 6. Asada HH, Shaltis P, Reisner A, Rhee S. Mobile monitoring with Crit Care Med. 2015;43(4):816–22. wearable photoplethysmographic biosensors. IEEE Eng Med Biol Mag. 29. Bano M, Chiaromanni F, Corrias M, Turco M, De Rui M, Amodio P, et al. The 2003;22(3):28–40. influence of environmental factors on sleep quality in hospitalized medical 7. Tamura T, Maeda Y, Sekine M, Yoshida M. Wearable photoplethysmographic patients. Front Neurol. 2014;5(4):267. sensors—past and present. Electronics. 2014;3(2):282–302. 30. Fitbit Inc. Fitbit help: how do I track my sleep? [internet]. help.fitbit.com. 8. Pantelopoulos A, Bourbakis NG. A survey on wearable sensor-based systems [accessed 2016 Oct 15]. Available from: https://help.fitbit.com/articles/en_ for health monitoring and prognosis. IEEE Trans Syst, Man, Cybern C. 2009; US/Help_article/1314 40(1):1–12. 31. Elliott R, McKinley S, Cistulli P, Fien M. Characterisation of sleep in intensive 9. Di Rienzo M, Rizzo F, Parati G, Brambilla G, Ferratini M, Castiglioni P. MagIC care using 24-hour polysomnography: an observational study. Crit Care. system: a new textile-based wearable device for biological signal 2013;17(2):R46. monitoring. Applicability in daily life and clinical setting. In: Engineering in medicine and biology society, 2005. IEEE-EMBS 2005. 27th annual international conference of the 2005 (pp. 7167–7169). Shanghai: IEEE. 10. de Zambotti M, Baker FC, Willoughby AR, Godino JG, Wing D, Patrick K, et al. Measures of sleep and cardiac functioning during sleep using a multi- sensory commercially-available wristband in adolescents. Physiol Behav. 2016;158:143–9. 11. Mantua J, Gravel N, Spencer R. Reliability of sleep measures from four personal health monitoring devices compared to research-based actigraphy and polysomnography. Sensors. 2016;16(5):646. 12. Young JS, Bourgeois JA, Hilty DM, Hardin KA. Sleep in hospitalized medical patients, part 1: factors affecting sleep. J Hosp Med. 2008;3(6):473–82. 13. Kamdar BB, Needham DM, Collop NA. Sleep deprivation in critical illness: its role in physical and psychological recovery. J Intensive Care Med. 2012; 27(2):97–111. 14. Litton E, Carnegie V, Elliott R, Webb SAR. The efficacy of earplugs as a sleep hygiene strategy for reducing delirium in the ICU: a systematic review and meta-analysis. Crit Care Med. 2016;44(5):992–9. 15. Richards KC, O’Sullivan PS, Phillips RL. Measurement of sleep in critically ill patients. J Nurs Meas. 2000;8(2):131–44. 16. Kamdar BB, Shah PA, King LM, Kho ME, Zhou X, Colantuoni E, et al. Patient- nurse interrater reliability and agreement of the Richards-Campbell Sleep Questionnaire. Am J Crit Care. 2012;21(4):261–9. 17. Flannery AH, Oyler DR, Weinhouse GL. The impact of interventions to improve sleep on delirium in the ICU: a systematic review and research framework. Crit Care Med. 2016;44(12):2231–40. 18. Kroll RR, Boyd JG, Maslove DM. Accuracy of a wrist-worn wearable device for monitoring heart rates in hospital inpatients: a prospective observational study. J Med Internet Res. 2016;18(9):e253. 19. Herridge MS, Chu LM, Matte A, Tomlinson G, Chan L, Thomas C, et al. The RECOVER program: disability risk groups and 1-year outcome after 7 or more days of mechanical ventilation. Am J Crit Care Med. 2016;194(7):831–44. 20. McNelly AS, Rawal J, Shrikrishna D, Hopkinson NS, Moxham J, Harridge SD, et al. An exploratory study of long-term outcome measures in critical illness survivors: construct validity of physical activity, frailty, and health-related quality of life measures. Crit Care Med. 2016;44(6):e362–9. Submit your next manuscript to BioMed Central 21. Roney JK, Whitley BE, Maples JC, Futrell LS, Stunkard KA, Long JD. Modified early warning scoring (MEWS): evaluating the evidence for tool inclusion of and we will help you at every step: sepsis screening criteria and impact on mortality and failure to rescue. J Clin • We accept pre-submission inquiries Nurs. 2015;24(23–24):3343–54. 22. Semple JL, Armstrong KA. Mobile applications for postoperative monitoring � Our selector tool helps you to find the most relevant journal after discharge. Can Med Assoc J. 2016. electronically published ahead of � We provide round the clock customer support print; https://doi.org/10.1503/cmaj.160195. � Convenient online submission 23. Jakicic JM, Davis KK, Rogers RJ, King WC, Marcus MD, Helsel D, et al. Effect of wearable technology combined with a lifestyle intervention on long-term � Thorough peer review weight loss: the IDEA randomized clinical trial. JAMA. 2016;316(11):1161–71. � Inclusion in PubMed and all major indexing services 24. Smith MEB, Chiovaro JC, O'Neil M, Kansagara D, Quiñones AR, Freeman M, � Maximum visibility for your research et al. Early warning system scores for clinical deterioration in hospitalized patients: a systematic review. Ann Am Thorac Soc. 2014;11(9):1454–65. Submit your manuscript at www.biomedcentral.com/submit

Journal

Journal of Intensive CareSpringer Journals

Published: Nov 21, 2017

There are no references for this article.