- Email: [email protected]
Housing tenure and affordability and mental health following disability acquisition in adulthood
Social Science & Medicine 151 (2016) 225e232
Contents lists available at ScienceDirect
Social Science & Medicine journal homepage: www.elsevier.com/locate/socscimed
Housing tenure and affordability and mental health following disability acquisition in adulthood Anne M. Kavanagh a, *, Zoe Aitken a, Emma Baker b, Anthony D. LaMontagne c, d, Allison Milner c, d, Rebecca Bentley a a
Gender and Women's Health, Centre for Health Equity, Melbourne School of Population and Global Health, The University of Melbourne, Australia School of Architecture & Built Environment, The University of Adelaide, Australia Population Health Strategic Research Centre, School of Health & Social Development, Deakin University, Australia d McCaughey VicHealth Centre for Community Wellbeing, Centre for Health Equity, Melbourne School of Population and Global Health, The University of Melbourne, Australia b c
a r t i c l e i n f o
a b s t r a c t
Article history: Received 22 July 2015 Received in revised form 19 October 2015 Accepted 7 January 2016 Available online 9 January 2016
Acquiring a disability in adulthood is associated with a reduction in mental health and access to secure and affordable housing is associated with better mental health. We hypothesised that the association between acquisition of disability and mental health is modified by housing tenure and affordability. We used twelve annual waves of data (2001e2012) (1913 participants, 13,037 observations) from the Household, Income and Labour Dynamics in Australia survey. Eligible participants reported at least two consecutive waves of disability preceded by two consecutive waves without disability. Effect measure modification, on the additive scale, was tested in three fixed-effects linear regression models (which remove time-invariant confounding) which included a cross-product term between disability and prior housing circumstances: housing tenure by disability; housing affordability by disability and, in a subsample (896 participants 5913 observations) with housing costs, tenure/affordability by disability. The outcome was the continuous mental component summary (MCS) of SF-36. Models adjusted for timevarying confounders. There was statistical evidence that prior housing modified the effect of disability acquisition on mental health. Our findings suggested that those in affordable housing had a 1.7 point deterioration in MCS (95% CI -2.1, -1.3) following disability acquisition and those in unaffordable housing had a 4.2 point reduction (95% CI -5.2, -1.4). Among people with housing costs, the largest declines in MCS were for people with unaffordable mortgages (5.3, 95% CI 8.8, 1.9) and private renters in unaffordable housing (4.0, 95% CI 6.3, 1.6), compared to a 1.4 reduction (95% CI 2.1, 0.7) for mortgagors in affordable housing. In sum, we used causally-robust fixed-effects regression and showed that deterioration in mental health following disability acquisition is modified by prior housing circumstance with the largest negative associations found for those in unaffordable housing. Future research should test whether providing secure, affordable housing when people acquire a disability prevents deterioration in mental health. © 2016 Published by Elsevier Ltd.
Keywords: Disability Mental health Housing tenure Housing affordability Longitudinal study Fixed-effects regression Effect measure modification
1. Introduction Internationally, people with disabilities e nearly 20% of the population e experience significant socio-economic disadvantage (Kavanagh et al., 2014; Kavanagh et al., 2013; World Health
* Corresponding author. Gender and Women's Health, Centre for Health Equity, Melbourne School of Population and Global Health, The University of Melbourne, Level 3, 207 Bouverie St, Carlton, 3010 Victoria, Australia. E-mail address: [email protected] (A.M. Kavanagh). http://dx.doi.org/10.1016/j.socscimed.2016.01.010 0277-9536/© 2016 Published by Elsevier Ltd.
Organization and World Bank Group, 2011). Disabled Australians have lower rates of employment, post-secondary education, and incomes and are more likely to experience housing-related disadvantage compared to those without a disability (Beer et al., 2012; Beer and Faulkner, 2008; Beer et al., 2011; Kavanagh et al., 2014, 2013; Parker and Fisher, 2010). They are over-represented in Australia's housing welfare sector (Dalton and Ong, 2007), are more likely to experience homelessness (Beer et al., 2012; Beer and Faulkner, 2008) and unaffordable housing (Kavanagh et al., 2014, 2013) e a situation found in other high-income countries (Kyle
226
A.M. Kavanagh et al. / Social Science & Medicine 151 (2016) 225e232
and Dunn, 2008; Papworth Trust, 2011; Wang, 2005; White et al., 1994). People with disabilities have poorer physical and mental health and some authors have argued that their disadvantaged circumstances (including inadequate housing) may play a significant role in explaining these health inequalities although the evidence is primarily limited to adolescents and young adults (E. Emerson and Hatton, 2007; E. Emerson, Llewellyn, Honey and Kariuki, 2012; Honey et al., 2011). A study of older adults (65 years and older) found that for those who reported a disability at baseline, social support modified the relationship between disability and depressive symptoms (Yang, 2006). On average, mental health tends to decline after acquisition of disability (Mandemakers and Monden, 2010; Turner and Noh, 1988), however this is not universal (Eric Emerson, Kariuki, Honey and Llewellyn, 2014; Honey et al., 2011); one small Australian study found that social support and financial stability prior to onset of disability protected against deterioration in mental health in young adults (Honey et al., 2011). A study of over 2000 Australian adults found that although being in the highest tertile of wealth before disability acquisition was associated with a small deterioration in mental health the deterioration was larger for those in the lowest and mid tertile of wealth (Kavanagh et al., 2015). Similar findings were found in a a UK study of adultswhere a greater reduction in subjective wellbeing was observed for people with below median wealth at time of disability acquisition (Smith et al., 2005). Elucidation of factors that predict deterioration in mental health with disability acquisition is of policy significance because it will enable better targeting of health and non-health sector interventions. Given housing's known relationship to health and its central role as a primary and regular household expenditure, it is an important potential modifier of the effect of disability acquisition on mental health e the focus of this paper. 1.1. Housing and health Housing is an important social determinant of health (Braubach, 2011; Gibson et al., 2011; Shaw, 2004), characteristics which have been linked to health include dwelling quality, location, tenure and affordability. There is a strong evidence-base linking structural characteristics of housing (e.g. toxins, damp, accessibility) (Evans et al., 2000; Free et al., 2010) to health including for people with disabilities (Imrie, 2006). Australian housing stock is of relatively good quality and because the majority of Australians live around the climatically mild coast, the structural aspects of housing are less important. Tenure type is strongly related to health, with outright owners and mortgagors having better health than people living in private rental or public housing (Ellaway and Macintyre, 1998; Macintyre et al., 2003; Pollack et al., 2004); however, whether this relationship is causal has been debated (E. Baker, Bentley and Mason, 2013). In Australia, an association between housing tenure and mental health was not found in causally-robust fixedeffects longitudinal regression analyses (E. Baker et al., 2013). In the context of rapidly increasing house prices and increasing household debt relative to income (ABS, 2014), Australia currently has one of the most unaffordable housing markets in the world (Demographia, 2014), making housing affordability an increasingly important determinant of health. Two longitudinal studies in the United Kingdom and Australia have demonstrated deleterious associations between living in unaffordable housing and mental health independent of general financial hardship (R. Bentley, Baker, Mason, Subramanian and Kavanagh, 2011; Pevalin et al., 2008). However this relationship appears to be modified by tenure type with recent research suggesting that this association is stronger for private renters than home-owners in Australia (Mason et al., 2013; C. Pollack, Griffin and Lynch, 2010).
1.2. Disability acquisition and economic security Acquiring a disability in adulthood may lead to concerns about future earnings due to difficulty maintaining employment or the need to reduce hours or move into lower-skilled jobs. Previous Australian research has shown that working-age adults who acquire a disability are more likely to become unemployed or underemployed (i.e. employed in jobs for which they are over-educated and over-skilled) (M. Jones, Mavromaras, Sloane and Wei, 2014; M. K. Jones and Sloane, 2010). In this situation housing tenure and affordability may be particularly salient as housing represents the largest category of household expenditure and assets for Australians (ABS, 2011, 2015). 1.3. Beyond ‘average effects’ of disability acquisition and mental health As Bauer (2014) has argued, population health researchers tend to focus on single, or average effects (e.g. disability acquisition and mental health) rather than exploring how multiple positions, processes, and structural factors intersect to produce heterogeneous effects (Bauer, 2014). Given that people who acquire a disability may have reduced economic security it is conceivable that those with less financial and social resources may be particularly vulnerable to experiencing negative health consequences. As discussed earlier there is some evidence that social support, financial security and wealth may modify the association between disability acquisition and health (Honey et al., 2011; Smith et al., 2005). No previous studies have examined the whether housing circumstances prior to disability acquisition modify the association between acquisition and mental health. Housing tenure may provide a potential buffer against the mental health effects of acquiring a disability as housing assets may reduce concerns about threats to future income and tenure security. Similarly affordable housing may be important as even short-term reductions in income will make it difficult to meet housing costs with potential flow-through effects to mental health. In this paper we investigate whether housing tenure and affordability are effect modifiers of the relationship between disability acquisition and mental health. We use data from a sample of 1913 people participating in the Household, Income and Labour Dynamics in Australia survey (HILDA) e a national populationbased survey of Australian adults e who acquired a disability in adulthood. We conducted fixed-effects longitudinal regression analyses to account for time-invariant confounding and present estimates of effect measure modification (EMM) on the additive scale. 2. Methods 2.1. Conceptual model Fig. 1 represents the Directed Acyclic Graph informing this analysis. DAGs visually the postulated causal relationships that are believed a priori to exist between the variables of interest to the research question using unidirectional arrows (Greenland et al., 1999; Williamson et al., 2014). Variables that have arrows leading both to the exposure and the outcome of interest represent prior common causes, or confounders. DAGs are invaluable in making the assumptions underlying statistical analyses explicit, and guiding the selection of variables for inclusion as confounders (Greenland et al., 1999) and mediators in statistical models. In our modified DAG, time-varying confounders include age, employment and occupation, and income. We represent the effect modifiers (the housing variables) by a unidirectional arrow to the pathway
A.M. Kavanagh et al. / Social Science & Medicine 151 (2016) 225e232
227
Fig. 1. Directed Acyclic Graph showing the postulated causal relationships between disability and mental health.
between disability acquisition and mental health. 2.2. Data source HILDA is a longitudinal nationally representative study of Australian households and individuals which includes data on a range of life domains including social, demographic, health and economic characteristics (Summerfield et al., 2013). HILDA has been conducted in annual waves since 2001. The original panel included 13,969 individuals from 7682 households, sampled using a national probability sample of private dwellings. Data were collected on each household member, and face-to-face interviews were sought from all household members aged 15 years or above. In later waves, survey members included all original participants, household members attaining the age of 15 as well as new participants added as a result of changes in household composition if new households were formed by existing survey participants. Response rates for the survey have been above 90% for continuing participants and 70% for new participants. 2.3. Outcome variable The Mental Component Summary (MCS) score is derived from the Short Form 36 (SF-36) health survey. The SF-36 is a widely used self-completion measure of health status that has been validated for use in the Australian population, and to detect within-person changes in health over time (Butterworth and Crosier, 2004). The MCS is comprised of components of the eight subscales, most heavily weighted on the mental health, social functioning, vitality and role limitations due to emotional problems subscales. The MCS captures mental health and wellbeing; it is not designed to capture a clinical state such as anxiety and depression. The SF-36 was included in every wave of the HILDA survey, and the mean score on the MCS across all twelve waves in the total HILDA sample was 48.8 (standard deviation 10.3). 2.4. Disability measures Information on long-term health conditions and disabilities was collected in all waves using a definition derived from the International Classification of Functioning, Disability and Health (ICF) framework (World Health Organization, 2002). Participants were asked if they had an ‘impairment, long-term health condition or disability which restricts their everyday activities that had lasted, or was likely to last, for a period of six months or more’. Participants were defined as having acquired a disability if they
did not report a disability for two consecutive waves followed by two consecutive waves with a reported disability, a definition used in previous studies of disability acquisition including studies using HILDA data (Burchardt, 2003; Jenkins and Rigg, 2004; Polidano and Vu, 2015). We used two waves so as to exclude people with transient disability (such as an injury from which they recovery or a very time-limited condition) and to reduce the potential for measurement error because disability status was based on self-report. If participants reported more than one episode of disability acquisition (according to our definition), only the first episode was included. All consecutive waves in which individuals did not report a disability prior to their first reported disability acquisition and all consecutive waves reporting a disability subsequent to disability acquisition were included (minimum of four, maximum of 12 consecutive waves contributed for each person). Supplementary Table A shows potential individual disability trajectories and how they would be coded. 2.5. Housing variables Data on housing variables were collected at every wave of HILDA. To represent housing characteristics prior to disability acquisition, housing variables were recorded two waves prior to disability acquisition. Housing tenure was categorised as outright owners, mortgagors, private renters and public renters. A measure of unaffordable housing was constructed which identified households with a disposable household income in the lowest 40% of the national distribution as defined annually by the Australian Bureau of Statistics (ABS) (ABS, 2013), who had housing payments that exceeded 30% of their gross household income. This ‘30/40’ approach is commonly used to measure housing affordability in Australia (Yates et al., 2007). A final housing variable was constructed which combined housing tenure and housing affordability, categorizing housing affordability for each type of housing tenure for people who incur housing costs (i.e., mortgagors in affordable housing, mortgagors in unaffordable housing, private renters in affordable housing, and private renters in unaffordable housing). 2.6. Other variables Age was collapsed into four categories: under 30, 30 to 44, 45 to 59 and 60 years and above. Information on labour force status and occupational skill level was combined into a measure of employment that we have used previously (Milner et al., 2014) with five mutually-exclusive categories: unemployed (actively seeking employment or currently unable to find work), not in the labour
228
A.M. Kavanagh et al. / Social Science & Medicine 151 (2016) 225e232
force (not actively seeking employment, for various reasons including education, retirement, infirmity/disability, or household duties), in a low skill job (sales workers, machinery operators and drivers, labourers), medium skill job (technicians and trades workers; community and personal service workers; clerical and administrative workers) and high skill job (managers; professionals). Equivalised household disposable income was calculated by summing income components for all adults in the household and equivalised using the modified OECD scale (Haagenars et al., 1994). The equivalised variable was converted to national quintiles using statistics published annually by the ABS (ABS, 2013). Approximately 20% of the observations in the sample had missing values for household income. Data were imputed by the Reserve Bank of Australia using the Little and Su imputation method (Little and Su, 1989; Watson, 2004). 2.7. Statistical analysis Twelve waves of HILDA data were included in the analysis (2001e2012). We described the characteristics of the sample at baseline (first wave of participation in HILDA). We presented MCS scores by age, sex, employment status and occupation, equivalised household income and prior housing variables for people with and without disabilities. Comparisons of MCS scores were made within individuals; therefore we present the pooled mean (between-persons) of the within-person mean MCS scores for waves with and without disabilities. This was estimated as follows: 1. Calculating the mean MCS score for each participant i (ўi according to disability status xd where d ¼ 0 for no disability and d ¼ 1 for disability within category j of the covariate z). Given x ¼ d, z ¼ j, for k waves,
yi ¼
12 X
yi
k¼1
1 k
where yi is the MCS score for participant i, and k is the number of observations for individual i within strata of x ¼ d and z ¼ j (Equation 1). 2. Calculating the sample pooled mean of within-person means for x ¼ d, z ¼ j Given x ¼ d, z ¼ j,
y¼
X
yi
where yi is the mean MCS score for each participant i within strata of x ¼ d and z ¼ j (Equation 2). We used fixed-effects longitudinal linear regression to estimate the association between disability acquisition and MCS score. Fixed effects models, unlike conventional regression or random effects regression models, make comparisons within individuals rather than between people, therefore each individual acts as their own control thereby controlling for characteristics of the individual which do not vary with time (Allison, 2009). Coefficients from the models represent within-individual average differences in MCS scores between waves in which individuals reported no disability and waves in which they reported disability. Fixed-effects models remove bias from time-invariant confounding from both measured and unobserved variables (Gunasekara et al., 2013). We included age group, employment and occupation and income as time-varying confounders. In this analysis housing
variables were time-invariant as they were measured at a single time point prior to disability acquisition; therefore within-person changes in these variables cannot be estimated. To assess whether the association between disability acquisition and mental health varied by prior housing circumstances we included a crossproduct term between disability acquisition and each housing variable, and assessed whether there was statistical evidence of an interaction using the P values of the product terms. EMM was measured on the additive scale only as the MCS is a continuous variable (VanderWeele and Knol, 2014). All analyses were conducted in Stata/SE 12 (StataCorp, College Station) (StataCorp, 2011), using the xtreg command with fixed-effects estimators and robust standard errors to fit regression models and the lincom command to compute effect estimates and 95% confidence intervals for each category of the housing variables. The data used in this paper were extracted from HILDA using the Add-On package PanelWhiz for Stata (Hahn & Haisken-DeNew, 2013). 2.8. Sensitivity analyses We conducted the following sensitivity analyses to test the robustness of our findings. These included: 1. Exclusion of people with psychological impairments because their MCS scores are on average lower and the relationships between disability, housing and mental health may differ for this subgroup; 2. Exclusion of people with imputed income data from analyses of housing affordability. 3. Results There were 2112 persons (15,562 observations) who met our criteria for disability acquisition. There were complete data available for 91% of people, resulting in a final analytic sample of 1913 persons (13,037 observations). The mean number of observations (contributed annual waves of data) per person was 6.8 observations. Further details of sample selection and missing data are in Fig. 2. Supplementary Table B compares the characteristics of people who had missing data with those without missing data based on baseline characteristics. Missing data was more common in younger and older age groups and people with lower incomes. 3.1. Descriptive analyses At baseline entry into the analytic sample, nearly a third were aged 60 years or older; 58% were employed; 13% were in the highest income quintile while nearly a third were in the lowest quintile. In terms of housing tenure, over three quarters of the sample were living in their own home (49% outright owners, 28% servicing a mortgage), nearly one in five were in private rental and a small minority in public housing (4%). Seven percent of the sample lived in unaffordable housing (5% of mortgagors, 9% of private renters) (Table 1). Table 2 shows the mean of the pooled within-individual MCS scores by disability status for each covariate (see Methods). The MCS score was approximately two points lower in waves in which disability was reported compared to waves with no disability (47.1 versus 49.0). In both waves reporting disability and no disability, mental health was low in the unemployed and people living in unaffordable housing. Owner occupiers had the highest mental health of all tenure types. The MCS score was positively associated with age and men reported higher MCS than women.
A.M. Kavanagh et al. / Social Science & Medicine 151 (2016) 225e232
229
Fig. 2. Flow diagram of sample selection and missing data (MCS ¼ mental component summary; obs ¼ observations; ppl ¼ people).
3.2. Regression analyses
3.6. Sensitivity analyses
There was statistical evidence of EMM of the association between disability acquisition and mental health by all housing variables examined, which supported the inclusion of a cross-product term between disability acquisition and housing characteristics.
Results of our main analyses were robust to sensitivity analyses. Exclusion of people with psychological impairments attenuated the results however the general patterns were similar (Supplementary Table C). The results of the complete case analysis related to housing affordability (excluding the observations where income was imputed) were almost identical to the main analyses (Supplementary Table D).
3.3. Housing tenure While the mental health of people from all tenure types decreased after disability acquisition, the largest difference was for those in private rental housing whose mental health was 2.8 points lower (95% CI 3.9, 1.8) in waves reporting disability, compared to a difference of 1.6 points for those in outright ownership (95% CI 2.4, 0.1) (Table 3). 3.4. Housing affordability Living in unaffordable housing was associated with a 4.2 point lower mental health score in waves reporting disability compared to no disability (4.2, 95% CI 5.2, 1.4), compared to a 1.7 point decline for those in affordable housing (1.7, 95% CI 2.1, 1.3) (Table 3). 3.5. Housing affordability and tenure type Both mortgagors and private renters in unaffordable housing had the largest reductions in mental health following acquisition of a disability with reductions of 5.3 (95% CI 8.8, 1.9) and 4.0 (95% CI 6.3, 1.6) respectively (Table 4).
4. Discussion Our results suggest that housing characteristics prior to disability acquisition in adulthood modify the effect of disability acquisition on mental health. While we observed a small reduction in mental health for people who were outright owners or mortgagors, the effect was largest for those living in private rental. Living in unaffordable housing was associated with a four point reduction in the mental health score. Unlike previous analyses in the general population (Mason et al., 2013) where affordability was important for private renters only, we found evidence that both mortgagors and renters who were in unaffordable housing had a large decline in mental health associated with disability acquisition. We are unaware of any previous studies that have examined whether housing characteristics are effect modifiers of the relationship between disability acquisition and mental health. A previous study in the United States found a greater decline in subjective wellbeing following disability acquisition among those with below median wealth however this was a much smaller study
230
A.M. Kavanagh et al. / Social Science & Medicine 151 (2016) 225e232
Table 1 Characteristics of the analytic sample at baseline (individuals not observations, n ¼ 1913). N Age group (years) <30 30e44 45e60 60þ Sex Women Men Occupation High skilled Medium skilled Low skilled Not in the labour force Unemployed Income Q5 (highest) Q4 Q3 Q2 Q1 (lowest) Housing tenure Outright owner Servicing a mortgage Private renter Public renter Housing affordability Affordable Unaffordable Housing tenure# affordability Mortgage e affordable Mortgage e unaffordable Private renter e affordable Private renter e unaffordable
%
305 492 527 589
15.9 25.7 27.6 30.8
1075 838
56.2 43.8
411 392 296 754 60
21.5 20.5 15.5 39.4 3.1
248 312 373 390 590
13.0 16.3 19.5 20.4 30.8
943 531 365 74
49.3 27.8 19.1 3.9
1783 130
93.2 6.8
489 42 286 79
54.6 4.7 31.9 8.8
(478 people) with fewer waves of data and did not specifically assess housing assets or affordability (Smith et al., 2005). In this paper we used fixed-effects regression to examine within-person change in mental health e an approach we and others have used in a number of other studies of psychosocial working conditions (Butterworth and Crosier, 2004), employment arrangements (LaMontagne et al., 2014), housing affordability (R. Bentley et al., 2011), and job quality (R Bentley et al., 2015; Milner et al., 2015) in HILDA. The effect estimates for other exposures have been in the range of 1e2.5 points on the MCS scale. In this paper we find larger effect sizes particularly relation to housing affordability, highlighting the importance of these findings. A strength of this study is that it is based on a nationally representative study of nearly 2000 people who acquire a disability. In addition, fixed-effects longitudinal regression removes confounding due to time-invariant variables because comparisons are made within people (Gunasekara et al., 2013). Limitations include the potential for error due to missing data particularly on income (20% missing observations) however complete case analysis of the housing affordability analysis yielded similar results. Dependent misclassification is also a possibility because both disability and mental health are self-reported. In theory individual characteristics such as negative affectivity might result in dependent measurement error; however, to the extent that these characteristics are stable over time, fixed-effects regression will remove this bias. Selection bias due to attrition is a potential limitation however retention rates in HILDA are high (>90% at most waves). We acknowledge that the ratio based definition of housing unaffordability (30/40 rule) is somewhat blunt, however it is an accepted and widely-used cut-off, and has been tested and applied in Australia (E. Baker et al., 2013; Nepal et al., 2010). Because housing
Table 2 Mean within-person MCS score in waves reporting disability and no disability.
Whole sample Age group (years) <30 30e44 45e60 60þ Sex Women Men Occupation High skilled Medium skilled Low skilled Not in the labour force Unemployed Income Q5 (highest) Q4 Q3 Q2 Q1 (lowest) Housing tenure Outright owner Servicing a mortgage Private renter Public renter Housing affordability Affordable Unaffordable Housing tenure# affordability Mortgage - affordable Mortgage - unaffordable Private renter - affordable Private renter e unaffordable
No disability
Disability
49.0 (48.6, 49.4)
47.1 (46.6, 47.6)
44.8 46.5 48.7 52.8
41.1 43.1 46.6 50.6
(43.7, (45.7, (48.0, (52.2,
45.9) 47.3) 49.4) 53.3)
(39.4, (42.0, (45.8, (50.0,
42.7) 44.1) 47.4) 51.2)
48.3 (47.7, 48.8) 49.9 (49.3, 50.4)
46.3 (45.6, 46.9) 48.1 (47.5, 48.8)
48.7 48.0 48.2 50.1 43.0
(48.0, (47.3, (47.3, (49.5, (41.0,
49.5) 48.7) 49.0) 50.7) 45.0)
47.9 46.2 45.1 47.2 40.1
(47.0, (45.3, (43.9, (46.5, (37.7,
48.8) 47.1) 46.4) 47.9) 42.6)
49.4 48.4 49.1 48.4 49.0
(48.5, (47.6, (48.4, (47.7, (48.4,
50.3) 49.2) 49.7) 49.1) 49.7)
47.5 47.3 46.8 46.7 46.5
(46.5, (46.4, (45.9, (45.9, (45.7,
48.6) 48.2) 47.6) 47.5) 47.3)
50.9 47.6 46.6 46.3
(50.3, (46.8, (45.7, (44.0,
51.4) 48.3) 47.6) 48.6)
49.3 45.8 43.7 44.5
(48.7, (45.0, (42.5, (41.9,
49.9) 46.7) 44.9) 47.1)
49.1 (48.7, 49.5) 47.0 (45.3, 48.7)
47.4 (46.9, 47.8) 43.3 (41.2, 45.4)
47.5 48.2 46.4 47.4
46.1 42.8 43.7 43.9
(46.8, (45.1, (45.4, (45.2,
48.3) 51.2) 47.5) 49.6)
(45.2, (38.7, (42.3, (41.2,
47.0) 47.0) 45.0) 46.6)
Table 3 Linear fixed-effects regression coefficients for the within-person difference in MCS score between waves reporting disability and no disability (n ¼ 1913, observations ¼ 13 037) for categories of housing variables separately*
Housing tenure Outright owner Servicing a mortgagea Private renterb Public renterc Housing affordability Affordable Unaffordabled
Coeff.
95% CI
P Value
1.6 1.7 2.8 2.2
2.1, 2.5, 3.9, 3.9,
1.1 1.0 1.8 0.5
<0.001 <0.001 <0.001 0.012
1.7 4.2
2.1, 1.3 5.2, 1.4
<0.001 <0.001
*Adjusted for age, employment and equivalised household disposable income. a Interaction term: mortgage (0.2, 95% CI -1.1, 0.7, p ¼ 0.700). b Interaction term: private renter (1.3, 95% CI -2.4, -0.1, p ¼ 0.034). c Interaction term: public renter (0.6, 95% CI -2.4, 1.2, p ¼ 0.492). d Interaction term: unaffordable housing (2.5, 95% CI -4.4, -0.5, p ¼ 0.015).
characteristics are modelled as time-invariant variables classified prior to disability acquisition, we cannot estimate the main effects of housing variables although we are able to estimate the values of their interaction terms with disability acquisition. People with severe disabilities, particularly intellectual and psychological disabilities, may be less likely to participate in HILDA and it is possible that the relationships between disability, housing and mental health are different in this group. While most Australians housing is physically sound, recent work (Baker et al., 2016) has highlighted that subgroups of the Australian population (e.g. indigenous Australians) live in housing of poor physical quality are vulnerable to the same health problems documented in other countries (Gibson et al., 2011; Krieger and Higgins, 2002); we have not investigated
A.M. Kavanagh et al. / Social Science & Medicine 151 (2016) 225e232 Table 4 Linear fixed-effects regression coefficients for the difference in MCS score between waves reporting disability and no disability according to tenure/affordability (n ¼ 896, observations ¼ 5913)*
Housing tenure# affordability Mortgage - affordable Mortgage - unaffordablea Private renter - affordableb Private renter - unaffordablec
Coeff.
95% CI
P Value
1.4 5.3 2.4 4.0
2.1, 8.8, 3.6, 6.3,
<0.001 0.003 <0.001 0.001
0.7 1.9 1.3 1.6
*Adjusted for age, employment and equivalised household disposable income. a Interaction term: mortgage - unaffordable (4.0, 95% CI -7.5, -0.4, p ¼ 0.030). b Interaction term: private renter - affordable (1.0, 95% CI -2.4, 0.3, p ¼ 0.144). c Interaction term: private renter e unaffordable (2.6, 95% CI -5.1, -0.1, p ¼ 0.042).
these issues in this study. Although we did include the variables we postulated to be the most important time-varying confounders (age, employment and occupation, and income) it is possible that confounding due to time-varying confounders not included in the model remains. Models that additionally tested for household structure and education had almost identical results. Ten percent of the eligible sample had missing data on one or more of the variables. Young and older age groups and people from the lowest two quintiles of income were more likely to have missing data. Selection bias is possible if the relationships between housing characteristics, disability and mental health varied for those with missing data. Finally, the findings of this study may not be generalizable to other settings as there are considerable variations between countries with respect to home ownership, affordability, tenure and housing policies. In sum, we find evidence to support EMM by housing affordability and tenure of the association between disability acquisition and mental health. Importantly, the small proportion of people living in unaffordable housing at the time of disability acquisition (approximately seven percent of our sample), are likely to experience more severe mental health effects with potential consequences for ongoing workforce participation and health costs. As this study models housing conditions prior to disability acquisition the findings suggest that the ideal time to intervene (e.g. with financial support) is prior to acquisition which cannot be known in advance. It is possible that financial support to meet housing costs for those living in unaffordable housing at the time of acquisition would also protect against deterioration in mental health however that would best be tested in an experimental or quasi-experimental study. Appendix A. Supplementary data Supplementary data related to this article can be found at http:// dx.doi.org/10.1016/j.socscimed.2016.01.010. References ABS, 2011. Household Expenditure Survey, Australia: Summary of Results, 20092010 (6530.0): ABS. ABS, 2013. Household Income and Income Distribution, Australia, 2011-12 (6523.0): ABS. ABS, 2014. Australian Social Trends (4102.0). March 2014: ABS. ABS, 2015. Australian National Accounts: Finance and Wealth, Dec 2014 (5232.0): ABS. Allison, P.D., 2009. Fixed Effects Regression Models (Quantitative Applications in the Social Sciences). Sage Publications, London. Baker, E., Bentley, R., Mason, K., 2013. The mental health effects of housing tenure: causal or compositional? Urban Stud. 50 (2), 426e442. http://dx.doi.org/ 10.1177/0042098012446992. Baker, E., Lester, L., Bentley, R., Beer, A., 2016. The health effects of poor quality housing: Australia's hidden fraction. J. Prev. Intervention Community (in press). Bauer, G.R., 2014. Incorporating intersectionality theory into population health
231
research methodology: challenges and the potential to advance health equity. Soc. Sci. Med. 110, 10e17. http://dx.doi.org/10.1016/j.socscimed.2014.03.022. Beer, A., Baker, E., Mallett, S., Batterham, D., Pate, A., Lester, L., 2012. Addressing Homelessness Amongst Persons with a Disability: Identifying and Enacting Best Practice. Department of Families, Housing, COmmunity Services and Indigenous Affairs, Adelaide. Beer, A., Faulkner, D., 2008. The Housing Careers of People with a Disability and Carers of People with a Disability. Beer, A., Faulkner, D., Paris, C., Clower, T., 2011. Housing Transitions Through the Life Course: Aspirations, Needs and Policy. The Policy Press, Bristol. Bentley, R., Baker, E., Mason, K., Subramanian, S.V., Kavanagh, A.M., 2011. Association between housing affordability and mental health: a longitudinal analysis of a nationally representative household survey in Australia reply. Am. J. Epidemiol. 174 (7), 753e760. http://dx.doi.org/10.1093/Aje/Kwr161. Bentley, R., Krnjacki, L., Kavanagh, A., Lamontagne, A.D., 2015. A longitudinal analysis of changes in job control and mental health. Am. J. Epidemiol. 182 (4), 328e334. Braubach, M., 2011. Key challenges of housing and health from WHO perspective. Int. J. Public Health 56, 579e580. Burchardt, T., 2003. Being and becoming: social exclusion and the onset of disability. In: Centre for Analysis of Social Exclusion. The London School of Economics.,, London. Butterworth, P., Crosier, T., 2004. 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 4. http://dx.doi.org/10.1186/1471-2458-4-44. Dalton, T., Ong, R., 2007. Welfare to work in Australia: disability income support, housing affordability and employment incentives. Int. J. Hous. Policy 7, 275e297. Demographia, 2014. 10th Annual Demographia International Housing Affordability Survey (2013: 3rd Quarter). Demographia, New York. Ellaway, A., Macintyre, S., 1998. Does housing tenure predict health in the UK because it exposes people to different levels of housing related hazards in the home or its surroundings? Health & Place 4 (2), 141e150. http://dx.doi.org/ 10.1016/S1353-8292(98) 00006e9. Pii S1353-8292(98)00006-9. Emerson, E., Hatton, C., 2007. Poverty, socio-economic position, social capital and the health of children and adolescents with intellectual disabilities in Britain: a replication. J. Intellect. Disabil. Res. 51 (11), 866e874. JIR951 [pii] 10. 1111/ j.1365e2788.2007.00951.x [doi]. Emerson, E., Kariuki, M., Honey, A., Llewellyn, G., 2014. 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. 7 (4), 448e456. http://dx.doi.org/10.1016/ j.dhjo.2014.03.004. Emerson, E., Llewellyn, G., Honey, A., Kariuki, M., 2012. Lower well-being of young Australian adults with self-reported disability reflects their poorer living conditions rather than health issues. Aust. N. Z. J. Public Health, 36 (2), 176e182. http://dx.doi.org/10.1111/j.1753-6405.2011.00810.x. Evans, J., Hyndman, S., Stewart-Brown, S., Smith, D., Petersen, S., 2000. An epidemiological study of the relative importance of damp housing in relation to adult health. J. Epidemiol. Community Health 54, 677e686. Free, S., Howden-Chapman, P., Pierse, N., Viggers, H., the Housing, H. a. H. S. R. T, 2010. More effective home heating reduces school absences for children with asthma. J. Epidemiol. Community Health 64, 379e386. Gibson, M., Petticrew, M., Bambra, C., AJ, S., Wright, K., Whitehead, M., 2011. Housing and health inequalities:A synthesis of systematic reviews of interventions aimed at different pathways linking housing and health. Health Place 17, 174e185. Greenland, S., Pearl, J., Robins, J.M., 1999. Causal diagrams for epidemiologic research. Epidemiology 10 (1), 37e48. Gunasekara, F.I., Richardson, K., Carter, K., Blakely, T., 2013. Fixed effects analysis of repeated measures data. Int. J. Epidemiol. 43 (1), 264e269. http://dx.doi.org/ 10.1093/ije/dyt221. Haagenars, A., de Vos, K., Zaidi, A., 1994. Poverty Statistics in the Late 1980s: Research Based on Micro-data. Office for Official Publications of the European Communities, Luxemburg. Hahn, M.H., Haisken-DeNew, J.P., 2013. PanelWhiz and the australian longitudinal data infrastructure in economics. Aust. Econ. Rev. 46 (3), 379e386. http:// dx.doi.org/10.1111/j.1467-8462.2013.12010.x. Honey, A., Emerson, E., Llewellyn, G., 2011. The mental health of young people with disabilities: impact of social conditions. Soc. Psychiatry Psychiatr. Epidemiol. 46 (1), 1e10. http://dx.doi.org/10.1007/s00127-009-0161-y. Imrie, R., 2006. Accessible Housing: Quality, Disability and Design. Taylor & Francis. Jenkins, S.P., Rigg, J.A., 2004. Disability and disadvantage: selection, onset, and duration effects. J. Soc. Policy 33, 479e501. http://dx.doi.org/10.1017/ S0047279404007780. Jones, M., Mavromaras, K., Sloane, P., Wei, Z., 2014. Disability, job mismatch, earnings and job satisfaction in Australia. Camb. J. Econ. 38 (5), 1221e1246. http:// dx.doi.org/10.1093/Cje/Beu014. Jones, M.K., Sloane, P.J., 2010. Disability and skill mismatch. Econ. Rec. 86, 101e114. http://dx.doi.org/10.1111/j.1475-4932.2010.00659.x. Kavanagh, A.M., Aitken, Z., Krnjacki, L., LaMontagne, A.D., Bentley, R., Milner, A., 2015. Mental health following acquisition of disability in adulthood-the impact of wealth. PLoS One 10 (10), e0139708. http://dx.doi.org/10.1371/ journal.pone.0139708.
232
A.M. Kavanagh et al. / Social Science & Medicine 151 (2016) 225e232
Kavanagh, A.M., Krnjacki, L., Aitken, Z., LaMontagne, A.D., Beer, A., Baker, E., Bentley, R., 2014. Intersections between disability, type of impairment, gender and socio-economic disadvantage in a nationally representative sample of 33,101 working-aged Australians. Disabil. Health J. 8 (2), 191e199. http:// dx.doi.org/10.1016/j.dhjo.2014.08.008. Kavanagh, A.M., Krnjacki, L., Beer, A., Lamontagne, A.D., Bentley, R., 2013. Time trends in socio-economic inequalities for women and men with disabilities in Australia: evidence of persisting inequalities. Int. J. Equity Health 12 (1), 73. http://dx.doi.org/10.1186/1475-9276-12-73. Krieger, J., Higgins, D.L., 2002. Housing and health: time again for public health action. Am. J. Public Health 92 (5), 758e768. http://dx.doi.org/10.2105/ Ajph.92.5.758. Kyle, T., Dunn, J., 2008. Effects of housing circumstances on health, quality of life and healthcare use for people with severe mental illness: a review. Health Soc. Care Community 16 (1), 1e15. LaMontagne, A.D., Milner, A., Krnjacki, L., Kavanagh, A.M., Blakely, T.A., Bentley, R., 2014. Employment arrangements and mental health in a cohort of working australians: are transitions from permanent to temporary employment associated with changes in mental health? Am. J. Epidemiol. 179 (12), 1467e1476. http://dx.doi.org/10.1093/Aje/Kwu093. Little, R.J.A., Su, H.-L., 1989. Item Nonresponse in Panel Surveys. John Wiley & Sons, New York. Macintyre, S., Ellaway, A., Hiscock, R., Kearns, A., Der, G., McKay, L., 2003. What features of the home and the area might help to explain observed relationships between housing tenure and health? evidence from the west of Scotland. Health & Place 9 (3), 207e218. http://dx.doi.org/10.1016/S1353-8292(02) 00040e0. Pii S1353-8292(02)00040-0. Mandemakers, J.J., Monden, C.W.S., 2010. Does education buffer the impact of disability on psychological distress? Soc. Sci. Med. 71 (2), 288e297. http:// dx.doi.org/10.1016/j.socscimed.2010.04.004. Mason, K., Baker, E., Blakely, T., Bentley, R., 2013. Housing affordability and mental health: does the relationship differ for renters and home purchasers? Soc. Sci. & Med. 94, 91e97. Milner, A., Butterworth, P., Bentley, R., Kavanagh, A.M., LaMontagne, A.D., 2015. Sickness absence and psychosocial job quality: an analysis from a longitudinal survey of working australians, 2005-2012. Am. J. Epidemiol. 181 (10), 781e788. http://dx.doi.org/10.1093/aje/kwu355. Milner, A., LaMontagne, A.D., Aitken, Z., Bentley, R., Kavanagh, A.M., 2014. Employment status and mental health among persons with and without a disability: evidence from an Australian cohort study. J. Epidemiol. Community Health 68 (11), 1064e1071. http://dx.doi.org/10.1136/jech-2014-204147. Nepal, B., Tanton, R., Harding, A., 2010. Measuring housing stress: how much do definitions matter? Urban Policy Res. 28 (2), 211e224. http://dx.doi.org/ 10.1080/08111141003797454. Papworth Trust, 2011. Disability in the United Kingdom 2011: Facts and Figures. Paworth Trust, Cambridge. Parker, S., Fisher, K., 2010. Facilitators and barriers in Australian disability housing support policies: using a human rights framework 30.3/4(2010). Disabil. Stud.
Q. 30. Pevalin, D.J., Taylor, M.P., Todd, J., 2008. The dynamics of unhealthy housing in the UK: a panel data analysis. Hous. Stud. 23 (5), 679e695. http://dx.doi.org/ 10.1080/02673030802253848. Polidano, C., Vu, H., 2015. Differential labour market impacts from disability onset. Health Econ. 24 (3), 302e317. http://dx.doi.org/10.1002/hec.3017. Pollack, C., Griffin, B., Lynch, J., 2010. Housing affordability and health among homeowners and renters. Am. J. Prev. Med. 39 (6), 515e521. Pollack, C.E., von dem Knesebeck, O., Siegrist, J., 2004. Housing and health in Germany. J. Epidemiol. Community Health 58 (3), 216e222. http://dx.doi.org/ 10.1136/jech.2003.012781. Shaw, M., 2004. Housing and public health. Annu. Rev. Public Health 25, 397e418. Smith, D.M., Langa, K.M., Kabeto, M.U., Ubel, P.A., 2005. Health, wealth, and happiness: financial resources buffer subjective well-being after the onset of a disability. Research Support, N.I.H., Extramural Research Support, Non-U.S. Gov't Research Support, U.S. Gov't, Non-P.H.S. Psychol. Sci. 16 (9), 663e666. http://dx.doi.org/10.1111/j.1467-9280.2005.01592.x. StataCorp, 2011. Stata Statistical Software: Release 12. StataCorp LP, College Station, TX. Summerfield, M., Freidin, S., Hahn, M., Ittak, P., Li, N., Macalalad, N., Wooden, M., 2013. HILDA user manual - release 12. In: Melbourne Institute of Applied Economic and Social Research. Turner, R.J., Noh, S., 1988. Physical disability and depression - a longitudinal analysis. J. Health Soc. Behav. 29 (1), 23e37. http://dx.doi.org/10.2307/2137178. VanderWeele, T.J., Knol, M.J., 2014. A Tutorial on Interaction Epidemiologic Methods, vol. 3, p. 33. Wang, 2005. Disability and American Families: 2000 Census Special Reports: US Census Bureau. Watson, N., 2004. Income and Wealth Imputation for Waves 1 and 2 HILDA Project Technical Paper Series No. 3/04, July 2004. Melbourne Institute of Applied Economic and Social Research, University of Melbourne, Melbourne. White, B.J., Peaslee, J., LaQuatra, J., 1994. Comparing housing affordability and quality among disability households: the United States and its regions. J. Fam. Econ. Issues 15 (4), 367e380. Williamson, E.J., Aitken, Z., Lawrie, J., Dharmage, S.C., Burgess, J.A., Forbes, A.B., 2014. Introduction to causal diagrams for confounder selection. Respirology 19 (3), 303e311. http://dx.doi.org/10.1111/Resp.12238. World Health Organization, 2002. Towards a Common Language for Functioning, Disability and Health: ICF the International Classification of Functioning, Disability and Health. WHO, Geneva. World Health Organization and World Bank Group, 2011. World Report on Disability. WHO, Geneva. Yang, Y., 2006. How does functional disability affect depressive symptoms in late life? the role of perceived social support and psychological resources. J. Health Soc. Behav. 47 (4), 355e372. Yates, J., Milligan, V., Berry, M., Burke, T., Gabriel, M., Pinnegar, S., 2007. Housing Affordability: A 21st Century Problem AHURI Final Report No. 105. Australian Housing and Urban Research Institute, Sydney Research Centre, Melbourne.
Contents lists available at ScienceDirect
Social Science & Medicine journal homepage: www.elsevier.com/locate/socscimed
Housing tenure and affordability and mental health following disability acquisition in adulthood Anne M. Kavanagh a, *, Zoe Aitken a, Emma Baker b, Anthony D. LaMontagne c, d, Allison Milner c, d, Rebecca Bentley a a
Gender and Women's Health, Centre for Health Equity, Melbourne School of Population and Global Health, The University of Melbourne, Australia School of Architecture & Built Environment, The University of Adelaide, Australia Population Health Strategic Research Centre, School of Health & Social Development, Deakin University, Australia d McCaughey VicHealth Centre for Community Wellbeing, Centre for Health Equity, Melbourne School of Population and Global Health, The University of Melbourne, Australia b c
a r t i c l e i n f o
a b s t r a c t
Article history: Received 22 July 2015 Received in revised form 19 October 2015 Accepted 7 January 2016 Available online 9 January 2016
Acquiring a disability in adulthood is associated with a reduction in mental health and access to secure and affordable housing is associated with better mental health. We hypothesised that the association between acquisition of disability and mental health is modified by housing tenure and affordability. We used twelve annual waves of data (2001e2012) (1913 participants, 13,037 observations) from the Household, Income and Labour Dynamics in Australia survey. Eligible participants reported at least two consecutive waves of disability preceded by two consecutive waves without disability. Effect measure modification, on the additive scale, was tested in three fixed-effects linear regression models (which remove time-invariant confounding) which included a cross-product term between disability and prior housing circumstances: housing tenure by disability; housing affordability by disability and, in a subsample (896 participants 5913 observations) with housing costs, tenure/affordability by disability. The outcome was the continuous mental component summary (MCS) of SF-36. Models adjusted for timevarying confounders. There was statistical evidence that prior housing modified the effect of disability acquisition on mental health. Our findings suggested that those in affordable housing had a 1.7 point deterioration in MCS (95% CI -2.1, -1.3) following disability acquisition and those in unaffordable housing had a 4.2 point reduction (95% CI -5.2, -1.4). Among people with housing costs, the largest declines in MCS were for people with unaffordable mortgages (5.3, 95% CI 8.8, 1.9) and private renters in unaffordable housing (4.0, 95% CI 6.3, 1.6), compared to a 1.4 reduction (95% CI 2.1, 0.7) for mortgagors in affordable housing. In sum, we used causally-robust fixed-effects regression and showed that deterioration in mental health following disability acquisition is modified by prior housing circumstance with the largest negative associations found for those in unaffordable housing. Future research should test whether providing secure, affordable housing when people acquire a disability prevents deterioration in mental health. © 2016 Published by Elsevier Ltd.
Keywords: Disability Mental health Housing tenure Housing affordability Longitudinal study Fixed-effects regression Effect measure modification
1. Introduction Internationally, people with disabilities e nearly 20% of the population e experience significant socio-economic disadvantage (Kavanagh et al., 2014; Kavanagh et al., 2013; World Health
* Corresponding author. Gender and Women's Health, Centre for Health Equity, Melbourne School of Population and Global Health, The University of Melbourne, Level 3, 207 Bouverie St, Carlton, 3010 Victoria, Australia. E-mail address: [email protected] (A.M. Kavanagh). http://dx.doi.org/10.1016/j.socscimed.2016.01.010 0277-9536/© 2016 Published by Elsevier Ltd.
Organization and World Bank Group, 2011). Disabled Australians have lower rates of employment, post-secondary education, and incomes and are more likely to experience housing-related disadvantage compared to those without a disability (Beer et al., 2012; Beer and Faulkner, 2008; Beer et al., 2011; Kavanagh et al., 2014, 2013; Parker and Fisher, 2010). They are over-represented in Australia's housing welfare sector (Dalton and Ong, 2007), are more likely to experience homelessness (Beer et al., 2012; Beer and Faulkner, 2008) and unaffordable housing (Kavanagh et al., 2014, 2013) e a situation found in other high-income countries (Kyle
226
A.M. Kavanagh et al. / Social Science & Medicine 151 (2016) 225e232
and Dunn, 2008; Papworth Trust, 2011; Wang, 2005; White et al., 1994). People with disabilities have poorer physical and mental health and some authors have argued that their disadvantaged circumstances (including inadequate housing) may play a significant role in explaining these health inequalities although the evidence is primarily limited to adolescents and young adults (E. Emerson and Hatton, 2007; E. Emerson, Llewellyn, Honey and Kariuki, 2012; Honey et al., 2011). A study of older adults (65 years and older) found that for those who reported a disability at baseline, social support modified the relationship between disability and depressive symptoms (Yang, 2006). On average, mental health tends to decline after acquisition of disability (Mandemakers and Monden, 2010; Turner and Noh, 1988), however this is not universal (Eric Emerson, Kariuki, Honey and Llewellyn, 2014; Honey et al., 2011); one small Australian study found that social support and financial stability prior to onset of disability protected against deterioration in mental health in young adults (Honey et al., 2011). A study of over 2000 Australian adults found that although being in the highest tertile of wealth before disability acquisition was associated with a small deterioration in mental health the deterioration was larger for those in the lowest and mid tertile of wealth (Kavanagh et al., 2015). Similar findings were found in a a UK study of adultswhere a greater reduction in subjective wellbeing was observed for people with below median wealth at time of disability acquisition (Smith et al., 2005). Elucidation of factors that predict deterioration in mental health with disability acquisition is of policy significance because it will enable better targeting of health and non-health sector interventions. Given housing's known relationship to health and its central role as a primary and regular household expenditure, it is an important potential modifier of the effect of disability acquisition on mental health e the focus of this paper. 1.1. Housing and health Housing is an important social determinant of health (Braubach, 2011; Gibson et al., 2011; Shaw, 2004), characteristics which have been linked to health include dwelling quality, location, tenure and affordability. There is a strong evidence-base linking structural characteristics of housing (e.g. toxins, damp, accessibility) (Evans et al., 2000; Free et al., 2010) to health including for people with disabilities (Imrie, 2006). Australian housing stock is of relatively good quality and because the majority of Australians live around the climatically mild coast, the structural aspects of housing are less important. Tenure type is strongly related to health, with outright owners and mortgagors having better health than people living in private rental or public housing (Ellaway and Macintyre, 1998; Macintyre et al., 2003; Pollack et al., 2004); however, whether this relationship is causal has been debated (E. Baker, Bentley and Mason, 2013). In Australia, an association between housing tenure and mental health was not found in causally-robust fixedeffects longitudinal regression analyses (E. Baker et al., 2013). In the context of rapidly increasing house prices and increasing household debt relative to income (ABS, 2014), Australia currently has one of the most unaffordable housing markets in the world (Demographia, 2014), making housing affordability an increasingly important determinant of health. Two longitudinal studies in the United Kingdom and Australia have demonstrated deleterious associations between living in unaffordable housing and mental health independent of general financial hardship (R. Bentley, Baker, Mason, Subramanian and Kavanagh, 2011; Pevalin et al., 2008). However this relationship appears to be modified by tenure type with recent research suggesting that this association is stronger for private renters than home-owners in Australia (Mason et al., 2013; C. Pollack, Griffin and Lynch, 2010).
1.2. Disability acquisition and economic security Acquiring a disability in adulthood may lead to concerns about future earnings due to difficulty maintaining employment or the need to reduce hours or move into lower-skilled jobs. Previous Australian research has shown that working-age adults who acquire a disability are more likely to become unemployed or underemployed (i.e. employed in jobs for which they are over-educated and over-skilled) (M. Jones, Mavromaras, Sloane and Wei, 2014; M. K. Jones and Sloane, 2010). In this situation housing tenure and affordability may be particularly salient as housing represents the largest category of household expenditure and assets for Australians (ABS, 2011, 2015). 1.3. Beyond ‘average effects’ of disability acquisition and mental health As Bauer (2014) has argued, population health researchers tend to focus on single, or average effects (e.g. disability acquisition and mental health) rather than exploring how multiple positions, processes, and structural factors intersect to produce heterogeneous effects (Bauer, 2014). Given that people who acquire a disability may have reduced economic security it is conceivable that those with less financial and social resources may be particularly vulnerable to experiencing negative health consequences. As discussed earlier there is some evidence that social support, financial security and wealth may modify the association between disability acquisition and health (Honey et al., 2011; Smith et al., 2005). No previous studies have examined the whether housing circumstances prior to disability acquisition modify the association between acquisition and mental health. Housing tenure may provide a potential buffer against the mental health effects of acquiring a disability as housing assets may reduce concerns about threats to future income and tenure security. Similarly affordable housing may be important as even short-term reductions in income will make it difficult to meet housing costs with potential flow-through effects to mental health. In this paper we investigate whether housing tenure and affordability are effect modifiers of the relationship between disability acquisition and mental health. We use data from a sample of 1913 people participating in the Household, Income and Labour Dynamics in Australia survey (HILDA) e a national populationbased survey of Australian adults e who acquired a disability in adulthood. We conducted fixed-effects longitudinal regression analyses to account for time-invariant confounding and present estimates of effect measure modification (EMM) on the additive scale. 2. Methods 2.1. Conceptual model Fig. 1 represents the Directed Acyclic Graph informing this analysis. DAGs visually the postulated causal relationships that are believed a priori to exist between the variables of interest to the research question using unidirectional arrows (Greenland et al., 1999; Williamson et al., 2014). Variables that have arrows leading both to the exposure and the outcome of interest represent prior common causes, or confounders. DAGs are invaluable in making the assumptions underlying statistical analyses explicit, and guiding the selection of variables for inclusion as confounders (Greenland et al., 1999) and mediators in statistical models. In our modified DAG, time-varying confounders include age, employment and occupation, and income. We represent the effect modifiers (the housing variables) by a unidirectional arrow to the pathway
A.M. Kavanagh et al. / Social Science & Medicine 151 (2016) 225e232
227
Fig. 1. Directed Acyclic Graph showing the postulated causal relationships between disability and mental health.
between disability acquisition and mental health. 2.2. Data source HILDA is a longitudinal nationally representative study of Australian households and individuals which includes data on a range of life domains including social, demographic, health and economic characteristics (Summerfield et al., 2013). HILDA has been conducted in annual waves since 2001. The original panel included 13,969 individuals from 7682 households, sampled using a national probability sample of private dwellings. Data were collected on each household member, and face-to-face interviews were sought from all household members aged 15 years or above. In later waves, survey members included all original participants, household members attaining the age of 15 as well as new participants added as a result of changes in household composition if new households were formed by existing survey participants. Response rates for the survey have been above 90% for continuing participants and 70% for new participants. 2.3. Outcome variable The Mental Component Summary (MCS) score is derived from the Short Form 36 (SF-36) health survey. The SF-36 is a widely used self-completion measure of health status that has been validated for use in the Australian population, and to detect within-person changes in health over time (Butterworth and Crosier, 2004). The MCS is comprised of components of the eight subscales, most heavily weighted on the mental health, social functioning, vitality and role limitations due to emotional problems subscales. The MCS captures mental health and wellbeing; it is not designed to capture a clinical state such as anxiety and depression. The SF-36 was included in every wave of the HILDA survey, and the mean score on the MCS across all twelve waves in the total HILDA sample was 48.8 (standard deviation 10.3). 2.4. Disability measures Information on long-term health conditions and disabilities was collected in all waves using a definition derived from the International Classification of Functioning, Disability and Health (ICF) framework (World Health Organization, 2002). Participants were asked if they had an ‘impairment, long-term health condition or disability which restricts their everyday activities that had lasted, or was likely to last, for a period of six months or more’. Participants were defined as having acquired a disability if they
did not report a disability for two consecutive waves followed by two consecutive waves with a reported disability, a definition used in previous studies of disability acquisition including studies using HILDA data (Burchardt, 2003; Jenkins and Rigg, 2004; Polidano and Vu, 2015). We used two waves so as to exclude people with transient disability (such as an injury from which they recovery or a very time-limited condition) and to reduce the potential for measurement error because disability status was based on self-report. If participants reported more than one episode of disability acquisition (according to our definition), only the first episode was included. All consecutive waves in which individuals did not report a disability prior to their first reported disability acquisition and all consecutive waves reporting a disability subsequent to disability acquisition were included (minimum of four, maximum of 12 consecutive waves contributed for each person). Supplementary Table A shows potential individual disability trajectories and how they would be coded. 2.5. Housing variables Data on housing variables were collected at every wave of HILDA. To represent housing characteristics prior to disability acquisition, housing variables were recorded two waves prior to disability acquisition. Housing tenure was categorised as outright owners, mortgagors, private renters and public renters. A measure of unaffordable housing was constructed which identified households with a disposable household income in the lowest 40% of the national distribution as defined annually by the Australian Bureau of Statistics (ABS) (ABS, 2013), who had housing payments that exceeded 30% of their gross household income. This ‘30/40’ approach is commonly used to measure housing affordability in Australia (Yates et al., 2007). A final housing variable was constructed which combined housing tenure and housing affordability, categorizing housing affordability for each type of housing tenure for people who incur housing costs (i.e., mortgagors in affordable housing, mortgagors in unaffordable housing, private renters in affordable housing, and private renters in unaffordable housing). 2.6. Other variables Age was collapsed into four categories: under 30, 30 to 44, 45 to 59 and 60 years and above. Information on labour force status and occupational skill level was combined into a measure of employment that we have used previously (Milner et al., 2014) with five mutually-exclusive categories: unemployed (actively seeking employment or currently unable to find work), not in the labour
228
A.M. Kavanagh et al. / Social Science & Medicine 151 (2016) 225e232
force (not actively seeking employment, for various reasons including education, retirement, infirmity/disability, or household duties), in a low skill job (sales workers, machinery operators and drivers, labourers), medium skill job (technicians and trades workers; community and personal service workers; clerical and administrative workers) and high skill job (managers; professionals). Equivalised household disposable income was calculated by summing income components for all adults in the household and equivalised using the modified OECD scale (Haagenars et al., 1994). The equivalised variable was converted to national quintiles using statistics published annually by the ABS (ABS, 2013). Approximately 20% of the observations in the sample had missing values for household income. Data were imputed by the Reserve Bank of Australia using the Little and Su imputation method (Little and Su, 1989; Watson, 2004). 2.7. Statistical analysis Twelve waves of HILDA data were included in the analysis (2001e2012). We described the characteristics of the sample at baseline (first wave of participation in HILDA). We presented MCS scores by age, sex, employment status and occupation, equivalised household income and prior housing variables for people with and without disabilities. Comparisons of MCS scores were made within individuals; therefore we present the pooled mean (between-persons) of the within-person mean MCS scores for waves with and without disabilities. This was estimated as follows: 1. Calculating the mean MCS score for each participant i (ўi according to disability status xd where d ¼ 0 for no disability and d ¼ 1 for disability within category j of the covariate z). Given x ¼ d, z ¼ j, for k waves,
yi ¼
12 X
yi
k¼1
1 k
where yi is the MCS score for participant i, and k is the number of observations for individual i within strata of x ¼ d and z ¼ j (Equation 1). 2. Calculating the sample pooled mean of within-person means for x ¼ d, z ¼ j Given x ¼ d, z ¼ j,
y¼
X
yi
where yi is the mean MCS score for each participant i within strata of x ¼ d and z ¼ j (Equation 2). We used fixed-effects longitudinal linear regression to estimate the association between disability acquisition and MCS score. Fixed effects models, unlike conventional regression or random effects regression models, make comparisons within individuals rather than between people, therefore each individual acts as their own control thereby controlling for characteristics of the individual which do not vary with time (Allison, 2009). Coefficients from the models represent within-individual average differences in MCS scores between waves in which individuals reported no disability and waves in which they reported disability. Fixed-effects models remove bias from time-invariant confounding from both measured and unobserved variables (Gunasekara et al., 2013). We included age group, employment and occupation and income as time-varying confounders. In this analysis housing
variables were time-invariant as they were measured at a single time point prior to disability acquisition; therefore within-person changes in these variables cannot be estimated. To assess whether the association between disability acquisition and mental health varied by prior housing circumstances we included a crossproduct term between disability acquisition and each housing variable, and assessed whether there was statistical evidence of an interaction using the P values of the product terms. EMM was measured on the additive scale only as the MCS is a continuous variable (VanderWeele and Knol, 2014). All analyses were conducted in Stata/SE 12 (StataCorp, College Station) (StataCorp, 2011), using the xtreg command with fixed-effects estimators and robust standard errors to fit regression models and the lincom command to compute effect estimates and 95% confidence intervals for each category of the housing variables. The data used in this paper were extracted from HILDA using the Add-On package PanelWhiz for Stata (Hahn & Haisken-DeNew, 2013). 2.8. Sensitivity analyses We conducted the following sensitivity analyses to test the robustness of our findings. These included: 1. Exclusion of people with psychological impairments because their MCS scores are on average lower and the relationships between disability, housing and mental health may differ for this subgroup; 2. Exclusion of people with imputed income data from analyses of housing affordability. 3. Results There were 2112 persons (15,562 observations) who met our criteria for disability acquisition. There were complete data available for 91% of people, resulting in a final analytic sample of 1913 persons (13,037 observations). The mean number of observations (contributed annual waves of data) per person was 6.8 observations. Further details of sample selection and missing data are in Fig. 2. Supplementary Table B compares the characteristics of people who had missing data with those without missing data based on baseline characteristics. Missing data was more common in younger and older age groups and people with lower incomes. 3.1. Descriptive analyses At baseline entry into the analytic sample, nearly a third were aged 60 years or older; 58% were employed; 13% were in the highest income quintile while nearly a third were in the lowest quintile. In terms of housing tenure, over three quarters of the sample were living in their own home (49% outright owners, 28% servicing a mortgage), nearly one in five were in private rental and a small minority in public housing (4%). Seven percent of the sample lived in unaffordable housing (5% of mortgagors, 9% of private renters) (Table 1). Table 2 shows the mean of the pooled within-individual MCS scores by disability status for each covariate (see Methods). The MCS score was approximately two points lower in waves in which disability was reported compared to waves with no disability (47.1 versus 49.0). In both waves reporting disability and no disability, mental health was low in the unemployed and people living in unaffordable housing. Owner occupiers had the highest mental health of all tenure types. The MCS score was positively associated with age and men reported higher MCS than women.
A.M. Kavanagh et al. / Social Science & Medicine 151 (2016) 225e232
229
Fig. 2. Flow diagram of sample selection and missing data (MCS ¼ mental component summary; obs ¼ observations; ppl ¼ people).
3.2. Regression analyses
3.6. Sensitivity analyses
There was statistical evidence of EMM of the association between disability acquisition and mental health by all housing variables examined, which supported the inclusion of a cross-product term between disability acquisition and housing characteristics.
Results of our main analyses were robust to sensitivity analyses. Exclusion of people with psychological impairments attenuated the results however the general patterns were similar (Supplementary Table C). The results of the complete case analysis related to housing affordability (excluding the observations where income was imputed) were almost identical to the main analyses (Supplementary Table D).
3.3. Housing tenure While the mental health of people from all tenure types decreased after disability acquisition, the largest difference was for those in private rental housing whose mental health was 2.8 points lower (95% CI 3.9, 1.8) in waves reporting disability, compared to a difference of 1.6 points for those in outright ownership (95% CI 2.4, 0.1) (Table 3). 3.4. Housing affordability Living in unaffordable housing was associated with a 4.2 point lower mental health score in waves reporting disability compared to no disability (4.2, 95% CI 5.2, 1.4), compared to a 1.7 point decline for those in affordable housing (1.7, 95% CI 2.1, 1.3) (Table 3). 3.5. Housing affordability and tenure type Both mortgagors and private renters in unaffordable housing had the largest reductions in mental health following acquisition of a disability with reductions of 5.3 (95% CI 8.8, 1.9) and 4.0 (95% CI 6.3, 1.6) respectively (Table 4).
4. Discussion Our results suggest that housing characteristics prior to disability acquisition in adulthood modify the effect of disability acquisition on mental health. While we observed a small reduction in mental health for people who were outright owners or mortgagors, the effect was largest for those living in private rental. Living in unaffordable housing was associated with a four point reduction in the mental health score. Unlike previous analyses in the general population (Mason et al., 2013) where affordability was important for private renters only, we found evidence that both mortgagors and renters who were in unaffordable housing had a large decline in mental health associated with disability acquisition. We are unaware of any previous studies that have examined whether housing characteristics are effect modifiers of the relationship between disability acquisition and mental health. A previous study in the United States found a greater decline in subjective wellbeing following disability acquisition among those with below median wealth however this was a much smaller study
230
A.M. Kavanagh et al. / Social Science & Medicine 151 (2016) 225e232
Table 1 Characteristics of the analytic sample at baseline (individuals not observations, n ¼ 1913). N Age group (years) <30 30e44 45e60 60þ Sex Women Men Occupation High skilled Medium skilled Low skilled Not in the labour force Unemployed Income Q5 (highest) Q4 Q3 Q2 Q1 (lowest) Housing tenure Outright owner Servicing a mortgage Private renter Public renter Housing affordability Affordable Unaffordable Housing tenure# affordability Mortgage e affordable Mortgage e unaffordable Private renter e affordable Private renter e unaffordable
%
305 492 527 589
15.9 25.7 27.6 30.8
1075 838
56.2 43.8
411 392 296 754 60
21.5 20.5 15.5 39.4 3.1
248 312 373 390 590
13.0 16.3 19.5 20.4 30.8
943 531 365 74
49.3 27.8 19.1 3.9
1783 130
93.2 6.8
489 42 286 79
54.6 4.7 31.9 8.8
(478 people) with fewer waves of data and did not specifically assess housing assets or affordability (Smith et al., 2005). In this paper we used fixed-effects regression to examine within-person change in mental health e an approach we and others have used in a number of other studies of psychosocial working conditions (Butterworth and Crosier, 2004), employment arrangements (LaMontagne et al., 2014), housing affordability (R. Bentley et al., 2011), and job quality (R Bentley et al., 2015; Milner et al., 2015) in HILDA. The effect estimates for other exposures have been in the range of 1e2.5 points on the MCS scale. In this paper we find larger effect sizes particularly relation to housing affordability, highlighting the importance of these findings. A strength of this study is that it is based on a nationally representative study of nearly 2000 people who acquire a disability. In addition, fixed-effects longitudinal regression removes confounding due to time-invariant variables because comparisons are made within people (Gunasekara et al., 2013). Limitations include the potential for error due to missing data particularly on income (20% missing observations) however complete case analysis of the housing affordability analysis yielded similar results. Dependent misclassification is also a possibility because both disability and mental health are self-reported. In theory individual characteristics such as negative affectivity might result in dependent measurement error; however, to the extent that these characteristics are stable over time, fixed-effects regression will remove this bias. Selection bias due to attrition is a potential limitation however retention rates in HILDA are high (>90% at most waves). We acknowledge that the ratio based definition of housing unaffordability (30/40 rule) is somewhat blunt, however it is an accepted and widely-used cut-off, and has been tested and applied in Australia (E. Baker et al., 2013; Nepal et al., 2010). Because housing
Table 2 Mean within-person MCS score in waves reporting disability and no disability.
Whole sample Age group (years) <30 30e44 45e60 60þ Sex Women Men Occupation High skilled Medium skilled Low skilled Not in the labour force Unemployed Income Q5 (highest) Q4 Q3 Q2 Q1 (lowest) Housing tenure Outright owner Servicing a mortgage Private renter Public renter Housing affordability Affordable Unaffordable Housing tenure# affordability Mortgage - affordable Mortgage - unaffordable Private renter - affordable Private renter e unaffordable
No disability
Disability
49.0 (48.6, 49.4)
47.1 (46.6, 47.6)
44.8 46.5 48.7 52.8
41.1 43.1 46.6 50.6
(43.7, (45.7, (48.0, (52.2,
45.9) 47.3) 49.4) 53.3)
(39.4, (42.0, (45.8, (50.0,
42.7) 44.1) 47.4) 51.2)
48.3 (47.7, 48.8) 49.9 (49.3, 50.4)
46.3 (45.6, 46.9) 48.1 (47.5, 48.8)
48.7 48.0 48.2 50.1 43.0
(48.0, (47.3, (47.3, (49.5, (41.0,
49.5) 48.7) 49.0) 50.7) 45.0)
47.9 46.2 45.1 47.2 40.1
(47.0, (45.3, (43.9, (46.5, (37.7,
48.8) 47.1) 46.4) 47.9) 42.6)
49.4 48.4 49.1 48.4 49.0
(48.5, (47.6, (48.4, (47.7, (48.4,
50.3) 49.2) 49.7) 49.1) 49.7)
47.5 47.3 46.8 46.7 46.5
(46.5, (46.4, (45.9, (45.9, (45.7,
48.6) 48.2) 47.6) 47.5) 47.3)
50.9 47.6 46.6 46.3
(50.3, (46.8, (45.7, (44.0,
51.4) 48.3) 47.6) 48.6)
49.3 45.8 43.7 44.5
(48.7, (45.0, (42.5, (41.9,
49.9) 46.7) 44.9) 47.1)
49.1 (48.7, 49.5) 47.0 (45.3, 48.7)
47.4 (46.9, 47.8) 43.3 (41.2, 45.4)
47.5 48.2 46.4 47.4
46.1 42.8 43.7 43.9
(46.8, (45.1, (45.4, (45.2,
48.3) 51.2) 47.5) 49.6)
(45.2, (38.7, (42.3, (41.2,
47.0) 47.0) 45.0) 46.6)
Table 3 Linear fixed-effects regression coefficients for the within-person difference in MCS score between waves reporting disability and no disability (n ¼ 1913, observations ¼ 13 037) for categories of housing variables separately*
Housing tenure Outright owner Servicing a mortgagea Private renterb Public renterc Housing affordability Affordable Unaffordabled
Coeff.
95% CI
P Value
1.6 1.7 2.8 2.2
2.1, 2.5, 3.9, 3.9,
1.1 1.0 1.8 0.5
<0.001 <0.001 <0.001 0.012
1.7 4.2
2.1, 1.3 5.2, 1.4
<0.001 <0.001
*Adjusted for age, employment and equivalised household disposable income. a Interaction term: mortgage (0.2, 95% CI -1.1, 0.7, p ¼ 0.700). b Interaction term: private renter (1.3, 95% CI -2.4, -0.1, p ¼ 0.034). c Interaction term: public renter (0.6, 95% CI -2.4, 1.2, p ¼ 0.492). d Interaction term: unaffordable housing (2.5, 95% CI -4.4, -0.5, p ¼ 0.015).
characteristics are modelled as time-invariant variables classified prior to disability acquisition, we cannot estimate the main effects of housing variables although we are able to estimate the values of their interaction terms with disability acquisition. People with severe disabilities, particularly intellectual and psychological disabilities, may be less likely to participate in HILDA and it is possible that the relationships between disability, housing and mental health are different in this group. While most Australians housing is physically sound, recent work (Baker et al., 2016) has highlighted that subgroups of the Australian population (e.g. indigenous Australians) live in housing of poor physical quality are vulnerable to the same health problems documented in other countries (Gibson et al., 2011; Krieger and Higgins, 2002); we have not investigated
A.M. Kavanagh et al. / Social Science & Medicine 151 (2016) 225e232 Table 4 Linear fixed-effects regression coefficients for the difference in MCS score between waves reporting disability and no disability according to tenure/affordability (n ¼ 896, observations ¼ 5913)*
Housing tenure# affordability Mortgage - affordable Mortgage - unaffordablea Private renter - affordableb Private renter - unaffordablec
Coeff.
95% CI
P Value
1.4 5.3 2.4 4.0
2.1, 8.8, 3.6, 6.3,
<0.001 0.003 <0.001 0.001
0.7 1.9 1.3 1.6
*Adjusted for age, employment and equivalised household disposable income. a Interaction term: mortgage - unaffordable (4.0, 95% CI -7.5, -0.4, p ¼ 0.030). b Interaction term: private renter - affordable (1.0, 95% CI -2.4, 0.3, p ¼ 0.144). c Interaction term: private renter e unaffordable (2.6, 95% CI -5.1, -0.1, p ¼ 0.042).
these issues in this study. Although we did include the variables we postulated to be the most important time-varying confounders (age, employment and occupation, and income) it is possible that confounding due to time-varying confounders not included in the model remains. Models that additionally tested for household structure and education had almost identical results. Ten percent of the eligible sample had missing data on one or more of the variables. Young and older age groups and people from the lowest two quintiles of income were more likely to have missing data. Selection bias is possible if the relationships between housing characteristics, disability and mental health varied for those with missing data. Finally, the findings of this study may not be generalizable to other settings as there are considerable variations between countries with respect to home ownership, affordability, tenure and housing policies. In sum, we find evidence to support EMM by housing affordability and tenure of the association between disability acquisition and mental health. Importantly, the small proportion of people living in unaffordable housing at the time of disability acquisition (approximately seven percent of our sample), are likely to experience more severe mental health effects with potential consequences for ongoing workforce participation and health costs. As this study models housing conditions prior to disability acquisition the findings suggest that the ideal time to intervene (e.g. with financial support) is prior to acquisition which cannot be known in advance. It is possible that financial support to meet housing costs for those living in unaffordable housing at the time of acquisition would also protect against deterioration in mental health however that would best be tested in an experimental or quasi-experimental study. Appendix A. Supplementary data Supplementary data related to this article can be found at http:// dx.doi.org/10.1016/j.socscimed.2016.01.010. References ABS, 2011. Household Expenditure Survey, Australia: Summary of Results, 20092010 (6530.0): ABS. ABS, 2013. Household Income and Income Distribution, Australia, 2011-12 (6523.0): ABS. ABS, 2014. Australian Social Trends (4102.0). March 2014: ABS. ABS, 2015. Australian National Accounts: Finance and Wealth, Dec 2014 (5232.0): ABS. Allison, P.D., 2009. Fixed Effects Regression Models (Quantitative Applications in the Social Sciences). Sage Publications, London. Baker, E., Bentley, R., Mason, K., 2013. The mental health effects of housing tenure: causal or compositional? Urban Stud. 50 (2), 426e442. http://dx.doi.org/ 10.1177/0042098012446992. Baker, E., Lester, L., Bentley, R., Beer, A., 2016. The health effects of poor quality housing: Australia's hidden fraction. J. Prev. Intervention Community (in press). Bauer, G.R., 2014. Incorporating intersectionality theory into population health
231
research methodology: challenges and the potential to advance health equity. Soc. Sci. Med. 110, 10e17. http://dx.doi.org/10.1016/j.socscimed.2014.03.022. Beer, A., Baker, E., Mallett, S., Batterham, D., Pate, A., Lester, L., 2012. Addressing Homelessness Amongst Persons with a Disability: Identifying and Enacting Best Practice. Department of Families, Housing, COmmunity Services and Indigenous Affairs, Adelaide. Beer, A., Faulkner, D., 2008. The Housing Careers of People with a Disability and Carers of People with a Disability. Beer, A., Faulkner, D., Paris, C., Clower, T., 2011. Housing Transitions Through the Life Course: Aspirations, Needs and Policy. The Policy Press, Bristol. Bentley, R., Baker, E., Mason, K., Subramanian, S.V., Kavanagh, A.M., 2011. Association between housing affordability and mental health: a longitudinal analysis of a nationally representative household survey in Australia reply. Am. J. Epidemiol. 174 (7), 753e760. http://dx.doi.org/10.1093/Aje/Kwr161. Bentley, R., Krnjacki, L., Kavanagh, A., Lamontagne, A.D., 2015. A longitudinal analysis of changes in job control and mental health. Am. J. Epidemiol. 182 (4), 328e334. Braubach, M., 2011. Key challenges of housing and health from WHO perspective. Int. J. Public Health 56, 579e580. Burchardt, T., 2003. Being and becoming: social exclusion and the onset of disability. In: Centre for Analysis of Social Exclusion. The London School of Economics.,, London. Butterworth, P., Crosier, T., 2004. 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 4. http://dx.doi.org/10.1186/1471-2458-4-44. Dalton, T., Ong, R., 2007. Welfare to work in Australia: disability income support, housing affordability and employment incentives. Int. J. Hous. Policy 7, 275e297. Demographia, 2014. 10th Annual Demographia International Housing Affordability Survey (2013: 3rd Quarter). Demographia, New York. Ellaway, A., Macintyre, S., 1998. Does housing tenure predict health in the UK because it exposes people to different levels of housing related hazards in the home or its surroundings? Health & Place 4 (2), 141e150. http://dx.doi.org/ 10.1016/S1353-8292(98) 00006e9. Pii S1353-8292(98)00006-9. Emerson, E., Hatton, C., 2007. Poverty, socio-economic position, social capital and the health of children and adolescents with intellectual disabilities in Britain: a replication. J. Intellect. Disabil. Res. 51 (11), 866e874. JIR951 [pii] 10. 1111/ j.1365e2788.2007.00951.x [doi]. Emerson, E., Kariuki, M., Honey, A., Llewellyn, G., 2014. 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. 7 (4), 448e456. http://dx.doi.org/10.1016/ j.dhjo.2014.03.004. Emerson, E., Llewellyn, G., Honey, A., Kariuki, M., 2012. Lower well-being of young Australian adults with self-reported disability reflects their poorer living conditions rather than health issues. Aust. N. Z. J. Public Health, 36 (2), 176e182. http://dx.doi.org/10.1111/j.1753-6405.2011.00810.x. Evans, J., Hyndman, S., Stewart-Brown, S., Smith, D., Petersen, S., 2000. An epidemiological study of the relative importance of damp housing in relation to adult health. J. Epidemiol. Community Health 54, 677e686. Free, S., Howden-Chapman, P., Pierse, N., Viggers, H., the Housing, H. a. H. S. R. T, 2010. More effective home heating reduces school absences for children with asthma. J. Epidemiol. Community Health 64, 379e386. Gibson, M., Petticrew, M., Bambra, C., AJ, S., Wright, K., Whitehead, M., 2011. Housing and health inequalities:A synthesis of systematic reviews of interventions aimed at different pathways linking housing and health. Health Place 17, 174e185. Greenland, S., Pearl, J., Robins, J.M., 1999. Causal diagrams for epidemiologic research. Epidemiology 10 (1), 37e48. Gunasekara, F.I., Richardson, K., Carter, K., Blakely, T., 2013. Fixed effects analysis of repeated measures data. Int. J. Epidemiol. 43 (1), 264e269. http://dx.doi.org/ 10.1093/ije/dyt221. Haagenars, A., de Vos, K., Zaidi, A., 1994. Poverty Statistics in the Late 1980s: Research Based on Micro-data. Office for Official Publications of the European Communities, Luxemburg. Hahn, M.H., Haisken-DeNew, J.P., 2013. PanelWhiz and the australian longitudinal data infrastructure in economics. Aust. Econ. Rev. 46 (3), 379e386. http:// dx.doi.org/10.1111/j.1467-8462.2013.12010.x. Honey, A., Emerson, E., Llewellyn, G., 2011. The mental health of young people with disabilities: impact of social conditions. Soc. Psychiatry Psychiatr. Epidemiol. 46 (1), 1e10. http://dx.doi.org/10.1007/s00127-009-0161-y. Imrie, R., 2006. Accessible Housing: Quality, Disability and Design. Taylor & Francis. Jenkins, S.P., Rigg, J.A., 2004. Disability and disadvantage: selection, onset, and duration effects. J. Soc. Policy 33, 479e501. http://dx.doi.org/10.1017/ S0047279404007780. Jones, M., Mavromaras, K., Sloane, P., Wei, Z., 2014. Disability, job mismatch, earnings and job satisfaction in Australia. Camb. J. Econ. 38 (5), 1221e1246. http:// dx.doi.org/10.1093/Cje/Beu014. Jones, M.K., Sloane, P.J., 2010. Disability and skill mismatch. Econ. Rec. 86, 101e114. http://dx.doi.org/10.1111/j.1475-4932.2010.00659.x. Kavanagh, A.M., Aitken, Z., Krnjacki, L., LaMontagne, A.D., Bentley, R., Milner, A., 2015. Mental health following acquisition of disability in adulthood-the impact of wealth. PLoS One 10 (10), e0139708. http://dx.doi.org/10.1371/ journal.pone.0139708.
232
A.M. Kavanagh et al. / Social Science & Medicine 151 (2016) 225e232
Kavanagh, A.M., Krnjacki, L., Aitken, Z., LaMontagne, A.D., Beer, A., Baker, E., Bentley, R., 2014. Intersections between disability, type of impairment, gender and socio-economic disadvantage in a nationally representative sample of 33,101 working-aged Australians. Disabil. Health J. 8 (2), 191e199. http:// dx.doi.org/10.1016/j.dhjo.2014.08.008. Kavanagh, A.M., Krnjacki, L., Beer, A., Lamontagne, A.D., Bentley, R., 2013. Time trends in socio-economic inequalities for women and men with disabilities in Australia: evidence of persisting inequalities. Int. J. Equity Health 12 (1), 73. http://dx.doi.org/10.1186/1475-9276-12-73. Krieger, J., Higgins, D.L., 2002. Housing and health: time again for public health action. Am. J. Public Health 92 (5), 758e768. http://dx.doi.org/10.2105/ Ajph.92.5.758. Kyle, T., Dunn, J., 2008. Effects of housing circumstances on health, quality of life and healthcare use for people with severe mental illness: a review. Health Soc. Care Community 16 (1), 1e15. LaMontagne, A.D., Milner, A., Krnjacki, L., Kavanagh, A.M., Blakely, T.A., Bentley, R., 2014. Employment arrangements and mental health in a cohort of working australians: are transitions from permanent to temporary employment associated with changes in mental health? Am. J. Epidemiol. 179 (12), 1467e1476. http://dx.doi.org/10.1093/Aje/Kwu093. Little, R.J.A., Su, H.-L., 1989. Item Nonresponse in Panel Surveys. John Wiley & Sons, New York. Macintyre, S., Ellaway, A., Hiscock, R., Kearns, A., Der, G., McKay, L., 2003. What features of the home and the area might help to explain observed relationships between housing tenure and health? evidence from the west of Scotland. Health & Place 9 (3), 207e218. http://dx.doi.org/10.1016/S1353-8292(02) 00040e0. Pii S1353-8292(02)00040-0. Mandemakers, J.J., Monden, C.W.S., 2010. Does education buffer the impact of disability on psychological distress? Soc. Sci. Med. 71 (2), 288e297. http:// dx.doi.org/10.1016/j.socscimed.2010.04.004. Mason, K., Baker, E., Blakely, T., Bentley, R., 2013. Housing affordability and mental health: does the relationship differ for renters and home purchasers? Soc. Sci. & Med. 94, 91e97. Milner, A., Butterworth, P., Bentley, R., Kavanagh, A.M., LaMontagne, A.D., 2015. Sickness absence and psychosocial job quality: an analysis from a longitudinal survey of working australians, 2005-2012. Am. J. Epidemiol. 181 (10), 781e788. http://dx.doi.org/10.1093/aje/kwu355. Milner, A., LaMontagne, A.D., Aitken, Z., Bentley, R., Kavanagh, A.M., 2014. Employment status and mental health among persons with and without a disability: evidence from an Australian cohort study. J. Epidemiol. Community Health 68 (11), 1064e1071. http://dx.doi.org/10.1136/jech-2014-204147. Nepal, B., Tanton, R., Harding, A., 2010. Measuring housing stress: how much do definitions matter? Urban Policy Res. 28 (2), 211e224. http://dx.doi.org/ 10.1080/08111141003797454. Papworth Trust, 2011. Disability in the United Kingdom 2011: Facts and Figures. Paworth Trust, Cambridge. Parker, S., Fisher, K., 2010. Facilitators and barriers in Australian disability housing support policies: using a human rights framework 30.3/4(2010). Disabil. Stud.
Q. 30. Pevalin, D.J., Taylor, M.P., Todd, J., 2008. The dynamics of unhealthy housing in the UK: a panel data analysis. Hous. Stud. 23 (5), 679e695. http://dx.doi.org/ 10.1080/02673030802253848. Polidano, C., Vu, H., 2015. Differential labour market impacts from disability onset. Health Econ. 24 (3), 302e317. http://dx.doi.org/10.1002/hec.3017. Pollack, C., Griffin, B., Lynch, J., 2010. Housing affordability and health among homeowners and renters. Am. J. Prev. Med. 39 (6), 515e521. Pollack, C.E., von dem Knesebeck, O., Siegrist, J., 2004. Housing and health in Germany. J. Epidemiol. Community Health 58 (3), 216e222. http://dx.doi.org/ 10.1136/jech.2003.012781. Shaw, M., 2004. Housing and public health. Annu. Rev. Public Health 25, 397e418. Smith, D.M., Langa, K.M., Kabeto, M.U., Ubel, P.A., 2005. Health, wealth, and happiness: financial resources buffer subjective well-being after the onset of a disability. Research Support, N.I.H., Extramural Research Support, Non-U.S. Gov't Research Support, U.S. Gov't, Non-P.H.S. Psychol. Sci. 16 (9), 663e666. http://dx.doi.org/10.1111/j.1467-9280.2005.01592.x. StataCorp, 2011. Stata Statistical Software: Release 12. StataCorp LP, College Station, TX. Summerfield, M., Freidin, S., Hahn, M., Ittak, P., Li, N., Macalalad, N., Wooden, M., 2013. HILDA user manual - release 12. In: Melbourne Institute of Applied Economic and Social Research. Turner, R.J., Noh, S., 1988. Physical disability and depression - a longitudinal analysis. J. Health Soc. Behav. 29 (1), 23e37. http://dx.doi.org/10.2307/2137178. VanderWeele, T.J., Knol, M.J., 2014. A Tutorial on Interaction Epidemiologic Methods, vol. 3, p. 33. Wang, 2005. Disability and American Families: 2000 Census Special Reports: US Census Bureau. Watson, N., 2004. Income and Wealth Imputation for Waves 1 and 2 HILDA Project Technical Paper Series No. 3/04, July 2004. Melbourne Institute of Applied Economic and Social Research, University of Melbourne, Melbourne. White, B.J., Peaslee, J., LaQuatra, J., 1994. Comparing housing affordability and quality among disability households: the United States and its regions. J. Fam. Econ. Issues 15 (4), 367e380. Williamson, E.J., Aitken, Z., Lawrie, J., Dharmage, S.C., Burgess, J.A., Forbes, A.B., 2014. Introduction to causal diagrams for confounder selection. Respirology 19 (3), 303e311. http://dx.doi.org/10.1111/Resp.12238. World Health Organization, 2002. Towards a Common Language for Functioning, Disability and Health: ICF the International Classification of Functioning, Disability and Health. WHO, Geneva. World Health Organization and World Bank Group, 2011. World Report on Disability. WHO, Geneva. Yang, Y., 2006. How does functional disability affect depressive symptoms in late life? the role of perceived social support and psychological resources. J. Health Soc. Behav. 47 (4), 355e372. Yates, J., Milligan, V., Berry, M., Burke, T., Gabriel, M., Pinnegar, S., 2007. Housing Affordability: A 21st Century Problem AHURI Final Report No. 105. Australian Housing and Urban Research Institute, Sydney Research Centre, Melbourne.