Aims
This study analysed a pre-collected, nationally representative household survey on disability and informal caring. The aims were to identify: (1) whether co-resident mental health carers are more disadvantaged in employment than carers of people with other types of disabilities and non-carers; (2) which factors are most strongly associated with employment for mental health carers; and (3) whether there are unique factors associated with mental health carers’ employment compared to carers for other conditions.
Survey and sample
The 2015 SDAC [35] was a nationally representative household survey carried out by the Australian Bureau of Statistics (ABS) between July and December 2015. Households were selected from a stratified, multi-stage area sample developed by the ABS. Basic demographic data on all household members were collected from a responsible adult in each household (i.e., the first adult with whom the interviewer made contact who was able to participate), by trained interviewers using a Computer-Assisted Personal Interview. The responsible adult also answered questions to identify the presence of a person with disability or carer in the household and, where possible, additional interviews were completed with persons with disability and confirmed primary carers (but not other carers) at the same time or a later date within the survey period. Proxy interviews were conducted for people unable to be interviewed due to language or impairment, children aged below 15 years, and people aged 15–17 years without parental consent to participate. The final household sample included 25,806 households comprising 63,515 persons (80.0% response rate).
Key variables
Persons with disability
Persons with a disability were identified by the responsible adult (e.g. “Does anyone in the household have a [nervous or emotional condition] that has lasted, or is likely to last for 6 months or more?”, “Are they restricted in everyday activities because of this condition?”, “Is anyone in the household receiving treatment or medication for any long-term conditions or ailments?”). Household members identified as having a disability were interviewed and provided additional information on: their main disabling condition; all conditions; level of activity limitations; and receipt of formal assistance (services) for their disability.
Informal carers
Carers were identified by the responsible adult (e.g. “Does anyone in the household help or supervise [another member of the household]/ [someone living elsewhere] who has a long-term health condition or disability with everyday types of activities?”, “Do they provide this help on a regular, unpaid, informal basis?”). If not initially identified by the responsible adult, carers could also be subsequently identified by a person with disability living in the household during their personal interview (e.g. “Have you received, or do you expect to receive, assistance to help with these tasks from a partner or spouse/parent, family, friends or neighbours for 6 months or more?”). The 2015 SDAC classified household members as carers where they provided support to someone with a limitation to their mobility, communication or self-care and this support was ongoing, or likely to be ongoing, for at least six months. The responsible adult (or person with disability) initially described the relationship of the carer to the person cared for and the number of people supported; carers were only asked to complete a personal interview if they confirmed that they were the primary carer of a person with disability. The 2015 SDAC identified confirmed primary carers, a subset of all carers in the survey, as the person providing the most assistance to someone with a disability. Confirmed primary carers aged 15 years or more were interviewed separately to collect additional information, including questions about the impact of their caring on employment and working hours.
This study focused on both primary and secondary carers aged 15–64 years to align with the youngest age for Australians commencing employment and aged pension eligibility, after which workforce participation drops significantly. Information on the disabling conditions of care recipients was only available for carers living in the same household, so the analysis was limited to co-resident primary and secondary carers. Four carer groups were created based on the main disabling condition of the person cared for: mental illness (e.g. psychosis, depression, anxiety, personality and behavioural disorders; n = 520); other cognitive/behavioural conditions (e.g. dementia, autism, intellectual disability, acquired brain injury; n = 312); and physical conditions (e.g. musculoskeletal, cardiovascular, neurological and sensory disabilities) with or without a secondary mental illness (n = 577 and n = 1455 respectively). Additional file 1: Table S1, includes the full list of conditions. Carers for more than one person with different conditions were grouped hierarchically, in that order (i.e. mental illness first). Cognitive conditions and secondary mental illness were separately identified because the required behaviour management and fluctuating care needs were expected to have a more detrimental impact on carers’ ability to maintain employment [11]. We focused on carers of adults with disabilities; those supporting only people aged below 15 years were excluded because of the complexities in separating the effects of informal caring on employment from those of normal parenting in a cross-sectional analysis. A comparison group of non-carers included people aged 15–64 years who did not support a person with disability of any age.
Employment
The 2015 SDAC recorded employment data for participants aged 15 and over. The main outcome of interest for this study was employment status—whether a person is employed or not (unemployed or not in the labour force). The 2015 SDAC defined employment as engaging in economic work of one hour or more in the survey reference week. Full-time employment is permanent, temporary or casual employment of 35 h or more per week (across all jobs), or working 35 h or more during the reference week even if the person usually works fewer hours [36]. Part-time employment is working fewer than 35 h per week [36]. Persons were classified as unemployed if they were aged 15 or over and worked less than one hour in the reference week, were actively looking for work in the previous four weeks, and were also available to start work [37]. Those who were not employed indicated their main activity since last looking for work. We also examined potential indicators of underemployment in the form of hours worked and occupational category.
Data analysis
The ABS supplied a Confidentialised Unit Record File of the 2015 SDAC (October 2016 version). Person-level, recipient-level and condition-level data files were merged to obtain estimates for all co-resident carers and their care recipients. Analyses were conducted in Stata 15 [38], using ABS-provided survey weights to account for possible selection and non-response biases, and differences between the sample and Australian population. Survey-weighted proportions described key demographic and employment characteristics of each carer group, and 95% confidence intervals (CIs) were calculated using jackknife repeated replication.
To compare mental health carers with non-carers and other disability carers on employment status, working hours and occupational group (aim 1), three simple logistic regression models were run with binary outcomes for employment status (not employed vs. employed), working hours (< 16 vs 16+ hours per week), and occupation (machinery operator, driver or labourer vs. technical and professional roles).
Factors potentially associated with employment for mental health carers were identified based on previous studies: carer age group; marital status; rurality; country of birth; highest education level; whether the carer has a disability; whether any person cared for receives formal services for their disability, and the type, frequency of and need for these services; and indicators of caring intensity—including being a confirmed primary carer, number of people cared for, caring for a close family member (spouse/partner or adult child), and caring for someone who is profoundly or severely limited in core activities [1, 6, 18, 30,31,32]. Education level was recorded as ‘not determined’ for 14 of 520 mental health carers and 45 of 2344 other carers; this coding was not significantly related to employment status (χ2(1, N = 2864) = 0.10, p = 0.75), so these carers were excluded from the multivariate regression analyses.
To identify factors associated with employment status for mental health carers (aim 2), multivariate logistic regression models were developed to calculate adjusted odds ratios (AORs) and 95% CIs. Separate models were conducted for male and female carers due to the potentially different relationships by gender [2, 18, 34]. Pairwise Cramer’s V associations between factors revealed moderate relationships (V = 0.37–0.64) between age group, marital status and caring for a partner/child, and between primary carer status and disability level of the person cared for (Additional file 1: Table S2). However, all variance inflation factors were below three, and since these factors each represented distinct constructs of interest they were retained. Unsurprisingly, receipt of formal services was highly correlated with receiving particular types of assistance, frequency of assistance and unmet need for assistance. The latter factors were entered separately into supplementary regression analyses replacing the former in each model (see notes on, Additional file 1: Tables S5-S6). All factors were initially entered into each model and a final model selected via backwards elimination until only predictors with a p-value of <.10 remained. This higher cut-off than the standard significance threshold of p < .05 was chosen to ensure no potentially important variables were excluded [39, 40].
Further logistic regression models by gender were run to identify whether any factors associated with employment were unique to mental health carers (aim 3). These models tested interactions between the disability group of the person supported, selected factors (as below), and employment status. Bivariate chi-square tests identified which factors were significantly different between mental health versus other carers (Additional file 1: Table S3). For each gender, all significant factors as well as those identified as significantly related to employment in the mental health carer models were included in the initial regression analyses as interaction terms with disability group. To minimise loss of statistical power with the addition of interaction terms, education level and disability group were converted to dichotomous variables. Models were reduced via backwards elimination as above.