Do material, psychosocial and behavioural factors mediate the relationship between disability acquisition and mental health? A sequential causal mediation analysis

Do material, psychosocial and behavioural factors mediate the relationship between disability... Abstract Background There is evidence of a causal relationship between disability acquisition and poor mental health; however, the mechanism by which disability affects mental health is poorly understood. This gap in understanding limits the development of effective interventions to improve the mental health of people with disabilities. Methods We used four waves of data from the Household, Income and Labour Dynamics in Australia Survey (2011–14) to compare self-reported mental health between individuals who acquired any disability (n=387) and those who remained disability-free (n=7936). We tested three possible pathways from disability acquisition to mental health, examining the effect of material, psychosocial and behavioural mediators. The effect was partitioned into natural direct and indirect effects through the mediators using a sequential causal mediation analysis approach. Multiple imputation using chained equations was used to assess the impact of missing data. Results Disability acquisition was estimated to cause a five-point decline in mental health [estimated mean difference: –5.3, 95% confidence interval (CI) –6.8, –3.7]. The indirect effect through material factors was estimated to be a 1.7-point difference (–1.7, 95% CI –2.8, –0.6), explaining 32% of the total effect, with a negligible proportion of the effect explained by the addition of psychosocial characteristics (material and psychosocial: –1.7, 95% CI –3.0, –0.5) and a further 5% by behavioural factors (material-psychosocial-behavioural: –2.0, 95% CI –3.4, –0.6). Conclusions The finding that the effect of disability acquisition on mental health operates predominantly through material rather than psychosocial and behavioural factors has important implications. The results highlight the need for better social protection, including income support, employment and education opportunities, and affordable housing for people who acquire a disability. disability, mental health, health inequalities, social epidemiology, causal mediation analysis Key Messages This paper investigated the mechanistic pathways linking disability acquisition and mental health using sequential causal mediation analysis to examining the effect of material, psychosocial and behavioural factors as mediators of the association. The total causal effect of disability acquisition on mental health was estimated to be a five-point decline in Mental Health Inventory (MHI) score. The effect was partially explained by the three sets of mediators, with 32% of the total effect mediated by material factors, a negligible proportion mediated by the addition of psychosocial factors and a further 5% by behavioural factors. The effect of disability acquisition on mental health operates predominantly through material factors, highlighting the need for better social protection, including income support, employment and education opportunities, and affordable housing for people who acquire a disability. Introduction Currently, more than a billion people, approximately 15% of the world’s population, live with a disability.1 People with disabilities experience substantial health inequalities and are at high risk of poor mental health.2,3 A causal relationship between disability and poor mental health has been suggested from analyses of prospective cohort studies.2–9 However, the mechanism by which disability leads to deterioration in mental helath is poorly understood. There are a number of different potential explanations for a causal link between disability and poor mental health. Supported by theoretical and empirical studies of the mechanisms underlying income-related health inqualities, three frameworks have become well established in explaining how social determinants influence health: material, psychosocial and behavioural pathways.10–13 First, the material pathway, by which differential exposure to structural and material living conditions leads to health inequalities, which posits that material conditions such as poverty and economic deprivation affect health directly, but also have indirect effects by enabling access to better living circumstances such as access to health care.14 Second, the psychosocial perspective emphasizes the importance of psychosocial and stress-related risk factors on health, with inequalities arising from the unequal distribution of psychosocial factors such as social support, home–work balance and personal control.15 Third, differences in health-related behaviour are thought to contribute to health inequalities, e.g. smoking, physical activity and diet.16 There has been considerable debate regarding the relative importance of these factors in explaining social inequalities in health.11 Most empirical studies have argued for the significance of material pathways,11,17 postulated to have a greater relative contribution because they exert both a direct effect on health as well as an indirect effect through psychosocial and behavioural pathways.12,18 With regard to the mechanisms driving disability-related mental health inequalities, disability acquisition may lead to changes in material, psychosocial and behavioural factors, which could explain, or mediate, the observed mental health deterioration. At present, it is not clear to what extent the effect of disability on mental health operates through these proposed pathways or through other mechanisms. Evidence regarding the importance of different pathways between disability and mental health is sparse; the research has mainly been conducted in cross-sectional studies of people with chronic illness, has only examined psychosocial pathways and no study has examined multiple pathways simultaneously. Three studies examined mediation of the effect through psychosocial resources and found evidence that some of the effect of disability acquisition on depressive symptoms6 and depression was operating through this pathway.6,19,20 Understanding the mechanisms underpinning these mental health inequalities is an important public health question because socio-economic intermediary variables are potential modifiable targets for interventions to mitigate the adverse effects of disability on people’s mental health.21 In this study, we use data from four waves of a longitudinal study of Australian adults and apply recently developed methods—sequential causal mediation analysis—to estimate the relative importance of three distinct mechanistic pathways leading from disability acquisition to poor mental health, quantifying the indirect effects through material, psychosocial and behavioural factors (Figure 1). Material factors are likely to affect mental health directly as well as indirectly through psychosocial factors such as latent consequences of employment (e.g. purposeful time use, self-esteem)22 and behavioural factors. Similarly, psychosocial factors are thought to exert a direct effect on mental health, and an indirect effect through behavioural factors. Figure 1 View largeDownload slide Casual diagram illustrating postulated causal relationships between disability acquisition and mental health. Figure 1 View largeDownload slide Casual diagram illustrating postulated causal relationships between disability acquisition and mental health. Methods Data source The Household, Income and Labour Dynamics in Australia (HILDA) Survey is a longitudinal study of Australian households, conducted annually since 2001.23 The survey collects information on demographic, social, economic and health characteristics of individuals using a combination of interviews and self-completed questionnaires. The original sample included 13 969 participants from 7682 households, randomly sampled using a national probability sample of private dwellings. A top-up sample was added in 2011 to maintain representativeness, leading to a sample size after 14 waves of 28 794 people. On average, for all waves of the survey, response proportions were 80% (ranging from 70% to 92%) and attrition was 5.7% between waves, ranging from 3.5% in 2014 to 13.2% in 2002. The analysis used four waves of the survey (2011 to 2014) to establish a temporal sequence between disability acquisition, the mediators and mental health. Disability acquisition Information on disability was collected in every wave, using a single question defining disability as ‘an impairment, disability or long-term health condition, which restricts everyday activities that had lasted for six months or more’. Disability acquisition was defined as two waves reporting no disability, followed immediately by two consecutive waves reporting a disability. We used two consecutive waves of disability so as to exclude people with transient disability and to reduce the potential for measurement error—a definition used in previous studies examining disability acquisition.24–27 Participants who acquired a disability were compared with those who reported no disability in any of the four waves. People who reported other patterns of exposure, such as a single wave of disability, were excluded. Eligibility for inclusion required participation and response to the disability question at all four waves. Mental health Mental health was assessed in the final wave (2014) using the Mental Health Inventory (MHI), a subscale of the Short Form 36 (SF-36, a widely used general health questionnaire that has been validated in the Australian population using data from the HILDA Survey).28 The MHI is a well-validated and reliable measure of mental health status.29 It measures symptoms of depression, anxiety and psychological wellbeing and has been shown to be an effective screening tool for mood and anxiety disorders and severe depressive symptomatology in comparisons with established mental health, wellbeing and depression scales30–33 as well as studies comparing against clinical diagnoses.34–37 The MHI has been shown to be psychometrically sound, with high internal consistency, discriminant validity and high test–retest reliability.38 It includes five items relating to mental health over the previous 4 weeks, each scored using five response categories. Total scores were transformed into a scale with a mean score of 74 (range: 0–100), as per standard practice, with higher scores reflecting better mental health. Previous research has suggested that a difference of four to five points on the MHI scale is likely to reflect a minimally important clinical difference in mental health.39,40 Mediators Mediator variables, described in Table 1, were measured in the third wave (2013). The choice of variables and their classification into three broad categories were motivated by reviewing empirical studies examining different explanations for socio-economic inequalities in health11,41–43 and selecting similar variables available in the HILDA Survey where possible. Table 1 Description of mediator variables Variables Type Definition/categorization Material factors  Occupation Categorical High skill; medium skill; low skill job; unemployed/not in labour force  Weekly income Continuous Equivalized household disposable income, $AUD  Financial hardship Categorical Prosperous/very comfortable; reasonably comfortable; just getting along/poor/very poor  Financial satisfaction Continuous Satisfaction with financial circumstances, ranked using an 11-point Likert scale ranging from ‘totally dissatisfied’ to ‘totally satisfied’  Housing tenure Categorical Outright owner; mortgager; renter  Housing affordability Binary Unaffordable defined as households in the lowest 40% of the income distribution with housing costs exceeding 30% of their gross income Psychosocial factors  Relationship status Binary Yes; no  Children Binary Yes; no  Social support Continuous Constructed using the average of 10 questions addressing aspects of emotional support, each rated on a seven-point Likert scale25  Socializing Continuous Frequency of socializing with friends or relatives, rated on a seven-point Likert scale ranging from daily to less than once every 3 months  Parent relationship Continuous Satisfaction with relationship with parents, rated on an 11-point Likert scale ranging from ‘completely dissatisfied’ to ‘completely satisfied’ Behavioural factors  Smoking Categorical Never; ex-smoker; current  Alcohol consumption Categorical Never; rarely; 1–2 days per week; >2 days per week  Physical activity Categorical >3 times per week; one to three times per week; less than once a week  Body mass index Continuous Self-reported, kg/m2  Healthy diet index Continuous Ranging from 0 ‘unhealthiest’ to 4 ‘healthiest’, derived from four binary questions: eating fruit every day; eating vegetables every day; eating fatty foods less than once a month; drinking low fat milk26  Quality of sleep Continuous Rated on a four-point Likert scale ranging from 1 ‘very good’ to 4 ‘very bad’ Variables Type Definition/categorization Material factors  Occupation Categorical High skill; medium skill; low skill job; unemployed/not in labour force  Weekly income Continuous Equivalized household disposable income, $AUD  Financial hardship Categorical Prosperous/very comfortable; reasonably comfortable; just getting along/poor/very poor  Financial satisfaction Continuous Satisfaction with financial circumstances, ranked using an 11-point Likert scale ranging from ‘totally dissatisfied’ to ‘totally satisfied’  Housing tenure Categorical Outright owner; mortgager; renter  Housing affordability Binary Unaffordable defined as households in the lowest 40% of the income distribution with housing costs exceeding 30% of their gross income Psychosocial factors  Relationship status Binary Yes; no  Children Binary Yes; no  Social support Continuous Constructed using the average of 10 questions addressing aspects of emotional support, each rated on a seven-point Likert scale25  Socializing Continuous Frequency of socializing with friends or relatives, rated on a seven-point Likert scale ranging from daily to less than once every 3 months  Parent relationship Continuous Satisfaction with relationship with parents, rated on an 11-point Likert scale ranging from ‘completely dissatisfied’ to ‘completely satisfied’ Behavioural factors  Smoking Categorical Never; ex-smoker; current  Alcohol consumption Categorical Never; rarely; 1–2 days per week; >2 days per week  Physical activity Categorical >3 times per week; one to three times per week; less than once a week  Body mass index Continuous Self-reported, kg/m2  Healthy diet index Continuous Ranging from 0 ‘unhealthiest’ to 4 ‘healthiest’, derived from four binary questions: eating fruit every day; eating vegetables every day; eating fatty foods less than once a month; drinking low fat milk26  Quality of sleep Continuous Rated on a four-point Likert scale ranging from 1 ‘very good’ to 4 ‘very bad’ Table 1 Description of mediator variables Variables Type Definition/categorization Material factors  Occupation Categorical High skill; medium skill; low skill job; unemployed/not in labour force  Weekly income Continuous Equivalized household disposable income, $AUD  Financial hardship Categorical Prosperous/very comfortable; reasonably comfortable; just getting along/poor/very poor  Financial satisfaction Continuous Satisfaction with financial circumstances, ranked using an 11-point Likert scale ranging from ‘totally dissatisfied’ to ‘totally satisfied’  Housing tenure Categorical Outright owner; mortgager; renter  Housing affordability Binary Unaffordable defined as households in the lowest 40% of the income distribution with housing costs exceeding 30% of their gross income Psychosocial factors  Relationship status Binary Yes; no  Children Binary Yes; no  Social support Continuous Constructed using the average of 10 questions addressing aspects of emotional support, each rated on a seven-point Likert scale25  Socializing Continuous Frequency of socializing with friends or relatives, rated on a seven-point Likert scale ranging from daily to less than once every 3 months  Parent relationship Continuous Satisfaction with relationship with parents, rated on an 11-point Likert scale ranging from ‘completely dissatisfied’ to ‘completely satisfied’ Behavioural factors  Smoking Categorical Never; ex-smoker; current  Alcohol consumption Categorical Never; rarely; 1–2 days per week; >2 days per week  Physical activity Categorical >3 times per week; one to three times per week; less than once a week  Body mass index Continuous Self-reported, kg/m2  Healthy diet index Continuous Ranging from 0 ‘unhealthiest’ to 4 ‘healthiest’, derived from four binary questions: eating fruit every day; eating vegetables every day; eating fatty foods less than once a month; drinking low fat milk26  Quality of sleep Continuous Rated on a four-point Likert scale ranging from 1 ‘very good’ to 4 ‘very bad’ Variables Type Definition/categorization Material factors  Occupation Categorical High skill; medium skill; low skill job; unemployed/not in labour force  Weekly income Continuous Equivalized household disposable income, $AUD  Financial hardship Categorical Prosperous/very comfortable; reasonably comfortable; just getting along/poor/very poor  Financial satisfaction Continuous Satisfaction with financial circumstances, ranked using an 11-point Likert scale ranging from ‘totally dissatisfied’ to ‘totally satisfied’  Housing tenure Categorical Outright owner; mortgager; renter  Housing affordability Binary Unaffordable defined as households in the lowest 40% of the income distribution with housing costs exceeding 30% of their gross income Psychosocial factors  Relationship status Binary Yes; no  Children Binary Yes; no  Social support Continuous Constructed using the average of 10 questions addressing aspects of emotional support, each rated on a seven-point Likert scale25  Socializing Continuous Frequency of socializing with friends or relatives, rated on a seven-point Likert scale ranging from daily to less than once every 3 months  Parent relationship Continuous Satisfaction with relationship with parents, rated on an 11-point Likert scale ranging from ‘completely dissatisfied’ to ‘completely satisfied’ Behavioural factors  Smoking Categorical Never; ex-smoker; current  Alcohol consumption Categorical Never; rarely; 1–2 days per week; >2 days per week  Physical activity Categorical >3 times per week; one to three times per week; less than once a week  Body mass index Continuous Self-reported, kg/m2  Healthy diet index Continuous Ranging from 0 ‘unhealthiest’ to 4 ‘healthiest’, derived from four binary questions: eating fruit every day; eating vegetables every day; eating fatty foods less than once a month; drinking low fat milk26  Quality of sleep Continuous Rated on a four-point Likert scale ranging from 1 ‘very good’ to 4 ‘very bad’ Baseline covariates Baseline covariates were measured in the first wave (2011), as a measure of people’s circumstances prior to disability acquisition. It is well documented in the literature that the incidence of disability is socially patterned, with people who experience socio-economic disadvantage being more likely to acquire a disability.25,44,45 Furthermore, according to the International Classification of Functioning, Disability and Health (ICF) framework, disability results from the interaction between health conditions, personal attributes and environmental factors.46,47 Conceived in this way, personal attributes such as the experience of financial strain, or characteristics of people’s social environment, such as their ability to access social support, are key determinants of disability as they influence the impact of people’s impairments on activity limitations and restriction to participation. Demographic characteristics included age, sex and country of birth (Australia; other) and socio-economic characteristics included education (bachelor’s degree and above; completion of secondary education; did not complete secondary) and parental occupation (high skill; medium skill; low skill or not in the labour force). Baseline levels of material, psychosocial and behavioural variables were recorded, categorized as described above, except for diet index and sleep quality, which were not measured in 2011. Mental health at baseline was measured using the MHI. Sequential causal mediation approach Mediation analysis aims to determine the extent to which an association between an exposure (here, incident disability) and an outcome (mental health) is due to the effect of the exposure on an intermediate variable (the mediator) which then influences the outcome. It aims to partition the total (causal) effect (TCE) of the exposure on the outcome into the effect that acts through the mediator, the indirect effect and the effect of exposure on outcome through mechanisms other than those that involve the mediator, the direct effect (‘direct’ in the sense that it by-passes the putative mediator). We sought to decompose the effect of disability acquisition on mental health into natural direct effects (NDE) and natural indirect effects (NIE) through material, psychosocial and behavioural factors using a sequential approach to causal mediation analysis (further details in Supplementary File 1, available as Supplementary Data at IJE online).48 This approach allows for mediation analysis through multiple causally related mediators and accommodates exposure–mediator interactions, one of the main sources of potential bias of the traditional approach to mediation. Based on our assumptions about the causal ordering of the mediators, this approach enabled us to estimate, in Model 1, the NIE through material factors (including paths that act through causal descendants of material factors but excluding paths that act only through psychosocial and/or behavioural factors), in Model 2, the NIE through both material and psychosocial factors (and through their causal descendants but excluding the path that acts only through behavioural factors) and, in Model 3, the NIE through material, psychosocial and behavioural factors, consisting of all possible paths except for the ‘direct’ path from exposure to outcome (Figure 2). Figure 2 View largeDownload slide Simplified causal diagrams illustrating estimated paths in Models 1–3, the NDE illustrated by the black lines (-) and the NIE by the dashed lines (- -) (A, disability acquisition (exposure of interest); Y, mental health (outcome); Mediators—M1, material factors; M2, psychosocial factors; M3, behavioural factors). Figure 2 View largeDownload slide Simplified causal diagrams illustrating estimated paths in Models 1–3, the NDE illustrated by the black lines (-) and the NIE by the dashed lines (- -) (A, disability acquisition (exposure of interest); Y, mental health (outcome); Mediators—M1, material factors; M2, psychosocial factors; M3, behavioural factors). Statistical analysis We used a weighting approach to estimate the marginal TCE, NDE and NIE for each set of mediators (further details in Supplementary File 1, available as Supplementary Data at IJE online). Inverse probability weighting was used to achieve exchangeability between the comparison groups and thus to account for possible confounding of the exposure–mediator and exposure–outcome associations by measured covariates.48–50 The MHI was modelled as a continuously valued outcome using linear regression models with and without the mediators, including all baseline variables as covariates. Interactions were included between the exposure and mediator variable if removal of an interaction term substantially changed the estimates of the NDE and NIE,51 measured as a change in the estimate of greater than half a standard error. Bootstrapping with 200 replications was used to calculate 95% confidence intervals (CIs). Missing data There were missing observations for the outcome, as well as several baseline covariates and mediators (Table S2.1, Supplementary File 2, available as Supplementary Data at IJE online). The distribution of baseline covariates was compared between participants with and without missing observations to determine whether missingness was associated with the values of measured variables. Participants with missing data had poorer mental health and greater socio-economic disadvantage across all measures compared with those with complete data (Table S2.2, Supplementary File 2, available as Supplementary Data at IJE online), suggesting that the data were not missing completely at random. Multiple imputation (MI) using chained equations with 50 imputations was performed to optimize the validity of the findings. The imputation models included all variables in the target analysis as well as additional auxiliary variables (Table S2.3, Supplementary File 2, available as Supplementary Data at IJE online). The sequential mediation analysis was conducted on each of the 50 imputed datasets and the mean of the estimates from each imputed dataset was calculated to give an overall MI estimate of the NDE and NIE. Standard errors were derived using Rubin rules for combining the between-imputation and within-imputation variance (obtained by bootstrapping the NDE and NIE estimates).52 Sensitivity analyses Three sensitivity analyses were conducted to test the robustness of findings. First, we performed a bias analysis for unmeasured confounding, which assessed the sensitivity of the results to unmeasured confounding of the mediator–outcome association, positing a range of plausible values for the strength of association of the potential confounder with mental health and the difference in prevalence of this confounder between those with and without disability (further details in Supplementary File 3, available as Supplementary Data at IJE online).53 Second, we removed participants with psychological impairments, defined as nervous or emotional conditions that require treatment, or any mental illness that requires help or supervision, as the effect of acquiring a psychological impairment on a general mental health score is likely to be different to other types of impairments. Third, we conducted a complete case analysis. Results Of the 28 794 people who participated in at least one wave of HILDA between 2001 and 2014, 14 534 participated in all four waves 2011 to 2014 and 14 518 of these (99.9%) responded to the disability question in all four waves. A total of 8323 individuals satisfied the definition of disability acquisition or reported no disability in any of the four waves, making them eligible for inclusion in the analysis (Figure S2.1, Supplementary File 2, available as Supplementary Data at IJE online). Complete data for all baseline covariates, mediators and mental health score were available for 4305 individuals (52% of the eligible sample). Baseline characteristics At baseline, people who went on to acquire a disability were older than those without disability (mean age of 53 vs 41 years, Table 2). They had poorer education, with 33.6% not completing secondary education compared with 24.9% of those without disability, were more likely to be unemployed or not in the labour force (38.8 vs 22.7%), had a lower mean weekly income (AU$834 vs AU$987) and experienced greater financial hardship (34.1 vs 25.1% reported being very poor or just getting by). People with disabilities were more likely to be in a relationship (71.3 vs 65.3%) and have children (72.9 vs 59.4%), more likely to be current (21.2 vs 17.2%) or ex-smokers (31.8 vs 23.7%), less likely to exercise regularly (34.2 vs 37.0%) and had higher mean BMI (27.4 vs 25.8 kg/m2). At baseline, they also reported poorer mental health than those without disability (mean MHI score of 73.3 vs 77.6). Table 2 Distribution of baseline characteristics for people who acquired a disability and the control sample (n=8323) Disability No disability n=387 n=7936 n % N % Age, years (mean (SD)) 387 52.5 (18.1) 7936 41.2 (15.4) Sex  Men 193 49.9 3817 48.1  Women 194 50.1 4119 51.9 Country of birth  Australia 297 76.7 6257 78.8  Other 90 23.3 1679 21.2 Parent occupation  High skill 181 47.4 3988 51.1  Medium skill 129 33.8 2645 33.9  Low skill/never worked 72 18.9 1168 15.0  Missing n=5 n=135 Education  Bachelor or higher 66 17.1 2184 27.5  Secondary, certificate, diploma 191 49.4 3777 47.6  Did not complete secondary 130 33.6 1975 24.9 Occupation  High skill 75 19.4 2373 29.9  Medium skill 97 25.1 2395 30.2  Low skill 65 16.8 1364 17.2  Unemployed/not in the labour force 150 38.8 1799 22.7  Missing n=0 n=5 Income, weekly $AUD (mean (SD)) 387 833.9 (476.8) 7936 986.9 (496.5) Wealth  High 138 35.7 2952 37.2  Medium 130 33.6 2616 33.0  Low 119 30.8 2368 29.8 Financial hardship  Prosperous/very comfortable 42 12.0 1444 20.3  Reasonably comfortable 188 53.9 3874 54.6  Just getting by/very poor 119 34.1 1781 25.1  Missing n=38 n=837 Financial satisfaction [mean (SD)]a 387 6.4 (2.3) 7931 6.7 (2.0)  Missing n=0 n=5 Housing tenure  Outright owner 148 38.2 2107 26.6  Mortgager 126 32.6 3393 42.8  Other 113 29.2 2425 30.6  Missing n=0 n=11 Housing affordability  Affordable 348 91.1 7263 92.4  Unaffordable 34 8.9 597 7.6  Missing n=5 n=76 Relationship  Yes 276 71.3 5173 65.3  No 111 28.7 2755 34.8  Missing n=0 n=8 Children  No 105 27.1 3224 40.6  Yes 282 72.9 4712 59.4 Social support [mean (SD)]b 347 5.3 (1.1) 7017 5.6 (1.0)  Missing n=40 n=919 Frequency of socializing [mean (SD)]c 347 3.8 (1.6) 7068 3.4 (1.4)  Missing n=40 n=868 Relationship with parents [mean (SD)]d 197 7.9 (2.2) 5833 8.1 (2.0)  Missing n=190 n=2103 Alcohol consumption  Never 59 16.9 1108 15.6  Rarely 129 36.9 2540 35.8  One or two times/week 62 17.7 1547 21.8  At least three times/week 100 28.6 1909 26.9  Missing n=37 n=832 Smoking  Never smoked 164 47.0 4210 59.2  Ex-smoker 111 31.8 1684 23.7  Current 74 21.2 1222 17.2  Missing n=38 n=820 Physical activity  At least four times/week 120 34.2 2638 37.0  One to three times/week 137 39.0 2998 42.1  Less than once/week 94 26.8 1493 20.9  Missing n=36 n=807 BMI, kg/m2 [mean (SD)] 332 27.4 (5.3) 6855 25.8 (5.0)  Missing n=55 n=1081 Mental health inventory (MHI) [mean (SD)]e 351 73.3 (18.9) 7125 77.6 (14.4)  Missing n=36 n=811 Disability No disability n=387 n=7936 n % N % Age, years (mean (SD)) 387 52.5 (18.1) 7936 41.2 (15.4) Sex  Men 193 49.9 3817 48.1  Women 194 50.1 4119 51.9 Country of birth  Australia 297 76.7 6257 78.8  Other 90 23.3 1679 21.2 Parent occupation  High skill 181 47.4 3988 51.1  Medium skill 129 33.8 2645 33.9  Low skill/never worked 72 18.9 1168 15.0  Missing n=5 n=135 Education  Bachelor or higher 66 17.1 2184 27.5  Secondary, certificate, diploma 191 49.4 3777 47.6  Did not complete secondary 130 33.6 1975 24.9 Occupation  High skill 75 19.4 2373 29.9  Medium skill 97 25.1 2395 30.2  Low skill 65 16.8 1364 17.2  Unemployed/not in the labour force 150 38.8 1799 22.7  Missing n=0 n=5 Income, weekly $AUD (mean (SD)) 387 833.9 (476.8) 7936 986.9 (496.5) Wealth  High 138 35.7 2952 37.2  Medium 130 33.6 2616 33.0  Low 119 30.8 2368 29.8 Financial hardship  Prosperous/very comfortable 42 12.0 1444 20.3  Reasonably comfortable 188 53.9 3874 54.6  Just getting by/very poor 119 34.1 1781 25.1  Missing n=38 n=837 Financial satisfaction [mean (SD)]a 387 6.4 (2.3) 7931 6.7 (2.0)  Missing n=0 n=5 Housing tenure  Outright owner 148 38.2 2107 26.6  Mortgager 126 32.6 3393 42.8  Other 113 29.2 2425 30.6  Missing n=0 n=11 Housing affordability  Affordable 348 91.1 7263 92.4  Unaffordable 34 8.9 597 7.6  Missing n=5 n=76 Relationship  Yes 276 71.3 5173 65.3  No 111 28.7 2755 34.8  Missing n=0 n=8 Children  No 105 27.1 3224 40.6  Yes 282 72.9 4712 59.4 Social support [mean (SD)]b 347 5.3 (1.1) 7017 5.6 (1.0)  Missing n=40 n=919 Frequency of socializing [mean (SD)]c 347 3.8 (1.6) 7068 3.4 (1.4)  Missing n=40 n=868 Relationship with parents [mean (SD)]d 197 7.9 (2.2) 5833 8.1 (2.0)  Missing n=190 n=2103 Alcohol consumption  Never 59 16.9 1108 15.6  Rarely 129 36.9 2540 35.8  One or two times/week 62 17.7 1547 21.8  At least three times/week 100 28.6 1909 26.9  Missing n=37 n=832 Smoking  Never smoked 164 47.0 4210 59.2  Ex-smoker 111 31.8 1684 23.7  Current 74 21.2 1222 17.2  Missing n=38 n=820 Physical activity  At least four times/week 120 34.2 2638 37.0  One to three times/week 137 39.0 2998 42.1  Less than once/week 94 26.8 1493 20.9  Missing n=36 n=807 BMI, kg/m2 [mean (SD)] 332 27.4 (5.3) 6855 25.8 (5.0)  Missing n=55 n=1081 Mental health inventory (MHI) [mean (SD)]e 351 73.3 (18.9) 7125 77.6 (14.4)  Missing n=36 n=811 a Satisfaction with financial circumstances, ranked using an 11-point Likert scale ranging from ‘totally dissatisfied’ to ‘totally satisfied’. b Constructed using the average of 10 questions addressing aspects of emotional support, each rated on a seven-point Likert scale. c Frequency of socializing with friends or relatives, rated on a seven-point Likert scale ranging from daily to less than once every 3 months. d Satisfaction with relationship with parents, rated on an 11-point Likert scale ranging from ‘completely dissatisfied’ to ‘completely satisfied’. e Measured using five questions from the SF-36, each of which is scored using five response categories, and the total scores are transformed into a scale ranging from 0 to 100, with higher scores reflecting better mental health. Table 2 Distribution of baseline characteristics for people who acquired a disability and the control sample (n=8323) Disability No disability n=387 n=7936 n % N % Age, years (mean (SD)) 387 52.5 (18.1) 7936 41.2 (15.4) Sex  Men 193 49.9 3817 48.1  Women 194 50.1 4119 51.9 Country of birth  Australia 297 76.7 6257 78.8  Other 90 23.3 1679 21.2 Parent occupation  High skill 181 47.4 3988 51.1  Medium skill 129 33.8 2645 33.9  Low skill/never worked 72 18.9 1168 15.0  Missing n=5 n=135 Education  Bachelor or higher 66 17.1 2184 27.5  Secondary, certificate, diploma 191 49.4 3777 47.6  Did not complete secondary 130 33.6 1975 24.9 Occupation  High skill 75 19.4 2373 29.9  Medium skill 97 25.1 2395 30.2  Low skill 65 16.8 1364 17.2  Unemployed/not in the labour force 150 38.8 1799 22.7  Missing n=0 n=5 Income, weekly $AUD (mean (SD)) 387 833.9 (476.8) 7936 986.9 (496.5) Wealth  High 138 35.7 2952 37.2  Medium 130 33.6 2616 33.0  Low 119 30.8 2368 29.8 Financial hardship  Prosperous/very comfortable 42 12.0 1444 20.3  Reasonably comfortable 188 53.9 3874 54.6  Just getting by/very poor 119 34.1 1781 25.1  Missing n=38 n=837 Financial satisfaction [mean (SD)]a 387 6.4 (2.3) 7931 6.7 (2.0)  Missing n=0 n=5 Housing tenure  Outright owner 148 38.2 2107 26.6  Mortgager 126 32.6 3393 42.8  Other 113 29.2 2425 30.6  Missing n=0 n=11 Housing affordability  Affordable 348 91.1 7263 92.4  Unaffordable 34 8.9 597 7.6  Missing n=5 n=76 Relationship  Yes 276 71.3 5173 65.3  No 111 28.7 2755 34.8  Missing n=0 n=8 Children  No 105 27.1 3224 40.6  Yes 282 72.9 4712 59.4 Social support [mean (SD)]b 347 5.3 (1.1) 7017 5.6 (1.0)  Missing n=40 n=919 Frequency of socializing [mean (SD)]c 347 3.8 (1.6) 7068 3.4 (1.4)  Missing n=40 n=868 Relationship with parents [mean (SD)]d 197 7.9 (2.2) 5833 8.1 (2.0)  Missing n=190 n=2103 Alcohol consumption  Never 59 16.9 1108 15.6  Rarely 129 36.9 2540 35.8  One or two times/week 62 17.7 1547 21.8  At least three times/week 100 28.6 1909 26.9  Missing n=37 n=832 Smoking  Never smoked 164 47.0 4210 59.2  Ex-smoker 111 31.8 1684 23.7  Current 74 21.2 1222 17.2  Missing n=38 n=820 Physical activity  At least four times/week 120 34.2 2638 37.0  One to three times/week 137 39.0 2998 42.1  Less than once/week 94 26.8 1493 20.9  Missing n=36 n=807 BMI, kg/m2 [mean (SD)] 332 27.4 (5.3) 6855 25.8 (5.0)  Missing n=55 n=1081 Mental health inventory (MHI) [mean (SD)]e 351 73.3 (18.9) 7125 77.6 (14.4)  Missing n=36 n=811 Disability No disability n=387 n=7936 n % N % Age, years (mean (SD)) 387 52.5 (18.1) 7936 41.2 (15.4) Sex  Men 193 49.9 3817 48.1  Women 194 50.1 4119 51.9 Country of birth  Australia 297 76.7 6257 78.8  Other 90 23.3 1679 21.2 Parent occupation  High skill 181 47.4 3988 51.1  Medium skill 129 33.8 2645 33.9  Low skill/never worked 72 18.9 1168 15.0  Missing n=5 n=135 Education  Bachelor or higher 66 17.1 2184 27.5  Secondary, certificate, diploma 191 49.4 3777 47.6  Did not complete secondary 130 33.6 1975 24.9 Occupation  High skill 75 19.4 2373 29.9  Medium skill 97 25.1 2395 30.2  Low skill 65 16.8 1364 17.2  Unemployed/not in the labour force 150 38.8 1799 22.7  Missing n=0 n=5 Income, weekly $AUD (mean (SD)) 387 833.9 (476.8) 7936 986.9 (496.5) Wealth  High 138 35.7 2952 37.2  Medium 130 33.6 2616 33.0  Low 119 30.8 2368 29.8 Financial hardship  Prosperous/very comfortable 42 12.0 1444 20.3  Reasonably comfortable 188 53.9 3874 54.6  Just getting by/very poor 119 34.1 1781 25.1  Missing n=38 n=837 Financial satisfaction [mean (SD)]a 387 6.4 (2.3) 7931 6.7 (2.0)  Missing n=0 n=5 Housing tenure  Outright owner 148 38.2 2107 26.6  Mortgager 126 32.6 3393 42.8  Other 113 29.2 2425 30.6  Missing n=0 n=11 Housing affordability  Affordable 348 91.1 7263 92.4  Unaffordable 34 8.9 597 7.6  Missing n=5 n=76 Relationship  Yes 276 71.3 5173 65.3  No 111 28.7 2755 34.8  Missing n=0 n=8 Children  No 105 27.1 3224 40.6  Yes 282 72.9 4712 59.4 Social support [mean (SD)]b 347 5.3 (1.1) 7017 5.6 (1.0)  Missing n=40 n=919 Frequency of socializing [mean (SD)]c 347 3.8 (1.6) 7068 3.4 (1.4)  Missing n=40 n=868 Relationship with parents [mean (SD)]d 197 7.9 (2.2) 5833 8.1 (2.0)  Missing n=190 n=2103 Alcohol consumption  Never 59 16.9 1108 15.6  Rarely 129 36.9 2540 35.8  One or two times/week 62 17.7 1547 21.8  At least three times/week 100 28.6 1909 26.9  Missing n=37 n=832 Smoking  Never smoked 164 47.0 4210 59.2  Ex-smoker 111 31.8 1684 23.7  Current 74 21.2 1222 17.2  Missing n=38 n=820 Physical activity  At least four times/week 120 34.2 2638 37.0  One to three times/week 137 39.0 2998 42.1  Less than once/week 94 26.8 1493 20.9  Missing n=36 n=807 BMI, kg/m2 [mean (SD)] 332 27.4 (5.3) 6855 25.8 (5.0)  Missing n=55 n=1081 Mental health inventory (MHI) [mean (SD)]e 351 73.3 (18.9) 7125 77.6 (14.4)  Missing n=36 n=811 a Satisfaction with financial circumstances, ranked using an 11-point Likert scale ranging from ‘totally dissatisfied’ to ‘totally satisfied’. b Constructed using the average of 10 questions addressing aspects of emotional support, each rated on a seven-point Likert scale. c Frequency of socializing with friends or relatives, rated on a seven-point Likert scale ranging from daily to less than once every 3 months. d Satisfaction with relationship with parents, rated on an 11-point Likert scale ranging from ‘completely dissatisfied’ to ‘completely satisfied’. e Measured using five questions from the SF-36, each of which is scored using five response categories, and the total scores are transformed into a scale ranging from 0 to 100, with higher scores reflecting better mental health. Sequential causal mediation analysis Interactions between the exposure and the following mediator variables were included in the regression models: material factors including occupation, housing affordability, housing tenure and satisfaction with financial circumstances; psychosocial factors including social support, frequency of socializing and relationship status; and behavioural factors including smoking, alcohol consumption, physical activity, BMI and diet. The TCE of disability acquisition was estimated to be a 5.3-point reduction in MHI score (95% CI –6.8, –3.7) (Table 3). In the sequential approach, we first considered the mediated effect through material factors and estimated a mean 1.7-point decline (95% CI –2.8, –0.6) in MHI was occurring through material factors, which corresponds to 32.1% of the total effect. We then considered the additional effect of psychosocial factors and found that 33.2% was explained by both material and psychosocial factors (NIE: –1.7, 95% CI –3.0, –0.5) and the additional effect of behavioural factors explained 38.6% of the decline (NIE: –2.0, 95% CI –3.4, –0.6). Table 3 Total causal effect (TCE), natural direct effect (NDE) and natural indirect effect (NIE) of disability acquisition on mental health, with mediation through material factors, psychosocial and behavioural factors Material factors + psychosocial factors + behavioural factors Coef.a (95% CI) Coef.a (95% CI) Coef.a (95% CI) TCE –5.3 (–6.8, –3.7) –5.3 (–6.8, –3.7) –5.3 (–6.8, –3.7) NDE –3.6 (–5.4, –1.8) –3.5 (–5.3, –1.7) –3.2 (–5.1, –1.4) NIE –1.7 (–2.8, –0.6) –1.7 (–3.0, –0.5) –2.0 (–3.4, –0.6) Proportion of effect explained (%) 32.1 (10.1, 54.1) 33.2 (8.5, 58.0) 38.6 (11.4, 65.9) Material factors + psychosocial factors + behavioural factors Coef.a (95% CI) Coef.a (95% CI) Coef.a (95% CI) TCE –5.3 (–6.8, –3.7) –5.3 (–6.8, –3.7) –5.3 (–6.8, –3.7) NDE –3.6 (–5.4, –1.8) –3.5 (–5.3, –1.7) –3.2 (–5.1, –1.4) NIE –1.7 (–2.8, –0.6) –1.7 (–3.0, –0.5) –2.0 (–3.4, –0.6) Proportion of effect explained (%) 32.1 (10.1, 54.1) 33.2 (8.5, 58.0) 38.6 (11.4, 65.9) a These primary analysis results were obtained using multiple imputation using chained equations with 50 imputed datasets. Table 3 Total causal effect (TCE), natural direct effect (NDE) and natural indirect effect (NIE) of disability acquisition on mental health, with mediation through material factors, psychosocial and behavioural factors Material factors + psychosocial factors + behavioural factors Coef.a (95% CI) Coef.a (95% CI) Coef.a (95% CI) TCE –5.3 (–6.8, –3.7) –5.3 (–6.8, –3.7) –5.3 (–6.8, –3.7) NDE –3.6 (–5.4, –1.8) –3.5 (–5.3, –1.7) –3.2 (–5.1, –1.4) NIE –1.7 (–2.8, –0.6) –1.7 (–3.0, –0.5) –2.0 (–3.4, –0.6) Proportion of effect explained (%) 32.1 (10.1, 54.1) 33.2 (8.5, 58.0) 38.6 (11.4, 65.9) Material factors + psychosocial factors + behavioural factors Coef.a (95% CI) Coef.a (95% CI) Coef.a (95% CI) TCE –5.3 (–6.8, –3.7) –5.3 (–6.8, –3.7) –5.3 (–6.8, –3.7) NDE –3.6 (–5.4, –1.8) –3.5 (–5.3, –1.7) –3.2 (–5.1, –1.4) NIE –1.7 (–2.8, –0.6) –1.7 (–3.0, –0.5) –2.0 (–3.4, –0.6) Proportion of effect explained (%) 32.1 (10.1, 54.1) 33.2 (8.5, 58.0) 38.6 (11.4, 65.9) a These primary analysis results were obtained using multiple imputation using chained equations with 50 imputed datasets. Sensitivity analysis The results were robust to the changes implied by the scenarios in the sensitivity analyses. The bias analysis demonstrated that the estimated indirect effects were unlikely to be explained by unmeasured confounding (Supplementary File 3, available as Supplementary Data at IJE online). Removing disabled people with psychological impairments (41 of 387) attenuated the effect estimates; however, the proportion of the effect mediated increased slightly. For the complete case analysis, only small changes in the magnitude of individual coefficients were observed (Table 4). Table 4 Results of the sensitivity analyses showing total causal effect (TCE), natural direct effect (NDE) and natural indirect effect (NIE) of disability acquisition on mental health, with mediation through material, psychosocial and behavioural factors Material factors + psychosocial factors + behavioural factors Coef. (95% CI) Coef. (95% CI) Coef. (95% CI) Psychological impairments removeda  TCE –4.3 (–5.9, –2.7) –4.3 (–5.9, –2.7) –4.3 (–5.9, –2.7)  NDE –2.7 (–4.6, –0.8) –2.6 (–4.4, –0.8) –2.2 (–4.0, –0.3)  NIE –1.6 (–2.7, –0.5) –1.7 (–3.0, –0.5) –2.1 (–3.5, –0.8)  Proportion of effect explained (%) 37.7 (7.0, 68.4) 40.2 (8.8, 71.6) 49.8 (14.6, 84.9) Complete case analysis  TCE –5.1 (–7.7, –2.5) –5.1 (–7.7, –2.5) –5.1 (–7.7, –2.5)  NDE –3.3 (–5.8, –0.8) –3.2 (–5.7, –0.7) –3.1 (–5.7, –0.5)  NIE –1.7 (–3.6, 0.1) –1.9 (–4.1, 0.3) –2.0 (–4.5, 0.6)  Proportion of effect explained (%) 34.4 (0.2, 68.7) 36.9 (–4.4, 78.2) 38.9 (–10.0, 87.8) Material factors + psychosocial factors + behavioural factors Coef. (95% CI) Coef. (95% CI) Coef. (95% CI) Psychological impairments removeda  TCE –4.3 (–5.9, –2.7) –4.3 (–5.9, –2.7) –4.3 (–5.9, –2.7)  NDE –2.7 (–4.6, –0.8) –2.6 (–4.4, –0.8) –2.2 (–4.0, –0.3)  NIE –1.6 (–2.7, –0.5) –1.7 (–3.0, –0.5) –2.1 (–3.5, –0.8)  Proportion of effect explained (%) 37.7 (7.0, 68.4) 40.2 (8.8, 71.6) 49.8 (14.6, 84.9) Complete case analysis  TCE –5.1 (–7.7, –2.5) –5.1 (–7.7, –2.5) –5.1 (–7.7, –2.5)  NDE –3.3 (–5.8, –0.8) –3.2 (–5.7, –0.7) –3.1 (–5.7, –0.5)  NIE –1.7 (–3.6, 0.1) –1.9 (–4.1, 0.3) –2.0 (–4.5, 0.6)  Proportion of effect explained (%) 34.4 (0.2, 68.7) 36.9 (–4.4, 78.2) 38.9 (–10.0, 87.8) a These sensitivity analysis results were obtained using multiple imputation using chained equations with 50 imputed datasets. Table 4 Results of the sensitivity analyses showing total causal effect (TCE), natural direct effect (NDE) and natural indirect effect (NIE) of disability acquisition on mental health, with mediation through material, psychosocial and behavioural factors Material factors + psychosocial factors + behavioural factors Coef. (95% CI) Coef. (95% CI) Coef. (95% CI) Psychological impairments removeda  TCE –4.3 (–5.9, –2.7) –4.3 (–5.9, –2.7) –4.3 (–5.9, –2.7)  NDE –2.7 (–4.6, –0.8) –2.6 (–4.4, –0.8) –2.2 (–4.0, –0.3)  NIE –1.6 (–2.7, –0.5) –1.7 (–3.0, –0.5) –2.1 (–3.5, –0.8)  Proportion of effect explained (%) 37.7 (7.0, 68.4) 40.2 (8.8, 71.6) 49.8 (14.6, 84.9) Complete case analysis  TCE –5.1 (–7.7, –2.5) –5.1 (–7.7, –2.5) –5.1 (–7.7, –2.5)  NDE –3.3 (–5.8, –0.8) –3.2 (–5.7, –0.7) –3.1 (–5.7, –0.5)  NIE –1.7 (–3.6, 0.1) –1.9 (–4.1, 0.3) –2.0 (–4.5, 0.6)  Proportion of effect explained (%) 34.4 (0.2, 68.7) 36.9 (–4.4, 78.2) 38.9 (–10.0, 87.8) Material factors + psychosocial factors + behavioural factors Coef. (95% CI) Coef. (95% CI) Coef. (95% CI) Psychological impairments removeda  TCE –4.3 (–5.9, –2.7) –4.3 (–5.9, –2.7) –4.3 (–5.9, –2.7)  NDE –2.7 (–4.6, –0.8) –2.6 (–4.4, –0.8) –2.2 (–4.0, –0.3)  NIE –1.6 (–2.7, –0.5) –1.7 (–3.0, –0.5) –2.1 (–3.5, –0.8)  Proportion of effect explained (%) 37.7 (7.0, 68.4) 40.2 (8.8, 71.6) 49.8 (14.6, 84.9) Complete case analysis  TCE –5.1 (–7.7, –2.5) –5.1 (–7.7, –2.5) –5.1 (–7.7, –2.5)  NDE –3.3 (–5.8, –0.8) –3.2 (–5.7, –0.7) –3.1 (–5.7, –0.5)  NIE –1.7 (–3.6, 0.1) –1.9 (–4.1, 0.3) –2.0 (–4.5, 0.6)  Proportion of effect explained (%) 34.4 (0.2, 68.7) 36.9 (–4.4, 78.2) 38.9 (–10.0, 87.8) a These sensitivity analysis results were obtained using multiple imputation using chained equations with 50 imputed datasets. Discussion Interpretation of findings In this analysis, we found that 32% of the effect of disability acquisition on mental health was mediated by material factors, with only a negligible proportion explained by the addition of psychosocial factors and 5% by behavioural factors. This is consistent with the majority of the literature explaining health inequalities, which found that health differences are predominantly attributable to material factors.11,17 The results were not consistent with studies that had shown that psychosocial resources accounted for some of the effect of disability on depression;6,19,20 however, these pathways are not mutually exclusive and it is possible that a large proportion of the effect through material factors is also operating through psychosocial pathways. Previous studies did not use a sequential causal mediation approach, which allows estimation of the additional contribution of psychosocial factors beyond the effect that is operating through material factors.48 The effect sizes estimated in this study were of clinical significance. Study participants who acquired a disability experienced on average a five-point decline in mental health, exceeding the four- to five-point difference considered to represent a clinically meaningful change.29,39,40 The effect mediated through material factors was estimated to be 32.1%, which can be interpreted as the proportion of the mental health decline that could be avoided if people with disabilities experienced the same material socio-economic circumstances as those without disabilities. About two-fifths of the effect (38.6%) was explained by all three sets of mediators, leaving a large proportion of the effect unexplained—it seems unlikely that the remaining 61.4% of the total effect is not mediated by any other factors and is therefore a true ‘direct’ effect. This is perhaps not surprising as, despite measuring a broad range of socio-economic characteristics, these measures capture only a snapshot of people’s socio-economic experiences at one point in time54 and do not capture the broader structural, political and economic processes they experience.11 Additionally, there were some factors that were not recorded in HILDA that could be important mediators, such as experience of discrimination, sense of personal control (asked only in 2011 and 2015), psychosocial working conditions and personal–work balance (asked only for those people who were employed). Therefore, the effect operating through psychosocial pathways may be underestimated. Strengths and limitations This study used data from a large longitudinal survey in Australia. The longitudinal nature of the data meant that we could characterize disability acquisition, based on a sample of people who reported no disability for two waves followed immediately by two waves of disability. Furthermore, we could measure disability acquisition, mediators and mental health at different time points, to establish a temporal sequence between them, and control for prior values of the mediator and mental health score so that the results can be interpreted as effects of changes in the mediators on the outcome. We used causal sequential mediation methods, which can address the limitations of traditional mediation methods, generating unbiased estimates of mediation through multiple causally ordered mediators, given a set of clearly specified assumptions of no confounding. There were also limitations with this study. The analysis rests on several strong assumptions about no confounding between disability acquisition, mediators and mental health. We used inverse probability weighting to account for (measured) confounding of the disability–mental health and the disability–mediator relationships. For the assumption relating to no uncontrolled confounding of the mediator–outcome relationship, we conducted a bias analysis which suggested that the NDE and NIE were unlikely to be explained by confounding by measured or unmeasured variables. The weighting approach is sensitive to outcome model misspecification, which can lead to biased estimates of natural direct and indirect effects; however, this approach was deemed most appropriate because of the large number of mediators.48 Furthermore, to ensure best specification of the outcome model, interactions between the exposure and each mediator were considered and tested. There were strong assumptions about the causal ordering of the mediators. The direction of causality between these contributory factors is likely to be bi-directional, e.g. the relationship between employment and social support. This may have led to overestimation of the proportion of the effect operating through material factors if these are consequences of psychosocial and behavioural factors, rather than a cause of them. However, for most of the variables considered, the effect is likely to be causally ordered from material to psychosocial to behavioural factors. There was a large proportion of missing data and this was higher in participants with poorer mental health and greater socio-economic disadvantage; however, the use of MI as the primary analysis should have reduced this selection bias. The concepts of disability and mental health are related, which makes it difficult to isolate the causal effect of one on the other. To address this limitation, first we chose to use the mental health subscale of the SF-36 health questionnaire (MHI), rather than the summary mental health score (MCS), therefore selecting parts of the SF-36 questionnaire that were less likely to overlap with the definition of disability. Furthermore, we conducted a sensitivity analysis in which we excluded people with psychological impairments to further minimize overlap between the concepts, which did not change the interpretation of the results though the magnitude of effect estimates was slightly attenuated. When we excluded people with psychological impairments, the proportion of the effect mediated was slightly larger. It is plausible that the mechanisms are different for people with psychological impairments compared with other types of disability, though the relative proportion through each of the three pathways was similar. It would be interesting to look at differences in these effects according to types of impairments; however, we lacked power to examine differences by disability characteristics. Finally, people with severe disabilities are less likely to participate in HILDA; therefore, our results are likely to underestimate the population effect of disability acquisition on mental health. Conclusions The finding that the effect of disability acquisition on mental health operates predominantly through material factors has important policy implications. These results highlight that social policy reforms that reduce socio-economic disadvantage among people who acquire a disability will improve mental health. This could be achieved through better social protection, including income support, but also through improved educational and employment opportunities for people with disabilities and access to affordable housing. It is important to further disentangle the mechanisms involved in the material pathway, to better understand the relative importance of specific factors and which social determinants are driving the mental health inequalities. This will help to better target policy interventions to improved the mental health of people with disabilities. Supplementary Data Supplementary data are available at IJE online. Funding This work was supported by an Australian Government Research Training Program Scholarship and a National Health and Medical Research Council Postgraduate Scholarship (1093740 to Z.A.), a National Health and Medical Research Council Senior Research Fellowship (1104975 to J.A.S.) and an Australian Research Council Future Fellowship (FT150100131 to R.B.). Conflict of interest: The authors have no conflicts of interest to declare. References 1 World Health Organization and World Bank Group . World Report on Disability . Geneva : WHO , 2011 . 2 Kavanagh AM , Aitken Z , Baker E et al. Housing tenure and affordability and mental health following disability acquisition in adulthood . Soc Sci Med 2016 ; 151 : 225 – 32 . Google Scholar CrossRef Search ADS PubMed 3 Kavanagh AM , Aitken Z , Krnjacki L , LaMontagne AD , Bentley R , Milner A. Mental health following acquisition of disability in adulthood: the impact of wealth . PLoS One 2015 ; 10 : e0139708 . Google Scholar CrossRef Search ADS PubMed 4 Mandemakers JJ , Monden CWS. Does education buffer the impact of disability on psychological distress? Soc Sci Med 2010 ; 71 : 288 – 97 . Google Scholar CrossRef Search ADS PubMed 5 Smith DM , Langa KM , Kabeto MU , Ubel PA. Health, wealth, and happiness: financial resources buffer subjective well-being after the onset of a disability . Psychol Sci 2005 ; 16 : 663 – 66 . Google Scholar CrossRef Search ADS PubMed 6 Yang Y. How does functional disability affect depressive symptoms in late life? The role of perceived social support and psychological resources . J Health Soc Behav 2006 ; 47 : 355 – 72 . Google Scholar CrossRef Search ADS PubMed 7 Kariuki M , Honey A , Emerson E , Llewellyn G. Mental health trajectories of young people after disability onset . Disabil Health J 2011 ; 4 : 91 – 101 . Google Scholar CrossRef Search ADS PubMed 8 Emerson E , Kariuki M , Honey A , Llewellyn G. Becoming disabled: the association between disability onset in younger adults and subsequent changes in productive engagement, social support, financial hardship and subjective wellbeing . Disabil Health J 2014 ; 7 : 448 – 56 . Google Scholar CrossRef Search ADS PubMed 9 Yang Y , George LK. Functional disability, disability transitions, and depressive symptoms in late life . J Aging Health 2005 ; 17 : 263 – 92 . Google Scholar CrossRef Search ADS PubMed 10 Lynch J. Income inequality and health: expanding the debate . Soc Sci Med 2000 ; 51 : 1001 – 05 . Google Scholar CrossRef Search ADS PubMed 11 Lynch JW , Smith GD , Kaplan GA , House JS. Income inequality and mortality: importance to health of individual income, psychosocial environment, or material conditions . Brit Med J 2000 ; 320 : 1200 – 04 . Google Scholar CrossRef Search ADS PubMed 12 Marmot M , Wilkinson R. Social Determinants of Health . Oxford : Oxford University Press , 1999 . 13 Bartley M. Health Inequality: An Introduction to Theories, Concepts and Methods . Cambridge : Polity Press , 2004 . 14 DHSS . Inequalities in Health: Report of a Research Working Group . London : DHSS , 1980 . 15 Wilkinson RG. Socioeconomic determinants of health—health inequalities: relative or absolute material standards? Brit Med J 1997 ; 314 : 591 – 95 . Google Scholar CrossRef Search ADS PubMed 16 Pocock SJ , Cook DG , Shaper AG , Phillips AN , Walker M. Social-class differences in ischemic-heart-disease in British men . Lancet 1987 ; 2 : 197 – 201 . Google Scholar CrossRef Search ADS PubMed 17 Davey Smith G , Bartley M. Explanations for socio-economic differentials in mortality . Eur J Public Health 1994 ; 4 : 131 – 44 . Google Scholar CrossRef Search ADS 18 Power C , Matthews S. Origins of health inequalities in a national population sample . Lancet 1997 ; 350 : 1584 – 89 . Google Scholar CrossRef Search ADS PubMed 19 Kiviruusu O , Huurre T , Aro H. Psychosocial resources and depression among chronically ill young adults: are males more vulnerable? Soc Sci Med 2007 ; 65 : 173 – 86 . Google Scholar CrossRef Search ADS PubMed 20 Bisschop MI , Knegsman DMW , Beekman ATF , Deeg DJH. Chronic diseases and depression: the modifying role of psychosocial resources . Soc Sci Med 2004 ; 59 : 721 – 33 . Google Scholar CrossRef Search ADS PubMed 21 Hafeman DM , Schwartz S. Opening the Black Box: a motivation for the assessment of mediation . Int J Epidemiol 2009 ; 38 : 838 – 45 . Google Scholar CrossRef Search ADS PubMed 22 Jahoda M. Employment and Unemployment: A Social-Psychological Analysis . Cambridge : Cambridge University Press , 1982 . 23 Summerfield M , Freidin S , Hahn M et al. HILDA User Manual—Release 14. Melbourne : Melbourne Institute of Applied Economic and Social Research , 2015 . 24 Burchardt T. Being and Becoming: Social Exclusion and the Onset of Disability . London : The London School of Economics , 2003 . 25 Jenkins SP , Rigg JA. Disability and disadvantage: selection, onset, and duration effects . J Soc Policy 2004 ; 33 : 479 – 501 . Google Scholar CrossRef Search ADS 26 Polidano C , Vu H. Differential labour market impacts from disability onset . Health Econ 2015 ; 24 : 302 – 17 . Google Scholar CrossRef Search ADS PubMed 27 Burkhauser R , Daly M. Employment and economic well-being following the onset of a disability: the role for public policy. In: Mashaw J , Reno V , Burkhauser R , Berkowitz M (eds). Disability, Work, and Cash Benefits . Kalamazoo MI : Upjohn Institute for Employment Research , 1996 . Google Scholar CrossRef Search ADS 28 Butterworth P , Crosier T. The validity of the SF-36 in an Australian National Household Survey: demonstrating the applicability of the Household Income and Labour Dynamics in Australia (HILDA) Survey to examination of health inequalities . BMC Public Health 2004 ; 4 : 44 . Google Scholar CrossRef Search ADS PubMed 29 Ware JE Jr , Kosinski M , Gandek B. SF-36 Health Survey: Manual and Interpretation Guide . Lincoln, RI : QualityMetric , 2000 . 30 McCabe CJ , Thomas KJ , Brazier JE , Coleman P. Measuring the mental health status of a population: a comparison of the GHQ-12 and the SF-36 (MHI-5) . British Journal of Psychiatry: The Journal of Mental Science 1996 ; 169 : 516 – 21 . Google Scholar CrossRef Search ADS 31 Strand BH , Dalgard OS , Tambs K , Rognerud M. Measuring the mental health status of the Norwegian population: a comparison of the instruments SCL-25, SCL-10, SCL-5 and MHI-5 (SF-36) . Nord J Psychiat 2003 ; 57 : 113 – 81 . Google Scholar CrossRef Search ADS 32 Weinstein MC , Berwick DM , Goldman PA , Murphy JM , Barsky AJ. A comparison of three psychiatric screening tests using receiver operating characteristic (ROC) analysis . Med Care 1989 ; 27 : 593 – 607 . Google Scholar CrossRef Search ADS PubMed 33 Yamazaki S , Fukuhara S , Green J. Usefulness of five-item and three-item Mental Health Inventories to screen for depressive symptoms in the general population of Japan . Health Qual Life Out 2005 ; 3 : 48 . Google Scholar CrossRef Search ADS 34 Berwick DM , Murphy JM , Goldman PA , Ware JE Jr , Barsky AJ , Weinstein MC. Performance of a five-item mental health screening test . Med Care 1991 ; 29 : 169 – 76 . Google Scholar CrossRef Search ADS PubMed 35 Cuijpers P , Smits N , Donker T , ten Have M , de Graaf R. Screening for mood and anxiety disorders with the five-item, the three-item, and the two-item Mental Health Inventory . Psychiat Res 2009 ; 168 : 250 – 55 . Google Scholar CrossRef Search ADS 36 Means-Christensen AJ , Arnau RC , Tonidandel AM , Bramson R , Meagher MW. An efficient method of identifying major depression and panic disorder in primary care . J Behav Med 2005 ; 28 : 565 – 72 . Google Scholar CrossRef Search ADS PubMed 37 Rumpf HJ , Meyer C , Hapke U , John U. Screening for mental health: validity of the MHI-5 using DSM-IV Axis I psychiatric disorders as gold standard . Psychiat Res 2001 ; 105 : 243 – 53 . Google Scholar CrossRef Search ADS 38 Pirkis J , Burgess P , Kirk P , Dodson S , Coombs T. Review of Standardised Measures Used in the National Outcomes and Casemix Collection (NOCC) . Sydney : NSW Institute of Psychiatry , 2005 . 39 Knox SA , King MT. Validation and calibration of the SF-36 health transition question against an external criterion of clinical change in health status . Qual Life Res 2009 ; 18 : 637 – 45 . Google Scholar CrossRef Search ADS PubMed 40 Mchorney CA , Ware JE , Raczek AE . The Mos 36-Item Short-Form Health Survey (Sf-36). 2. Psychometric and clinical-tests of validity in measuring physical and mental-health constructs . Med Care 1993 ; 31 : 247 – 63 . Google Scholar CrossRef Search ADS PubMed 41 Moor I , Spallek J , Richter M. Explaining socioeconomic inequalities in self-rated health: a systematic review of the relative contribution of material, psychosocial and behavioural factors . J Epidemiol Commun H 2017 ; 71 : 565 – 75 . Google Scholar CrossRef Search ADS 42 Skalicka V , van Lenthe F , Bambra C , Krokstad S , Mackenbach J. Material, psychosocial, behavioural and biomedical factors in the explanation of relative socio-economic inequalities in mortality: evidence from the HUNT study . Int J Epidemiol 2009 ; 38 : 1272 – 84 . Google Scholar CrossRef Search ADS PubMed 43 van Oort FVA , van Lenthe FJ , Mackenbach JP. Material, psychosocial, and behavioural factors in the explanation of educational inequalities in mortality in the Netherlands . J Epidemiol Commun H 2005 ; 59 : 214 – 20 . Google Scholar CrossRef Search ADS 44 McMunn A , Nazroo J , Breeze E. Inequalities in health at older ages: a longitudinal investigation of the onset of illness and survival effects in England . Age Ageing 2009 ; 38 : 181 – 87 . Google Scholar CrossRef Search ADS PubMed 45 Emerson E, , Madden R, , Robertson J, , Graham H , Hatton C , Llewellyn G. Intellectual and Physical Disability, Social Mobility, Social Inclusion and Health: Background Paper for the Marmot Review . Lancaster : Centre for Disability Research, Lancaster University , 2009 . 46 World Health Organisation . The International Classification of Functioning, Disability and Health . Geneva : WHO , 2001 . 47 World Health Organization . Towards a Common Language for Functioning, Disability and Health: ICF the International Classification of Functioning, Disability and Health . Geneva : WHO , 2002 . 48 VanderWeele TJ , Vansteelandt S. Mediation analysis with multiple mediators . Epidemiologic Methods 2014 ; 2 : 95 – 115 . Google Scholar CrossRef Search ADS PubMed 49 Vansteelandt S , Bekaert M , Lange T. Imputation strategies for the estimation of natural direct and indirect effects . Epidemiologic Methods 2012 ; 1 : 131 – 58 . Google Scholar CrossRef Search ADS 50 Hernan MA , Robins JM. Estimating causal effects from epidemiological data . J Epidemiol Commun H 2006 ; 60 : 578 – 86 . Google Scholar CrossRef Search ADS 51 VanderWeele TJ. Explanation in Causal Inference: Methods for Mediation and Interaction . New York : Oxford University Press , 2015 . 52 Rubin D. Multiple Imputation for Nonresponse in Surveys . New York : Wiley , 1987 . Google Scholar CrossRef Search ADS 53 VanderWeele TJ. Bias formulas for sensitivity analysis for direct and indirect effects . Epidemiology 2010 ; 21 : 540 – 51 . Google Scholar CrossRef Search ADS PubMed 54 De Stavola BL , Daniel RM. Commentary: Incorporating concepts and methods from causal inference into life course epidemiology . Int J Epidemiol 2016 ; 45 : 1006 – 10 . Google Scholar CrossRef Search ADS PubMed © The Author(s) 2018; all rights reserved. Published by Oxford University Press on behalf of the International Epidemiological Association This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png International Journal of Epidemiology Oxford University Press

Do material, psychosocial and behavioural factors mediate the relationship between disability acquisition and mental health? A sequential causal mediation analysis

Loading next page...
 
/lp/ou_press/do-material-psychosocial-and-behavioural-factors-mediate-the-rQbAovhJG1
Publisher
Oxford University Press
Copyright
© The Author(s) 2018; all rights reserved. Published by Oxford University Press on behalf of the International Epidemiological Association
ISSN
0300-5771
eISSN
1464-3685
D.O.I.
10.1093/ije/dyx277
Publisher site
See Article on Publisher Site

Abstract

Abstract Background There is evidence of a causal relationship between disability acquisition and poor mental health; however, the mechanism by which disability affects mental health is poorly understood. This gap in understanding limits the development of effective interventions to improve the mental health of people with disabilities. Methods We used four waves of data from the Household, Income and Labour Dynamics in Australia Survey (2011–14) to compare self-reported mental health between individuals who acquired any disability (n=387) and those who remained disability-free (n=7936). We tested three possible pathways from disability acquisition to mental health, examining the effect of material, psychosocial and behavioural mediators. The effect was partitioned into natural direct and indirect effects through the mediators using a sequential causal mediation analysis approach. Multiple imputation using chained equations was used to assess the impact of missing data. Results Disability acquisition was estimated to cause a five-point decline in mental health [estimated mean difference: –5.3, 95% confidence interval (CI) –6.8, –3.7]. The indirect effect through material factors was estimated to be a 1.7-point difference (–1.7, 95% CI –2.8, –0.6), explaining 32% of the total effect, with a negligible proportion of the effect explained by the addition of psychosocial characteristics (material and psychosocial: –1.7, 95% CI –3.0, –0.5) and a further 5% by behavioural factors (material-psychosocial-behavioural: –2.0, 95% CI –3.4, –0.6). Conclusions The finding that the effect of disability acquisition on mental health operates predominantly through material rather than psychosocial and behavioural factors has important implications. The results highlight the need for better social protection, including income support, employment and education opportunities, and affordable housing for people who acquire a disability. disability, mental health, health inequalities, social epidemiology, causal mediation analysis Key Messages This paper investigated the mechanistic pathways linking disability acquisition and mental health using sequential causal mediation analysis to examining the effect of material, psychosocial and behavioural factors as mediators of the association. The total causal effect of disability acquisition on mental health was estimated to be a five-point decline in Mental Health Inventory (MHI) score. The effect was partially explained by the three sets of mediators, with 32% of the total effect mediated by material factors, a negligible proportion mediated by the addition of psychosocial factors and a further 5% by behavioural factors. The effect of disability acquisition on mental health operates predominantly through material factors, highlighting the need for better social protection, including income support, employment and education opportunities, and affordable housing for people who acquire a disability. Introduction Currently, more than a billion people, approximately 15% of the world’s population, live with a disability.1 People with disabilities experience substantial health inequalities and are at high risk of poor mental health.2,3 A causal relationship between disability and poor mental health has been suggested from analyses of prospective cohort studies.2–9 However, the mechanism by which disability leads to deterioration in mental helath is poorly understood. There are a number of different potential explanations for a causal link between disability and poor mental health. Supported by theoretical and empirical studies of the mechanisms underlying income-related health inqualities, three frameworks have become well established in explaining how social determinants influence health: material, psychosocial and behavioural pathways.10–13 First, the material pathway, by which differential exposure to structural and material living conditions leads to health inequalities, which posits that material conditions such as poverty and economic deprivation affect health directly, but also have indirect effects by enabling access to better living circumstances such as access to health care.14 Second, the psychosocial perspective emphasizes the importance of psychosocial and stress-related risk factors on health, with inequalities arising from the unequal distribution of psychosocial factors such as social support, home–work balance and personal control.15 Third, differences in health-related behaviour are thought to contribute to health inequalities, e.g. smoking, physical activity and diet.16 There has been considerable debate regarding the relative importance of these factors in explaining social inequalities in health.11 Most empirical studies have argued for the significance of material pathways,11,17 postulated to have a greater relative contribution because they exert both a direct effect on health as well as an indirect effect through psychosocial and behavioural pathways.12,18 With regard to the mechanisms driving disability-related mental health inequalities, disability acquisition may lead to changes in material, psychosocial and behavioural factors, which could explain, or mediate, the observed mental health deterioration. At present, it is not clear to what extent the effect of disability on mental health operates through these proposed pathways or through other mechanisms. Evidence regarding the importance of different pathways between disability and mental health is sparse; the research has mainly been conducted in cross-sectional studies of people with chronic illness, has only examined psychosocial pathways and no study has examined multiple pathways simultaneously. Three studies examined mediation of the effect through psychosocial resources and found evidence that some of the effect of disability acquisition on depressive symptoms6 and depression was operating through this pathway.6,19,20 Understanding the mechanisms underpinning these mental health inequalities is an important public health question because socio-economic intermediary variables are potential modifiable targets for interventions to mitigate the adverse effects of disability on people’s mental health.21 In this study, we use data from four waves of a longitudinal study of Australian adults and apply recently developed methods—sequential causal mediation analysis—to estimate the relative importance of three distinct mechanistic pathways leading from disability acquisition to poor mental health, quantifying the indirect effects through material, psychosocial and behavioural factors (Figure 1). Material factors are likely to affect mental health directly as well as indirectly through psychosocial factors such as latent consequences of employment (e.g. purposeful time use, self-esteem)22 and behavioural factors. Similarly, psychosocial factors are thought to exert a direct effect on mental health, and an indirect effect through behavioural factors. Figure 1 View largeDownload slide Casual diagram illustrating postulated causal relationships between disability acquisition and mental health. Figure 1 View largeDownload slide Casual diagram illustrating postulated causal relationships between disability acquisition and mental health. Methods Data source The Household, Income and Labour Dynamics in Australia (HILDA) Survey is a longitudinal study of Australian households, conducted annually since 2001.23 The survey collects information on demographic, social, economic and health characteristics of individuals using a combination of interviews and self-completed questionnaires. The original sample included 13 969 participants from 7682 households, randomly sampled using a national probability sample of private dwellings. A top-up sample was added in 2011 to maintain representativeness, leading to a sample size after 14 waves of 28 794 people. On average, for all waves of the survey, response proportions were 80% (ranging from 70% to 92%) and attrition was 5.7% between waves, ranging from 3.5% in 2014 to 13.2% in 2002. The analysis used four waves of the survey (2011 to 2014) to establish a temporal sequence between disability acquisition, the mediators and mental health. Disability acquisition Information on disability was collected in every wave, using a single question defining disability as ‘an impairment, disability or long-term health condition, which restricts everyday activities that had lasted for six months or more’. Disability acquisition was defined as two waves reporting no disability, followed immediately by two consecutive waves reporting a disability. We used two consecutive waves of disability so as to exclude people with transient disability and to reduce the potential for measurement error—a definition used in previous studies examining disability acquisition.24–27 Participants who acquired a disability were compared with those who reported no disability in any of the four waves. People who reported other patterns of exposure, such as a single wave of disability, were excluded. Eligibility for inclusion required participation and response to the disability question at all four waves. Mental health Mental health was assessed in the final wave (2014) using the Mental Health Inventory (MHI), a subscale of the Short Form 36 (SF-36, a widely used general health questionnaire that has been validated in the Australian population using data from the HILDA Survey).28 The MHI is a well-validated and reliable measure of mental health status.29 It measures symptoms of depression, anxiety and psychological wellbeing and has been shown to be an effective screening tool for mood and anxiety disorders and severe depressive symptomatology in comparisons with established mental health, wellbeing and depression scales30–33 as well as studies comparing against clinical diagnoses.34–37 The MHI has been shown to be psychometrically sound, with high internal consistency, discriminant validity and high test–retest reliability.38 It includes five items relating to mental health over the previous 4 weeks, each scored using five response categories. Total scores were transformed into a scale with a mean score of 74 (range: 0–100), as per standard practice, with higher scores reflecting better mental health. Previous research has suggested that a difference of four to five points on the MHI scale is likely to reflect a minimally important clinical difference in mental health.39,40 Mediators Mediator variables, described in Table 1, were measured in the third wave (2013). The choice of variables and their classification into three broad categories were motivated by reviewing empirical studies examining different explanations for socio-economic inequalities in health11,41–43 and selecting similar variables available in the HILDA Survey where possible. Table 1 Description of mediator variables Variables Type Definition/categorization Material factors  Occupation Categorical High skill; medium skill; low skill job; unemployed/not in labour force  Weekly income Continuous Equivalized household disposable income, $AUD  Financial hardship Categorical Prosperous/very comfortable; reasonably comfortable; just getting along/poor/very poor  Financial satisfaction Continuous Satisfaction with financial circumstances, ranked using an 11-point Likert scale ranging from ‘totally dissatisfied’ to ‘totally satisfied’  Housing tenure Categorical Outright owner; mortgager; renter  Housing affordability Binary Unaffordable defined as households in the lowest 40% of the income distribution with housing costs exceeding 30% of their gross income Psychosocial factors  Relationship status Binary Yes; no  Children Binary Yes; no  Social support Continuous Constructed using the average of 10 questions addressing aspects of emotional support, each rated on a seven-point Likert scale25  Socializing Continuous Frequency of socializing with friends or relatives, rated on a seven-point Likert scale ranging from daily to less than once every 3 months  Parent relationship Continuous Satisfaction with relationship with parents, rated on an 11-point Likert scale ranging from ‘completely dissatisfied’ to ‘completely satisfied’ Behavioural factors  Smoking Categorical Never; ex-smoker; current  Alcohol consumption Categorical Never; rarely; 1–2 days per week; >2 days per week  Physical activity Categorical >3 times per week; one to three times per week; less than once a week  Body mass index Continuous Self-reported, kg/m2  Healthy diet index Continuous Ranging from 0 ‘unhealthiest’ to 4 ‘healthiest’, derived from four binary questions: eating fruit every day; eating vegetables every day; eating fatty foods less than once a month; drinking low fat milk26  Quality of sleep Continuous Rated on a four-point Likert scale ranging from 1 ‘very good’ to 4 ‘very bad’ Variables Type Definition/categorization Material factors  Occupation Categorical High skill; medium skill; low skill job; unemployed/not in labour force  Weekly income Continuous Equivalized household disposable income, $AUD  Financial hardship Categorical Prosperous/very comfortable; reasonably comfortable; just getting along/poor/very poor  Financial satisfaction Continuous Satisfaction with financial circumstances, ranked using an 11-point Likert scale ranging from ‘totally dissatisfied’ to ‘totally satisfied’  Housing tenure Categorical Outright owner; mortgager; renter  Housing affordability Binary Unaffordable defined as households in the lowest 40% of the income distribution with housing costs exceeding 30% of their gross income Psychosocial factors  Relationship status Binary Yes; no  Children Binary Yes; no  Social support Continuous Constructed using the average of 10 questions addressing aspects of emotional support, each rated on a seven-point Likert scale25  Socializing Continuous Frequency of socializing with friends or relatives, rated on a seven-point Likert scale ranging from daily to less than once every 3 months  Parent relationship Continuous Satisfaction with relationship with parents, rated on an 11-point Likert scale ranging from ‘completely dissatisfied’ to ‘completely satisfied’ Behavioural factors  Smoking Categorical Never; ex-smoker; current  Alcohol consumption Categorical Never; rarely; 1–2 days per week; >2 days per week  Physical activity Categorical >3 times per week; one to three times per week; less than once a week  Body mass index Continuous Self-reported, kg/m2  Healthy diet index Continuous Ranging from 0 ‘unhealthiest’ to 4 ‘healthiest’, derived from four binary questions: eating fruit every day; eating vegetables every day; eating fatty foods less than once a month; drinking low fat milk26  Quality of sleep Continuous Rated on a four-point Likert scale ranging from 1 ‘very good’ to 4 ‘very bad’ Table 1 Description of mediator variables Variables Type Definition/categorization Material factors  Occupation Categorical High skill; medium skill; low skill job; unemployed/not in labour force  Weekly income Continuous Equivalized household disposable income, $AUD  Financial hardship Categorical Prosperous/very comfortable; reasonably comfortable; just getting along/poor/very poor  Financial satisfaction Continuous Satisfaction with financial circumstances, ranked using an 11-point Likert scale ranging from ‘totally dissatisfied’ to ‘totally satisfied’  Housing tenure Categorical Outright owner; mortgager; renter  Housing affordability Binary Unaffordable defined as households in the lowest 40% of the income distribution with housing costs exceeding 30% of their gross income Psychosocial factors  Relationship status Binary Yes; no  Children Binary Yes; no  Social support Continuous Constructed using the average of 10 questions addressing aspects of emotional support, each rated on a seven-point Likert scale25  Socializing Continuous Frequency of socializing with friends or relatives, rated on a seven-point Likert scale ranging from daily to less than once every 3 months  Parent relationship Continuous Satisfaction with relationship with parents, rated on an 11-point Likert scale ranging from ‘completely dissatisfied’ to ‘completely satisfied’ Behavioural factors  Smoking Categorical Never; ex-smoker; current  Alcohol consumption Categorical Never; rarely; 1–2 days per week; >2 days per week  Physical activity Categorical >3 times per week; one to three times per week; less than once a week  Body mass index Continuous Self-reported, kg/m2  Healthy diet index Continuous Ranging from 0 ‘unhealthiest’ to 4 ‘healthiest’, derived from four binary questions: eating fruit every day; eating vegetables every day; eating fatty foods less than once a month; drinking low fat milk26  Quality of sleep Continuous Rated on a four-point Likert scale ranging from 1 ‘very good’ to 4 ‘very bad’ Variables Type Definition/categorization Material factors  Occupation Categorical High skill; medium skill; low skill job; unemployed/not in labour force  Weekly income Continuous Equivalized household disposable income, $AUD  Financial hardship Categorical Prosperous/very comfortable; reasonably comfortable; just getting along/poor/very poor  Financial satisfaction Continuous Satisfaction with financial circumstances, ranked using an 11-point Likert scale ranging from ‘totally dissatisfied’ to ‘totally satisfied’  Housing tenure Categorical Outright owner; mortgager; renter  Housing affordability Binary Unaffordable defined as households in the lowest 40% of the income distribution with housing costs exceeding 30% of their gross income Psychosocial factors  Relationship status Binary Yes; no  Children Binary Yes; no  Social support Continuous Constructed using the average of 10 questions addressing aspects of emotional support, each rated on a seven-point Likert scale25  Socializing Continuous Frequency of socializing with friends or relatives, rated on a seven-point Likert scale ranging from daily to less than once every 3 months  Parent relationship Continuous Satisfaction with relationship with parents, rated on an 11-point Likert scale ranging from ‘completely dissatisfied’ to ‘completely satisfied’ Behavioural factors  Smoking Categorical Never; ex-smoker; current  Alcohol consumption Categorical Never; rarely; 1–2 days per week; >2 days per week  Physical activity Categorical >3 times per week; one to three times per week; less than once a week  Body mass index Continuous Self-reported, kg/m2  Healthy diet index Continuous Ranging from 0 ‘unhealthiest’ to 4 ‘healthiest’, derived from four binary questions: eating fruit every day; eating vegetables every day; eating fatty foods less than once a month; drinking low fat milk26  Quality of sleep Continuous Rated on a four-point Likert scale ranging from 1 ‘very good’ to 4 ‘very bad’ Baseline covariates Baseline covariates were measured in the first wave (2011), as a measure of people’s circumstances prior to disability acquisition. It is well documented in the literature that the incidence of disability is socially patterned, with people who experience socio-economic disadvantage being more likely to acquire a disability.25,44,45 Furthermore, according to the International Classification of Functioning, Disability and Health (ICF) framework, disability results from the interaction between health conditions, personal attributes and environmental factors.46,47 Conceived in this way, personal attributes such as the experience of financial strain, or characteristics of people’s social environment, such as their ability to access social support, are key determinants of disability as they influence the impact of people’s impairments on activity limitations and restriction to participation. Demographic characteristics included age, sex and country of birth (Australia; other) and socio-economic characteristics included education (bachelor’s degree and above; completion of secondary education; did not complete secondary) and parental occupation (high skill; medium skill; low skill or not in the labour force). Baseline levels of material, psychosocial and behavioural variables were recorded, categorized as described above, except for diet index and sleep quality, which were not measured in 2011. Mental health at baseline was measured using the MHI. Sequential causal mediation approach Mediation analysis aims to determine the extent to which an association between an exposure (here, incident disability) and an outcome (mental health) is due to the effect of the exposure on an intermediate variable (the mediator) which then influences the outcome. It aims to partition the total (causal) effect (TCE) of the exposure on the outcome into the effect that acts through the mediator, the indirect effect and the effect of exposure on outcome through mechanisms other than those that involve the mediator, the direct effect (‘direct’ in the sense that it by-passes the putative mediator). We sought to decompose the effect of disability acquisition on mental health into natural direct effects (NDE) and natural indirect effects (NIE) through material, psychosocial and behavioural factors using a sequential approach to causal mediation analysis (further details in Supplementary File 1, available as Supplementary Data at IJE online).48 This approach allows for mediation analysis through multiple causally related mediators and accommodates exposure–mediator interactions, one of the main sources of potential bias of the traditional approach to mediation. Based on our assumptions about the causal ordering of the mediators, this approach enabled us to estimate, in Model 1, the NIE through material factors (including paths that act through causal descendants of material factors but excluding paths that act only through psychosocial and/or behavioural factors), in Model 2, the NIE through both material and psychosocial factors (and through their causal descendants but excluding the path that acts only through behavioural factors) and, in Model 3, the NIE through material, psychosocial and behavioural factors, consisting of all possible paths except for the ‘direct’ path from exposure to outcome (Figure 2). Figure 2 View largeDownload slide Simplified causal diagrams illustrating estimated paths in Models 1–3, the NDE illustrated by the black lines (-) and the NIE by the dashed lines (- -) (A, disability acquisition (exposure of interest); Y, mental health (outcome); Mediators—M1, material factors; M2, psychosocial factors; M3, behavioural factors). Figure 2 View largeDownload slide Simplified causal diagrams illustrating estimated paths in Models 1–3, the NDE illustrated by the black lines (-) and the NIE by the dashed lines (- -) (A, disability acquisition (exposure of interest); Y, mental health (outcome); Mediators—M1, material factors; M2, psychosocial factors; M3, behavioural factors). Statistical analysis We used a weighting approach to estimate the marginal TCE, NDE and NIE for each set of mediators (further details in Supplementary File 1, available as Supplementary Data at IJE online). Inverse probability weighting was used to achieve exchangeability between the comparison groups and thus to account for possible confounding of the exposure–mediator and exposure–outcome associations by measured covariates.48–50 The MHI was modelled as a continuously valued outcome using linear regression models with and without the mediators, including all baseline variables as covariates. Interactions were included between the exposure and mediator variable if removal of an interaction term substantially changed the estimates of the NDE and NIE,51 measured as a change in the estimate of greater than half a standard error. Bootstrapping with 200 replications was used to calculate 95% confidence intervals (CIs). Missing data There were missing observations for the outcome, as well as several baseline covariates and mediators (Table S2.1, Supplementary File 2, available as Supplementary Data at IJE online). The distribution of baseline covariates was compared between participants with and without missing observations to determine whether missingness was associated with the values of measured variables. Participants with missing data had poorer mental health and greater socio-economic disadvantage across all measures compared with those with complete data (Table S2.2, Supplementary File 2, available as Supplementary Data at IJE online), suggesting that the data were not missing completely at random. Multiple imputation (MI) using chained equations with 50 imputations was performed to optimize the validity of the findings. The imputation models included all variables in the target analysis as well as additional auxiliary variables (Table S2.3, Supplementary File 2, available as Supplementary Data at IJE online). The sequential mediation analysis was conducted on each of the 50 imputed datasets and the mean of the estimates from each imputed dataset was calculated to give an overall MI estimate of the NDE and NIE. Standard errors were derived using Rubin rules for combining the between-imputation and within-imputation variance (obtained by bootstrapping the NDE and NIE estimates).52 Sensitivity analyses Three sensitivity analyses were conducted to test the robustness of findings. First, we performed a bias analysis for unmeasured confounding, which assessed the sensitivity of the results to unmeasured confounding of the mediator–outcome association, positing a range of plausible values for the strength of association of the potential confounder with mental health and the difference in prevalence of this confounder between those with and without disability (further details in Supplementary File 3, available as Supplementary Data at IJE online).53 Second, we removed participants with psychological impairments, defined as nervous or emotional conditions that require treatment, or any mental illness that requires help or supervision, as the effect of acquiring a psychological impairment on a general mental health score is likely to be different to other types of impairments. Third, we conducted a complete case analysis. Results Of the 28 794 people who participated in at least one wave of HILDA between 2001 and 2014, 14 534 participated in all four waves 2011 to 2014 and 14 518 of these (99.9%) responded to the disability question in all four waves. A total of 8323 individuals satisfied the definition of disability acquisition or reported no disability in any of the four waves, making them eligible for inclusion in the analysis (Figure S2.1, Supplementary File 2, available as Supplementary Data at IJE online). Complete data for all baseline covariates, mediators and mental health score were available for 4305 individuals (52% of the eligible sample). Baseline characteristics At baseline, people who went on to acquire a disability were older than those without disability (mean age of 53 vs 41 years, Table 2). They had poorer education, with 33.6% not completing secondary education compared with 24.9% of those without disability, were more likely to be unemployed or not in the labour force (38.8 vs 22.7%), had a lower mean weekly income (AU$834 vs AU$987) and experienced greater financial hardship (34.1 vs 25.1% reported being very poor or just getting by). People with disabilities were more likely to be in a relationship (71.3 vs 65.3%) and have children (72.9 vs 59.4%), more likely to be current (21.2 vs 17.2%) or ex-smokers (31.8 vs 23.7%), less likely to exercise regularly (34.2 vs 37.0%) and had higher mean BMI (27.4 vs 25.8 kg/m2). At baseline, they also reported poorer mental health than those without disability (mean MHI score of 73.3 vs 77.6). Table 2 Distribution of baseline characteristics for people who acquired a disability and the control sample (n=8323) Disability No disability n=387 n=7936 n % N % Age, years (mean (SD)) 387 52.5 (18.1) 7936 41.2 (15.4) Sex  Men 193 49.9 3817 48.1  Women 194 50.1 4119 51.9 Country of birth  Australia 297 76.7 6257 78.8  Other 90 23.3 1679 21.2 Parent occupation  High skill 181 47.4 3988 51.1  Medium skill 129 33.8 2645 33.9  Low skill/never worked 72 18.9 1168 15.0  Missing n=5 n=135 Education  Bachelor or higher 66 17.1 2184 27.5  Secondary, certificate, diploma 191 49.4 3777 47.6  Did not complete secondary 130 33.6 1975 24.9 Occupation  High skill 75 19.4 2373 29.9  Medium skill 97 25.1 2395 30.2  Low skill 65 16.8 1364 17.2  Unemployed/not in the labour force 150 38.8 1799 22.7  Missing n=0 n=5 Income, weekly $AUD (mean (SD)) 387 833.9 (476.8) 7936 986.9 (496.5) Wealth  High 138 35.7 2952 37.2  Medium 130 33.6 2616 33.0  Low 119 30.8 2368 29.8 Financial hardship  Prosperous/very comfortable 42 12.0 1444 20.3  Reasonably comfortable 188 53.9 3874 54.6  Just getting by/very poor 119 34.1 1781 25.1  Missing n=38 n=837 Financial satisfaction [mean (SD)]a 387 6.4 (2.3) 7931 6.7 (2.0)  Missing n=0 n=5 Housing tenure  Outright owner 148 38.2 2107 26.6  Mortgager 126 32.6 3393 42.8  Other 113 29.2 2425 30.6  Missing n=0 n=11 Housing affordability  Affordable 348 91.1 7263 92.4  Unaffordable 34 8.9 597 7.6  Missing n=5 n=76 Relationship  Yes 276 71.3 5173 65.3  No 111 28.7 2755 34.8  Missing n=0 n=8 Children  No 105 27.1 3224 40.6  Yes 282 72.9 4712 59.4 Social support [mean (SD)]b 347 5.3 (1.1) 7017 5.6 (1.0)  Missing n=40 n=919 Frequency of socializing [mean (SD)]c 347 3.8 (1.6) 7068 3.4 (1.4)  Missing n=40 n=868 Relationship with parents [mean (SD)]d 197 7.9 (2.2) 5833 8.1 (2.0)  Missing n=190 n=2103 Alcohol consumption  Never 59 16.9 1108 15.6  Rarely 129 36.9 2540 35.8  One or two times/week 62 17.7 1547 21.8  At least three times/week 100 28.6 1909 26.9  Missing n=37 n=832 Smoking  Never smoked 164 47.0 4210 59.2  Ex-smoker 111 31.8 1684 23.7  Current 74 21.2 1222 17.2  Missing n=38 n=820 Physical activity  At least four times/week 120 34.2 2638 37.0  One to three times/week 137 39.0 2998 42.1  Less than once/week 94 26.8 1493 20.9  Missing n=36 n=807 BMI, kg/m2 [mean (SD)] 332 27.4 (5.3) 6855 25.8 (5.0)  Missing n=55 n=1081 Mental health inventory (MHI) [mean (SD)]e 351 73.3 (18.9) 7125 77.6 (14.4)  Missing n=36 n=811 Disability No disability n=387 n=7936 n % N % Age, years (mean (SD)) 387 52.5 (18.1) 7936 41.2 (15.4) Sex  Men 193 49.9 3817 48.1  Women 194 50.1 4119 51.9 Country of birth  Australia 297 76.7 6257 78.8  Other 90 23.3 1679 21.2 Parent occupation  High skill 181 47.4 3988 51.1  Medium skill 129 33.8 2645 33.9  Low skill/never worked 72 18.9 1168 15.0  Missing n=5 n=135 Education  Bachelor or higher 66 17.1 2184 27.5  Secondary, certificate, diploma 191 49.4 3777 47.6  Did not complete secondary 130 33.6 1975 24.9 Occupation  High skill 75 19.4 2373 29.9  Medium skill 97 25.1 2395 30.2  Low skill 65 16.8 1364 17.2  Unemployed/not in the labour force 150 38.8 1799 22.7  Missing n=0 n=5 Income, weekly $AUD (mean (SD)) 387 833.9 (476.8) 7936 986.9 (496.5) Wealth  High 138 35.7 2952 37.2  Medium 130 33.6 2616 33.0  Low 119 30.8 2368 29.8 Financial hardship  Prosperous/very comfortable 42 12.0 1444 20.3  Reasonably comfortable 188 53.9 3874 54.6  Just getting by/very poor 119 34.1 1781 25.1  Missing n=38 n=837 Financial satisfaction [mean (SD)]a 387 6.4 (2.3) 7931 6.7 (2.0)  Missing n=0 n=5 Housing tenure  Outright owner 148 38.2 2107 26.6  Mortgager 126 32.6 3393 42.8  Other 113 29.2 2425 30.6  Missing n=0 n=11 Housing affordability  Affordable 348 91.1 7263 92.4  Unaffordable 34 8.9 597 7.6  Missing n=5 n=76 Relationship  Yes 276 71.3 5173 65.3  No 111 28.7 2755 34.8  Missing n=0 n=8 Children  No 105 27.1 3224 40.6  Yes 282 72.9 4712 59.4 Social support [mean (SD)]b 347 5.3 (1.1) 7017 5.6 (1.0)  Missing n=40 n=919 Frequency of socializing [mean (SD)]c 347 3.8 (1.6) 7068 3.4 (1.4)  Missing n=40 n=868 Relationship with parents [mean (SD)]d 197 7.9 (2.2) 5833 8.1 (2.0)  Missing n=190 n=2103 Alcohol consumption  Never 59 16.9 1108 15.6  Rarely 129 36.9 2540 35.8  One or two times/week 62 17.7 1547 21.8  At least three times/week 100 28.6 1909 26.9  Missing n=37 n=832 Smoking  Never smoked 164 47.0 4210 59.2  Ex-smoker 111 31.8 1684 23.7  Current 74 21.2 1222 17.2  Missing n=38 n=820 Physical activity  At least four times/week 120 34.2 2638 37.0  One to three times/week 137 39.0 2998 42.1  Less than once/week 94 26.8 1493 20.9  Missing n=36 n=807 BMI, kg/m2 [mean (SD)] 332 27.4 (5.3) 6855 25.8 (5.0)  Missing n=55 n=1081 Mental health inventory (MHI) [mean (SD)]e 351 73.3 (18.9) 7125 77.6 (14.4)  Missing n=36 n=811 a Satisfaction with financial circumstances, ranked using an 11-point Likert scale ranging from ‘totally dissatisfied’ to ‘totally satisfied’. b Constructed using the average of 10 questions addressing aspects of emotional support, each rated on a seven-point Likert scale. c Frequency of socializing with friends or relatives, rated on a seven-point Likert scale ranging from daily to less than once every 3 months. d Satisfaction with relationship with parents, rated on an 11-point Likert scale ranging from ‘completely dissatisfied’ to ‘completely satisfied’. e Measured using five questions from the SF-36, each of which is scored using five response categories, and the total scores are transformed into a scale ranging from 0 to 100, with higher scores reflecting better mental health. Table 2 Distribution of baseline characteristics for people who acquired a disability and the control sample (n=8323) Disability No disability n=387 n=7936 n % N % Age, years (mean (SD)) 387 52.5 (18.1) 7936 41.2 (15.4) Sex  Men 193 49.9 3817 48.1  Women 194 50.1 4119 51.9 Country of birth  Australia 297 76.7 6257 78.8  Other 90 23.3 1679 21.2 Parent occupation  High skill 181 47.4 3988 51.1  Medium skill 129 33.8 2645 33.9  Low skill/never worked 72 18.9 1168 15.0  Missing n=5 n=135 Education  Bachelor or higher 66 17.1 2184 27.5  Secondary, certificate, diploma 191 49.4 3777 47.6  Did not complete secondary 130 33.6 1975 24.9 Occupation  High skill 75 19.4 2373 29.9  Medium skill 97 25.1 2395 30.2  Low skill 65 16.8 1364 17.2  Unemployed/not in the labour force 150 38.8 1799 22.7  Missing n=0 n=5 Income, weekly $AUD (mean (SD)) 387 833.9 (476.8) 7936 986.9 (496.5) Wealth  High 138 35.7 2952 37.2  Medium 130 33.6 2616 33.0  Low 119 30.8 2368 29.8 Financial hardship  Prosperous/very comfortable 42 12.0 1444 20.3  Reasonably comfortable 188 53.9 3874 54.6  Just getting by/very poor 119 34.1 1781 25.1  Missing n=38 n=837 Financial satisfaction [mean (SD)]a 387 6.4 (2.3) 7931 6.7 (2.0)  Missing n=0 n=5 Housing tenure  Outright owner 148 38.2 2107 26.6  Mortgager 126 32.6 3393 42.8  Other 113 29.2 2425 30.6  Missing n=0 n=11 Housing affordability  Affordable 348 91.1 7263 92.4  Unaffordable 34 8.9 597 7.6  Missing n=5 n=76 Relationship  Yes 276 71.3 5173 65.3  No 111 28.7 2755 34.8  Missing n=0 n=8 Children  No 105 27.1 3224 40.6  Yes 282 72.9 4712 59.4 Social support [mean (SD)]b 347 5.3 (1.1) 7017 5.6 (1.0)  Missing n=40 n=919 Frequency of socializing [mean (SD)]c 347 3.8 (1.6) 7068 3.4 (1.4)  Missing n=40 n=868 Relationship with parents [mean (SD)]d 197 7.9 (2.2) 5833 8.1 (2.0)  Missing n=190 n=2103 Alcohol consumption  Never 59 16.9 1108 15.6  Rarely 129 36.9 2540 35.8  One or two times/week 62 17.7 1547 21.8  At least three times/week 100 28.6 1909 26.9  Missing n=37 n=832 Smoking  Never smoked 164 47.0 4210 59.2  Ex-smoker 111 31.8 1684 23.7  Current 74 21.2 1222 17.2  Missing n=38 n=820 Physical activity  At least four times/week 120 34.2 2638 37.0  One to three times/week 137 39.0 2998 42.1  Less than once/week 94 26.8 1493 20.9  Missing n=36 n=807 BMI, kg/m2 [mean (SD)] 332 27.4 (5.3) 6855 25.8 (5.0)  Missing n=55 n=1081 Mental health inventory (MHI) [mean (SD)]e 351 73.3 (18.9) 7125 77.6 (14.4)  Missing n=36 n=811 Disability No disability n=387 n=7936 n % N % Age, years (mean (SD)) 387 52.5 (18.1) 7936 41.2 (15.4) Sex  Men 193 49.9 3817 48.1  Women 194 50.1 4119 51.9 Country of birth  Australia 297 76.7 6257 78.8  Other 90 23.3 1679 21.2 Parent occupation  High skill 181 47.4 3988 51.1  Medium skill 129 33.8 2645 33.9  Low skill/never worked 72 18.9 1168 15.0  Missing n=5 n=135 Education  Bachelor or higher 66 17.1 2184 27.5  Secondary, certificate, diploma 191 49.4 3777 47.6  Did not complete secondary 130 33.6 1975 24.9 Occupation  High skill 75 19.4 2373 29.9  Medium skill 97 25.1 2395 30.2  Low skill 65 16.8 1364 17.2  Unemployed/not in the labour force 150 38.8 1799 22.7  Missing n=0 n=5 Income, weekly $AUD (mean (SD)) 387 833.9 (476.8) 7936 986.9 (496.5) Wealth  High 138 35.7 2952 37.2  Medium 130 33.6 2616 33.0  Low 119 30.8 2368 29.8 Financial hardship  Prosperous/very comfortable 42 12.0 1444 20.3  Reasonably comfortable 188 53.9 3874 54.6  Just getting by/very poor 119 34.1 1781 25.1  Missing n=38 n=837 Financial satisfaction [mean (SD)]a 387 6.4 (2.3) 7931 6.7 (2.0)  Missing n=0 n=5 Housing tenure  Outright owner 148 38.2 2107 26.6  Mortgager 126 32.6 3393 42.8  Other 113 29.2 2425 30.6  Missing n=0 n=11 Housing affordability  Affordable 348 91.1 7263 92.4  Unaffordable 34 8.9 597 7.6  Missing n=5 n=76 Relationship  Yes 276 71.3 5173 65.3  No 111 28.7 2755 34.8  Missing n=0 n=8 Children  No 105 27.1 3224 40.6  Yes 282 72.9 4712 59.4 Social support [mean (SD)]b 347 5.3 (1.1) 7017 5.6 (1.0)  Missing n=40 n=919 Frequency of socializing [mean (SD)]c 347 3.8 (1.6) 7068 3.4 (1.4)  Missing n=40 n=868 Relationship with parents [mean (SD)]d 197 7.9 (2.2) 5833 8.1 (2.0)  Missing n=190 n=2103 Alcohol consumption  Never 59 16.9 1108 15.6  Rarely 129 36.9 2540 35.8  One or two times/week 62 17.7 1547 21.8  At least three times/week 100 28.6 1909 26.9  Missing n=37 n=832 Smoking  Never smoked 164 47.0 4210 59.2  Ex-smoker 111 31.8 1684 23.7  Current 74 21.2 1222 17.2  Missing n=38 n=820 Physical activity  At least four times/week 120 34.2 2638 37.0  One to three times/week 137 39.0 2998 42.1  Less than once/week 94 26.8 1493 20.9  Missing n=36 n=807 BMI, kg/m2 [mean (SD)] 332 27.4 (5.3) 6855 25.8 (5.0)  Missing n=55 n=1081 Mental health inventory (MHI) [mean (SD)]e 351 73.3 (18.9) 7125 77.6 (14.4)  Missing n=36 n=811 a Satisfaction with financial circumstances, ranked using an 11-point Likert scale ranging from ‘totally dissatisfied’ to ‘totally satisfied’. b Constructed using the average of 10 questions addressing aspects of emotional support, each rated on a seven-point Likert scale. c Frequency of socializing with friends or relatives, rated on a seven-point Likert scale ranging from daily to less than once every 3 months. d Satisfaction with relationship with parents, rated on an 11-point Likert scale ranging from ‘completely dissatisfied’ to ‘completely satisfied’. e Measured using five questions from the SF-36, each of which is scored using five response categories, and the total scores are transformed into a scale ranging from 0 to 100, with higher scores reflecting better mental health. Sequential causal mediation analysis Interactions between the exposure and the following mediator variables were included in the regression models: material factors including occupation, housing affordability, housing tenure and satisfaction with financial circumstances; psychosocial factors including social support, frequency of socializing and relationship status; and behavioural factors including smoking, alcohol consumption, physical activity, BMI and diet. The TCE of disability acquisition was estimated to be a 5.3-point reduction in MHI score (95% CI –6.8, –3.7) (Table 3). In the sequential approach, we first considered the mediated effect through material factors and estimated a mean 1.7-point decline (95% CI –2.8, –0.6) in MHI was occurring through material factors, which corresponds to 32.1% of the total effect. We then considered the additional effect of psychosocial factors and found that 33.2% was explained by both material and psychosocial factors (NIE: –1.7, 95% CI –3.0, –0.5) and the additional effect of behavioural factors explained 38.6% of the decline (NIE: –2.0, 95% CI –3.4, –0.6). Table 3 Total causal effect (TCE), natural direct effect (NDE) and natural indirect effect (NIE) of disability acquisition on mental health, with mediation through material factors, psychosocial and behavioural factors Material factors + psychosocial factors + behavioural factors Coef.a (95% CI) Coef.a (95% CI) Coef.a (95% CI) TCE –5.3 (–6.8, –3.7) –5.3 (–6.8, –3.7) –5.3 (–6.8, –3.7) NDE –3.6 (–5.4, –1.8) –3.5 (–5.3, –1.7) –3.2 (–5.1, –1.4) NIE –1.7 (–2.8, –0.6) –1.7 (–3.0, –0.5) –2.0 (–3.4, –0.6) Proportion of effect explained (%) 32.1 (10.1, 54.1) 33.2 (8.5, 58.0) 38.6 (11.4, 65.9) Material factors + psychosocial factors + behavioural factors Coef.a (95% CI) Coef.a (95% CI) Coef.a (95% CI) TCE –5.3 (–6.8, –3.7) –5.3 (–6.8, –3.7) –5.3 (–6.8, –3.7) NDE –3.6 (–5.4, –1.8) –3.5 (–5.3, –1.7) –3.2 (–5.1, –1.4) NIE –1.7 (–2.8, –0.6) –1.7 (–3.0, –0.5) –2.0 (–3.4, –0.6) Proportion of effect explained (%) 32.1 (10.1, 54.1) 33.2 (8.5, 58.0) 38.6 (11.4, 65.9) a These primary analysis results were obtained using multiple imputation using chained equations with 50 imputed datasets. Table 3 Total causal effect (TCE), natural direct effect (NDE) and natural indirect effect (NIE) of disability acquisition on mental health, with mediation through material factors, psychosocial and behavioural factors Material factors + psychosocial factors + behavioural factors Coef.a (95% CI) Coef.a (95% CI) Coef.a (95% CI) TCE –5.3 (–6.8, –3.7) –5.3 (–6.8, –3.7) –5.3 (–6.8, –3.7) NDE –3.6 (–5.4, –1.8) –3.5 (–5.3, –1.7) –3.2 (–5.1, –1.4) NIE –1.7 (–2.8, –0.6) –1.7 (–3.0, –0.5) –2.0 (–3.4, –0.6) Proportion of effect explained (%) 32.1 (10.1, 54.1) 33.2 (8.5, 58.0) 38.6 (11.4, 65.9) Material factors + psychosocial factors + behavioural factors Coef.a (95% CI) Coef.a (95% CI) Coef.a (95% CI) TCE –5.3 (–6.8, –3.7) –5.3 (–6.8, –3.7) –5.3 (–6.8, –3.7) NDE –3.6 (–5.4, –1.8) –3.5 (–5.3, –1.7) –3.2 (–5.1, –1.4) NIE –1.7 (–2.8, –0.6) –1.7 (–3.0, –0.5) –2.0 (–3.4, –0.6) Proportion of effect explained (%) 32.1 (10.1, 54.1) 33.2 (8.5, 58.0) 38.6 (11.4, 65.9) a These primary analysis results were obtained using multiple imputation using chained equations with 50 imputed datasets. Sensitivity analysis The results were robust to the changes implied by the scenarios in the sensitivity analyses. The bias analysis demonstrated that the estimated indirect effects were unlikely to be explained by unmeasured confounding (Supplementary File 3, available as Supplementary Data at IJE online). Removing disabled people with psychological impairments (41 of 387) attenuated the effect estimates; however, the proportion of the effect mediated increased slightly. For the complete case analysis, only small changes in the magnitude of individual coefficients were observed (Table 4). Table 4 Results of the sensitivity analyses showing total causal effect (TCE), natural direct effect (NDE) and natural indirect effect (NIE) of disability acquisition on mental health, with mediation through material, psychosocial and behavioural factors Material factors + psychosocial factors + behavioural factors Coef. (95% CI) Coef. (95% CI) Coef. (95% CI) Psychological impairments removeda  TCE –4.3 (–5.9, –2.7) –4.3 (–5.9, –2.7) –4.3 (–5.9, –2.7)  NDE –2.7 (–4.6, –0.8) –2.6 (–4.4, –0.8) –2.2 (–4.0, –0.3)  NIE –1.6 (–2.7, –0.5) –1.7 (–3.0, –0.5) –2.1 (–3.5, –0.8)  Proportion of effect explained (%) 37.7 (7.0, 68.4) 40.2 (8.8, 71.6) 49.8 (14.6, 84.9) Complete case analysis  TCE –5.1 (–7.7, –2.5) –5.1 (–7.7, –2.5) –5.1 (–7.7, –2.5)  NDE –3.3 (–5.8, –0.8) –3.2 (–5.7, –0.7) –3.1 (–5.7, –0.5)  NIE –1.7 (–3.6, 0.1) –1.9 (–4.1, 0.3) –2.0 (–4.5, 0.6)  Proportion of effect explained (%) 34.4 (0.2, 68.7) 36.9 (–4.4, 78.2) 38.9 (–10.0, 87.8) Material factors + psychosocial factors + behavioural factors Coef. (95% CI) Coef. (95% CI) Coef. (95% CI) Psychological impairments removeda  TCE –4.3 (–5.9, –2.7) –4.3 (–5.9, –2.7) –4.3 (–5.9, –2.7)  NDE –2.7 (–4.6, –0.8) –2.6 (–4.4, –0.8) –2.2 (–4.0, –0.3)  NIE –1.6 (–2.7, –0.5) –1.7 (–3.0, –0.5) –2.1 (–3.5, –0.8)  Proportion of effect explained (%) 37.7 (7.0, 68.4) 40.2 (8.8, 71.6) 49.8 (14.6, 84.9) Complete case analysis  TCE –5.1 (–7.7, –2.5) –5.1 (–7.7, –2.5) –5.1 (–7.7, –2.5)  NDE –3.3 (–5.8, –0.8) –3.2 (–5.7, –0.7) –3.1 (–5.7, –0.5)  NIE –1.7 (–3.6, 0.1) –1.9 (–4.1, 0.3) –2.0 (–4.5, 0.6)  Proportion of effect explained (%) 34.4 (0.2, 68.7) 36.9 (–4.4, 78.2) 38.9 (–10.0, 87.8) a These sensitivity analysis results were obtained using multiple imputation using chained equations with 50 imputed datasets. Table 4 Results of the sensitivity analyses showing total causal effect (TCE), natural direct effect (NDE) and natural indirect effect (NIE) of disability acquisition on mental health, with mediation through material, psychosocial and behavioural factors Material factors + psychosocial factors + behavioural factors Coef. (95% CI) Coef. (95% CI) Coef. (95% CI) Psychological impairments removeda  TCE –4.3 (–5.9, –2.7) –4.3 (–5.9, –2.7) –4.3 (–5.9, –2.7)  NDE –2.7 (–4.6, –0.8) –2.6 (–4.4, –0.8) –2.2 (–4.0, –0.3)  NIE –1.6 (–2.7, –0.5) –1.7 (–3.0, –0.5) –2.1 (–3.5, –0.8)  Proportion of effect explained (%) 37.7 (7.0, 68.4) 40.2 (8.8, 71.6) 49.8 (14.6, 84.9) Complete case analysis  TCE –5.1 (–7.7, –2.5) –5.1 (–7.7, –2.5) –5.1 (–7.7, –2.5)  NDE –3.3 (–5.8, –0.8) –3.2 (–5.7, –0.7) –3.1 (–5.7, –0.5)  NIE –1.7 (–3.6, 0.1) –1.9 (–4.1, 0.3) –2.0 (–4.5, 0.6)  Proportion of effect explained (%) 34.4 (0.2, 68.7) 36.9 (–4.4, 78.2) 38.9 (–10.0, 87.8) Material factors + psychosocial factors + behavioural factors Coef. (95% CI) Coef. (95% CI) Coef. (95% CI) Psychological impairments removeda  TCE –4.3 (–5.9, –2.7) –4.3 (–5.9, –2.7) –4.3 (–5.9, –2.7)  NDE –2.7 (–4.6, –0.8) –2.6 (–4.4, –0.8) –2.2 (–4.0, –0.3)  NIE –1.6 (–2.7, –0.5) –1.7 (–3.0, –0.5) –2.1 (–3.5, –0.8)  Proportion of effect explained (%) 37.7 (7.0, 68.4) 40.2 (8.8, 71.6) 49.8 (14.6, 84.9) Complete case analysis  TCE –5.1 (–7.7, –2.5) –5.1 (–7.7, –2.5) –5.1 (–7.7, –2.5)  NDE –3.3 (–5.8, –0.8) –3.2 (–5.7, –0.7) –3.1 (–5.7, –0.5)  NIE –1.7 (–3.6, 0.1) –1.9 (–4.1, 0.3) –2.0 (–4.5, 0.6)  Proportion of effect explained (%) 34.4 (0.2, 68.7) 36.9 (–4.4, 78.2) 38.9 (–10.0, 87.8) a These sensitivity analysis results were obtained using multiple imputation using chained equations with 50 imputed datasets. Discussion Interpretation of findings In this analysis, we found that 32% of the effect of disability acquisition on mental health was mediated by material factors, with only a negligible proportion explained by the addition of psychosocial factors and 5% by behavioural factors. This is consistent with the majority of the literature explaining health inequalities, which found that health differences are predominantly attributable to material factors.11,17 The results were not consistent with studies that had shown that psychosocial resources accounted for some of the effect of disability on depression;6,19,20 however, these pathways are not mutually exclusive and it is possible that a large proportion of the effect through material factors is also operating through psychosocial pathways. Previous studies did not use a sequential causal mediation approach, which allows estimation of the additional contribution of psychosocial factors beyond the effect that is operating through material factors.48 The effect sizes estimated in this study were of clinical significance. Study participants who acquired a disability experienced on average a five-point decline in mental health, exceeding the four- to five-point difference considered to represent a clinically meaningful change.29,39,40 The effect mediated through material factors was estimated to be 32.1%, which can be interpreted as the proportion of the mental health decline that could be avoided if people with disabilities experienced the same material socio-economic circumstances as those without disabilities. About two-fifths of the effect (38.6%) was explained by all three sets of mediators, leaving a large proportion of the effect unexplained—it seems unlikely that the remaining 61.4% of the total effect is not mediated by any other factors and is therefore a true ‘direct’ effect. This is perhaps not surprising as, despite measuring a broad range of socio-economic characteristics, these measures capture only a snapshot of people’s socio-economic experiences at one point in time54 and do not capture the broader structural, political and economic processes they experience.11 Additionally, there were some factors that were not recorded in HILDA that could be important mediators, such as experience of discrimination, sense of personal control (asked only in 2011 and 2015), psychosocial working conditions and personal–work balance (asked only for those people who were employed). Therefore, the effect operating through psychosocial pathways may be underestimated. Strengths and limitations This study used data from a large longitudinal survey in Australia. The longitudinal nature of the data meant that we could characterize disability acquisition, based on a sample of people who reported no disability for two waves followed immediately by two waves of disability. Furthermore, we could measure disability acquisition, mediators and mental health at different time points, to establish a temporal sequence between them, and control for prior values of the mediator and mental health score so that the results can be interpreted as effects of changes in the mediators on the outcome. We used causal sequential mediation methods, which can address the limitations of traditional mediation methods, generating unbiased estimates of mediation through multiple causally ordered mediators, given a set of clearly specified assumptions of no confounding. There were also limitations with this study. The analysis rests on several strong assumptions about no confounding between disability acquisition, mediators and mental health. We used inverse probability weighting to account for (measured) confounding of the disability–mental health and the disability–mediator relationships. For the assumption relating to no uncontrolled confounding of the mediator–outcome relationship, we conducted a bias analysis which suggested that the NDE and NIE were unlikely to be explained by confounding by measured or unmeasured variables. The weighting approach is sensitive to outcome model misspecification, which can lead to biased estimates of natural direct and indirect effects; however, this approach was deemed most appropriate because of the large number of mediators.48 Furthermore, to ensure best specification of the outcome model, interactions between the exposure and each mediator were considered and tested. There were strong assumptions about the causal ordering of the mediators. The direction of causality between these contributory factors is likely to be bi-directional, e.g. the relationship between employment and social support. This may have led to overestimation of the proportion of the effect operating through material factors if these are consequences of psychosocial and behavioural factors, rather than a cause of them. However, for most of the variables considered, the effect is likely to be causally ordered from material to psychosocial to behavioural factors. There was a large proportion of missing data and this was higher in participants with poorer mental health and greater socio-economic disadvantage; however, the use of MI as the primary analysis should have reduced this selection bias. The concepts of disability and mental health are related, which makes it difficult to isolate the causal effect of one on the other. To address this limitation, first we chose to use the mental health subscale of the SF-36 health questionnaire (MHI), rather than the summary mental health score (MCS), therefore selecting parts of the SF-36 questionnaire that were less likely to overlap with the definition of disability. Furthermore, we conducted a sensitivity analysis in which we excluded people with psychological impairments to further minimize overlap between the concepts, which did not change the interpretation of the results though the magnitude of effect estimates was slightly attenuated. When we excluded people with psychological impairments, the proportion of the effect mediated was slightly larger. It is plausible that the mechanisms are different for people with psychological impairments compared with other types of disability, though the relative proportion through each of the three pathways was similar. It would be interesting to look at differences in these effects according to types of impairments; however, we lacked power to examine differences by disability characteristics. Finally, people with severe disabilities are less likely to participate in HILDA; therefore, our results are likely to underestimate the population effect of disability acquisition on mental health. Conclusions The finding that the effect of disability acquisition on mental health operates predominantly through material factors has important policy implications. These results highlight that social policy reforms that reduce socio-economic disadvantage among people who acquire a disability will improve mental health. This could be achieved through better social protection, including income support, but also through improved educational and employment opportunities for people with disabilities and access to affordable housing. It is important to further disentangle the mechanisms involved in the material pathway, to better understand the relative importance of specific factors and which social determinants are driving the mental health inequalities. This will help to better target policy interventions to improved the mental health of people with disabilities. Supplementary Data Supplementary data are available at IJE online. Funding This work was supported by an Australian Government Research Training Program Scholarship and a National Health and Medical Research Council Postgraduate Scholarship (1093740 to Z.A.), a National Health and Medical Research Council Senior Research Fellowship (1104975 to J.A.S.) and an Australian Research Council Future Fellowship (FT150100131 to R.B.). Conflict of interest: The authors have no conflicts of interest to declare. References 1 World Health Organization and World Bank Group . World Report on Disability . Geneva : WHO , 2011 . 2 Kavanagh AM , Aitken Z , Baker E et al. Housing tenure and affordability and mental health following disability acquisition in adulthood . Soc Sci Med 2016 ; 151 : 225 – 32 . Google Scholar CrossRef Search ADS PubMed 3 Kavanagh AM , Aitken Z , Krnjacki L , LaMontagne AD , Bentley R , Milner A. Mental health following acquisition of disability in adulthood: the impact of wealth . PLoS One 2015 ; 10 : e0139708 . Google Scholar CrossRef Search ADS PubMed 4 Mandemakers JJ , Monden CWS. Does education buffer the impact of disability on psychological distress? Soc Sci Med 2010 ; 71 : 288 – 97 . Google Scholar CrossRef Search ADS PubMed 5 Smith DM , Langa KM , Kabeto MU , Ubel PA. Health, wealth, and happiness: financial resources buffer subjective well-being after the onset of a disability . Psychol Sci 2005 ; 16 : 663 – 66 . Google Scholar CrossRef Search ADS PubMed 6 Yang Y. How does functional disability affect depressive symptoms in late life? The role of perceived social support and psychological resources . J Health Soc Behav 2006 ; 47 : 355 – 72 . Google Scholar CrossRef Search ADS PubMed 7 Kariuki M , Honey A , Emerson E , Llewellyn G. Mental health trajectories of young people after disability onset . Disabil Health J 2011 ; 4 : 91 – 101 . Google Scholar CrossRef Search ADS PubMed 8 Emerson E , Kariuki M , Honey A , Llewellyn G. Becoming disabled: the association between disability onset in younger adults and subsequent changes in productive engagement, social support, financial hardship and subjective wellbeing . Disabil Health J 2014 ; 7 : 448 – 56 . Google Scholar CrossRef Search ADS PubMed 9 Yang Y , George LK. Functional disability, disability transitions, and depressive symptoms in late life . J Aging Health 2005 ; 17 : 263 – 92 . Google Scholar CrossRef Search ADS PubMed 10 Lynch J. Income inequality and health: expanding the debate . Soc Sci Med 2000 ; 51 : 1001 – 05 . Google Scholar CrossRef Search ADS PubMed 11 Lynch JW , Smith GD , Kaplan GA , House JS. Income inequality and mortality: importance to health of individual income, psychosocial environment, or material conditions . Brit Med J 2000 ; 320 : 1200 – 04 . Google Scholar CrossRef Search ADS PubMed 12 Marmot M , Wilkinson R. Social Determinants of Health . Oxford : Oxford University Press , 1999 . 13 Bartley M. Health Inequality: An Introduction to Theories, Concepts and Methods . Cambridge : Polity Press , 2004 . 14 DHSS . Inequalities in Health: Report of a Research Working Group . London : DHSS , 1980 . 15 Wilkinson RG. Socioeconomic determinants of health—health inequalities: relative or absolute material standards? Brit Med J 1997 ; 314 : 591 – 95 . Google Scholar CrossRef Search ADS PubMed 16 Pocock SJ , Cook DG , Shaper AG , Phillips AN , Walker M. Social-class differences in ischemic-heart-disease in British men . Lancet 1987 ; 2 : 197 – 201 . Google Scholar CrossRef Search ADS PubMed 17 Davey Smith G , Bartley M. Explanations for socio-economic differentials in mortality . Eur J Public Health 1994 ; 4 : 131 – 44 . Google Scholar CrossRef Search ADS 18 Power C , Matthews S. Origins of health inequalities in a national population sample . Lancet 1997 ; 350 : 1584 – 89 . Google Scholar CrossRef Search ADS PubMed 19 Kiviruusu O , Huurre T , Aro H. Psychosocial resources and depression among chronically ill young adults: are males more vulnerable? Soc Sci Med 2007 ; 65 : 173 – 86 . Google Scholar CrossRef Search ADS PubMed 20 Bisschop MI , Knegsman DMW , Beekman ATF , Deeg DJH. Chronic diseases and depression: the modifying role of psychosocial resources . Soc Sci Med 2004 ; 59 : 721 – 33 . Google Scholar CrossRef Search ADS PubMed 21 Hafeman DM , Schwartz S. Opening the Black Box: a motivation for the assessment of mediation . Int J Epidemiol 2009 ; 38 : 838 – 45 . Google Scholar CrossRef Search ADS PubMed 22 Jahoda M. Employment and Unemployment: A Social-Psychological Analysis . Cambridge : Cambridge University Press , 1982 . 23 Summerfield M , Freidin S , Hahn M et al. HILDA User Manual—Release 14. Melbourne : Melbourne Institute of Applied Economic and Social Research , 2015 . 24 Burchardt T. Being and Becoming: Social Exclusion and the Onset of Disability . London : The London School of Economics , 2003 . 25 Jenkins SP , Rigg JA. Disability and disadvantage: selection, onset, and duration effects . J Soc Policy 2004 ; 33 : 479 – 501 . Google Scholar CrossRef Search ADS 26 Polidano C , Vu H. Differential labour market impacts from disability onset . Health Econ 2015 ; 24 : 302 – 17 . Google Scholar CrossRef Search ADS PubMed 27 Burkhauser R , Daly M. Employment and economic well-being following the onset of a disability: the role for public policy. In: Mashaw J , Reno V , Burkhauser R , Berkowitz M (eds). Disability, Work, and Cash Benefits . Kalamazoo MI : Upjohn Institute for Employment Research , 1996 . Google Scholar CrossRef Search ADS 28 Butterworth P , Crosier T. The validity of the SF-36 in an Australian National Household Survey: demonstrating the applicability of the Household Income and Labour Dynamics in Australia (HILDA) Survey to examination of health inequalities . BMC Public Health 2004 ; 4 : 44 . Google Scholar CrossRef Search ADS PubMed 29 Ware JE Jr , Kosinski M , Gandek B. SF-36 Health Survey: Manual and Interpretation Guide . Lincoln, RI : QualityMetric , 2000 . 30 McCabe CJ , Thomas KJ , Brazier JE , Coleman P. Measuring the mental health status of a population: a comparison of the GHQ-12 and the SF-36 (MHI-5) . British Journal of Psychiatry: The Journal of Mental Science 1996 ; 169 : 516 – 21 . Google Scholar CrossRef Search ADS 31 Strand BH , Dalgard OS , Tambs K , Rognerud M. Measuring the mental health status of the Norwegian population: a comparison of the instruments SCL-25, SCL-10, SCL-5 and MHI-5 (SF-36) . Nord J Psychiat 2003 ; 57 : 113 – 81 . Google Scholar CrossRef Search ADS 32 Weinstein MC , Berwick DM , Goldman PA , Murphy JM , Barsky AJ. A comparison of three psychiatric screening tests using receiver operating characteristic (ROC) analysis . Med Care 1989 ; 27 : 593 – 607 . Google Scholar CrossRef Search ADS PubMed 33 Yamazaki S , Fukuhara S , Green J. Usefulness of five-item and three-item Mental Health Inventories to screen for depressive symptoms in the general population of Japan . Health Qual Life Out 2005 ; 3 : 48 . Google Scholar CrossRef Search ADS 34 Berwick DM , Murphy JM , Goldman PA , Ware JE Jr , Barsky AJ , Weinstein MC. Performance of a five-item mental health screening test . Med Care 1991 ; 29 : 169 – 76 . Google Scholar CrossRef Search ADS PubMed 35 Cuijpers P , Smits N , Donker T , ten Have M , de Graaf R. Screening for mood and anxiety disorders with the five-item, the three-item, and the two-item Mental Health Inventory . Psychiat Res 2009 ; 168 : 250 – 55 . Google Scholar CrossRef Search ADS 36 Means-Christensen AJ , Arnau RC , Tonidandel AM , Bramson R , Meagher MW. An efficient method of identifying major depression and panic disorder in primary care . J Behav Med 2005 ; 28 : 565 – 72 . Google Scholar CrossRef Search ADS PubMed 37 Rumpf HJ , Meyer C , Hapke U , John U. Screening for mental health: validity of the MHI-5 using DSM-IV Axis I psychiatric disorders as gold standard . Psychiat Res 2001 ; 105 : 243 – 53 . Google Scholar CrossRef Search ADS 38 Pirkis J , Burgess P , Kirk P , Dodson S , Coombs T. Review of Standardised Measures Used in the National Outcomes and Casemix Collection (NOCC) . Sydney : NSW Institute of Psychiatry , 2005 . 39 Knox SA , King MT. Validation and calibration of the SF-36 health transition question against an external criterion of clinical change in health status . Qual Life Res 2009 ; 18 : 637 – 45 . Google Scholar CrossRef Search ADS PubMed 40 Mchorney CA , Ware JE , Raczek AE . The Mos 36-Item Short-Form Health Survey (Sf-36). 2. Psychometric and clinical-tests of validity in measuring physical and mental-health constructs . Med Care 1993 ; 31 : 247 – 63 . Google Scholar CrossRef Search ADS PubMed 41 Moor I , Spallek J , Richter M. Explaining socioeconomic inequalities in self-rated health: a systematic review of the relative contribution of material, psychosocial and behavioural factors . J Epidemiol Commun H 2017 ; 71 : 565 – 75 . Google Scholar CrossRef Search ADS 42 Skalicka V , van Lenthe F , Bambra C , Krokstad S , Mackenbach J. Material, psychosocial, behavioural and biomedical factors in the explanation of relative socio-economic inequalities in mortality: evidence from the HUNT study . Int J Epidemiol 2009 ; 38 : 1272 – 84 . Google Scholar CrossRef Search ADS PubMed 43 van Oort FVA , van Lenthe FJ , Mackenbach JP. Material, psychosocial, and behavioural factors in the explanation of educational inequalities in mortality in the Netherlands . J Epidemiol Commun H 2005 ; 59 : 214 – 20 . Google Scholar CrossRef Search ADS 44 McMunn A , Nazroo J , Breeze E. Inequalities in health at older ages: a longitudinal investigation of the onset of illness and survival effects in England . Age Ageing 2009 ; 38 : 181 – 87 . Google Scholar CrossRef Search ADS PubMed 45 Emerson E, , Madden R, , Robertson J, , Graham H , Hatton C , Llewellyn G. Intellectual and Physical Disability, Social Mobility, Social Inclusion and Health: Background Paper for the Marmot Review . Lancaster : Centre for Disability Research, Lancaster University , 2009 . 46 World Health Organisation . The International Classification of Functioning, Disability and Health . Geneva : WHO , 2001 . 47 World Health Organization . Towards a Common Language for Functioning, Disability and Health: ICF the International Classification of Functioning, Disability and Health . Geneva : WHO , 2002 . 48 VanderWeele TJ , Vansteelandt S. Mediation analysis with multiple mediators . Epidemiologic Methods 2014 ; 2 : 95 – 115 . Google Scholar CrossRef Search ADS PubMed 49 Vansteelandt S , Bekaert M , Lange T. Imputation strategies for the estimation of natural direct and indirect effects . Epidemiologic Methods 2012 ; 1 : 131 – 58 . Google Scholar CrossRef Search ADS 50 Hernan MA , Robins JM. Estimating causal effects from epidemiological data . J Epidemiol Commun H 2006 ; 60 : 578 – 86 . Google Scholar CrossRef Search ADS 51 VanderWeele TJ. Explanation in Causal Inference: Methods for Mediation and Interaction . New York : Oxford University Press , 2015 . 52 Rubin D. Multiple Imputation for Nonresponse in Surveys . New York : Wiley , 1987 . Google Scholar CrossRef Search ADS 53 VanderWeele TJ. Bias formulas for sensitivity analysis for direct and indirect effects . Epidemiology 2010 ; 21 : 540 – 51 . Google Scholar CrossRef Search ADS PubMed 54 De Stavola BL , Daniel RM. Commentary: Incorporating concepts and methods from causal inference into life course epidemiology . Int J Epidemiol 2016 ; 45 : 1006 – 10 . Google Scholar CrossRef Search ADS PubMed © The Author(s) 2018; all rights reserved. Published by Oxford University Press on behalf of the International Epidemiological Association This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices)

Journal

International Journal of EpidemiologyOxford University Press

Published: Jan 29, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off