Occupation type, family demands and mental health: analysis of linked administrative data.

Background: While employment generally promotes positive health and wellbeing, some jobs may be less salutogenic than others. Few studies have examined mental health across a range of broadly defined occupation types using a large population sample. Aims: To examine the prevalence of mental health problems across a wide range of occupation types, and further examine the association of family demands, controlling for key social determinants and health-related factors. Methods: We used linked administrative data from 2011 NI Census returns; NI Properties data; and Enhanced Prescribing Data (EPD) 2011/12. We examined self-reported mental health problems and receipt of psychotropic medication among 553,925 workers aged 25 and 59 years. Results: Self-reported chronic mental ill health was more prevalent among workers in lower paid occupations, while “ public-facing ” occupations had the highest rates of medication. In fully adjusted models, informal caregivers were less likely to report mental health problems but more likely to be in receipt of psychotropic medication, as were lone parents. The association of family demands also varied across occupational groupings. Conclusion: Future development of mental health at work plans should take cognisance of occupation specific mental health risk and wider family circumstances to support workers ’ mental wellbeing most effectively.


Introduction
As an important aspect of personal identity and social status, employment, generally, promotes positive health and wellbeing (Modini et al., 2016), providing structure, social interaction, and purpose.However, some jobs and related activities may be less salutogenic than others.For example, the influential "Whitehall Studies" of work, stress and health among British civil servants observed a clear "social gradient" for a wide range of physical and mental health conditions (Bell et al., 2004).Also, a longitudinal study (O'Reilly and Rosato, 2013) noted an increased risk in allcause mortality among men who worked in "routine occupations" for more than 55 hours per week.
Few large representative studies have examined mental health across a wide range of occupation types and results from available studies vary substantially (Inoue et al., 2010;Llena-Nozal et al., 2004;Marchand, 2007;Marchand et al., 2003;Riise et al., 2003;Sanne et al., 2003;Stansfeld et al., 2011).Using data from the second British Psychiatric Morbidity Survey, Stansfeld et al. (2011) examined the prevalence and risk of common mental disorders (CMDs) among working age individuals: occupations with high CMD prevalence included teachers, welfare and youth workers, and care assistants, while general managers in government and large organisations were among the professions with elevated risk of CMD, after controlling for other characteristics.In a large representative study of 77,377 Canadian workers, Marchand (2007) found high prevalence of mental ill health among machine operators, labourers and cleaners.In a cross-sectional study of 23,312 workers in Hordaland County in Norway, Riise et al. (2003) found that agricultural and fishery workers had the poorest mental health scores, while Sanne et al. (2003) found a distinct inverse relationship between skill levels and depression and anxiety scores in a separate analysis of data from the same study.In contrast, both Marchand et al. (2003) and Inoue et al. (2010) found little evidence of large variation in mental health across occupations.
Aside from job related characteristics, it is likely that family/home-related demands will have, not only a direct impact on mental health outcomes, but will also interact with working demands in influencing mental health.In Canada, Cadieux and Marchand (2014) found distress was associated with marital status, marital stress and parental stress, after controlling for occupation type, working conditions and time.The Whitehall II studies (Bell et al., 2004) found both "family-to-work" and "work-to-family" conflicts had deleterious effects on mental and physical health and again alluded to important gender differences, which may reflect traditional male/female work and family roles and variation in working patterns.Differential findings for males and females with respect to the association of occupation and mental health from other studies also support a genderbased approach to analysis in this area.In an examination of data from the UK National Child Development Survey, Llena-Nozal et al. (2004) found large occupation effects for females but not for males, while employment status was an important predictor of poorer mental health in males but not in females.Echoing the Whitehall II studies, Lennon and Rosenfield (1992) argue that work and family conditions interact in their association with mental health and propose a combined developed a "job" and "family" model in their study of mental ill health.
Despite an extensive wider literature on occupations and mental health, in-depth studies in this area require a large sample of workers across a wide range of occupations with a common measure of mental health (Stansfeld et al., 2011).As far as can be detected, the current study represents the first population-based administrative data study from within the UK to examine variation in mental disorder across a wide range of standard occupation types.Linkage of large clinically relevant datasets has allowed an in-depth examination of mental health and occupations, facilitating analysis of mental health, jobs and family responsibilities that accounts for a wide range of sociodemographic and socioeconomic contextual factors.

Aim
To examine the prevalence of mental health problems across standard occupational classifications (SOCs) among the Northern Ireland (NI) working population.In addition, we further examine the association of family demands with mental ill health across occupation types, controlling for key social determinants and health-related factors.Previous evidence points to gender differences in "family-to-work" and "work-to family" conflict, and preliminary analysis points to significant gender differences in mental health across most occupation types.Thus, we additionally aimed to explore the interaction of sex and family responsibilities with mental health across broad occupational groupings.

Data
We used linked NI administrative data, facilitated within the Administrative Data Research Centre NI (ADRC-NI), under the auspices of Administrative Data Research UK.The following datasets were merged: 2011 individual and household NI Census returns; NI Properties data; and Enhanced Prescribing Data (EPD) 2011/12.Consent was not required for this study as data were provided in anonymised format meaning the rights of individuals are respected with adequate privacy protection.These data are classed as confidential and subject to strict controls: it is held in a safe setting by the NI Statistics and Research Agency (NISRA); is available only to accredited research team members, who access de-identified data after signing Data Access Agreements.These contracts and organisational controls ensure no individual can be re-identified during the analysis or in any of the outputs.All outputs are scrutinised by data custodians, ensuring that privacy and disclosure safeguards are robust.Use of the NI Census data for this research was approved by the East Midlands -Leicester South Research Ethics Committee (REC reference: 18/EM/0053).

Outcome measures
We derived two indicators of the presence of a mental health problem: (1) self-reported chronic mental health problem, based on individuals who responded YES to a question in the 2011 NI Census which asked if an individual had an emotional or mental health condition that lasted, or was expected to last, at least 12 months and (2) receipt of psychotropic medication, which identified individuals prescribed medication in the following Chapter 4 British National Formulary (BNF) categories at any point in the 12month period from April 2011 to March 2012: Hypnotics and anxiolytics; drugs used in psychosis and related disorders and anti-depressants.It should be noted that these two measures are not comparable: the Census question requires persistence over 12 months and therefore may be viewed as a measure of more chronic mental health disorder; while the second indicator, based on receipt of psychotropic medication at any point within a 12-month period, will include individuals with short-term emotional or mental health problems as well as those with more persistent problems.We therefore anticipate higher prevalence rates based on the latter measure.

Explanatory factors
We examined the likelihood of a mental health problem across occupation groups as defined by the Standard Occupational Classification 2000 coding schedule (SOC2000), a hierarchically organised system with nine major and 25 sub-major groupings.Contextual indicators from the Census included: sex (male/female); number of hours worked (less than 15, 16-30, 31-48 and 49 hours or more); and three indications of family-related demands: dependent children (no/yes), informal caregiving (no/yes) and lone parent status (no/yes).Models were controlled for individual characteristics known to be associated with mental disorder, including other self-reported health conditions; age group (five-year bands); marital status (grouped as married, never married and a single group of separated/divorced/widowed); and locale of residence (urban, intermediate, and rural).Three variables representing socio-economic circumstance (SEC) were included: (i) household car availability (grouped as two or more cars, one car, no car access); (ii) educational attainment (university level, intermediate level, no qualifications recorded) and (iii) a combination of housing tenure and rateable value of the property.Rateable value had been derived as part of an exercise by central government in 2010 to determine the level of local residential taxes and combined with tenure to produce a meaningful sixfold gradation: private and social renting; and, for owner-occupiers, five categories ranging from less than £75,000 to over £200,000.

Analysis
Analysis was restricted to persons aged between 25 and 59 years, living in private households and normally resident in NI, who were in paid employment at the 2011 Census (N ¼ 553,925; males ¼ 277,148 and females ¼ 276,777).Descriptive statistics report the prevalence of both selfreported mental disorder (that has lasted or was expected to last for 12 months) and prescription of psychotropic medication at any point over a 12-month period across major and sub-major SOCs.Given consistently reported higher levels of mental disorder among females in the general population, descriptive analysis was stratified by sex.Logistic regression models for the overall sample of working adults report odds ratios (ORs) of both self-reported CMD and psychotropic medication associated with major standard occupational classification, working hours and family demands.In further logistic regression models, stratified by major SOC, a sex interaction term was introduced to examine the differential association of family responsibilities for males and females across broad occupation types.The models were informed by likelihood ratio tests, which compared (1) the main association of sex and each of the indicators of family responsibilities with the two indicators of mental ill health with (2) an interaction term for sex and each family demands indicator.These comparisons were stratified by each of the nine major SOCs and interaction terms included in fully adjusted models where there was evidence of effect modification.Fully adjusted models controlled for any other self-reported health condition and individual socio-demographic and socio-economic characteristics.All analyses were completed using Stata version 15.0 (StataCorp, 2015).The results of our study are reported according to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement guidelines (Vandenbroucke et al., 2007).

Prevalence of mental health problems
Table 1 shows male and female prevalence of a mental health problem based on both self-report and psychotropic medication for across nine major and 20 sub-major SOCs.Overall, 3.10% of workers reported a chronic mental health condition lasting 12 months, while 17.95% had any mental health prescription over a 12-month period.Prevalence rates among all workers are consistent with gender differences, with females more likely to have a mental health condition based on both outcomes.Analysis relating to any self-reported mental health Table 1.Prevalence of common mental disorder across major and sub-major SOCs for all working adults, males and females aged 25-59.

Self-reported mental disorder
Any 12-month psychotropic medication prescription SOCmajor and sub-major categories Overall % (n condition across the major SOCs shows the highest prevalence rates (all approximately 4%) among elementary occupations, sales and customer service, and process, plant, and machine operatives.Within sub-major categories highest rates (also approximately 4%) were among customer service occupations, elementary and related trades/elementary administrative and service, and sales occupations.Results based on psychotropic medication show a different pattern.Of the major SOCs caring, leisure and other service (26.33%), sales and customer service (24.00%) and administrative and secretarial (21.71%) occupations recorded the highest prevalence rates.Of the submajor categories, caring and personal service (27.16%), secretarial and related (24.72%) and sales (24.17%) occupations recorded the highest rates.

Risk of mental ill health associated with occupation type and family demands
Self-reported mental health Univariable analysis of self-reported mental health disorder (Table 2), shows that compared to "managers, directors and senior officials", likelihoods were higher among all other major SOCs, except for "professional occupations", which recorded lower likelihood (OR ¼ 0.80, 95% CI 0.75-0.86).In fully adjusted models, higher likelihoods remained among "associate professional and technical", "administrative", "skilled trades" and "sales and customer service", and "process, plant and machine operative" occupations.Females were more likely than men to have a self-reported disorder, an effect attenuated in the fully adjusted model (OR ¼ 1.05, 1.01-1.10).We noted a trend of reduced likelihood of self-reported disorder as working hours increase.Informal caregiving appears to have minor influence in the fully adjusted model (OR ¼ 0.96, 0.92-0.99)and the higher risk among lone parents in univariable analysis diminishes in fully adjusted analysis.Workers with dependent children had reduced risk of self-reported CMD (OR ¼ 0.86, 0.85-0.89).

Psychotropic medication
For psychotropic medication (Table 2), in both univariable and fully adjusted models, there were higher likelihoods among "associate professional and technical", "administrative and secretarial"; "caring, leisure and other service"; and "sales and customer service occupations".In contrast, "skilled trades" and "process, plant and machine operatives" had lower likelihood of prescriptions (OR ¼ 0.89, 0.85-0.92and OR ¼ 0.92, 0.88-0.95 in fully adjusted models).Female workers were more likely to have received a prescription, while working longer hours and presence of dependent children were again protective.Informal caregiving and lone-parenthood were associated with higher risk of psychotropic medication prescriptions (OR ¼ 1.15, 1.13-1.17and OR ¼ 1. 19, 1.15-1.22, respectively).
Full results from the fully adjusted models presented in Table 2, showing associations with all explanatory variables, are available in supplementary material.groupings, from fully adjusted logistic regression models, reporting the ORs associated with family demands.They also include interaction terms for gender and each of the family responsibility indicators where there was evidence of an interaction effect based on likelihood ratio tests.
While there was no evidence of a main association between informal caregiving and medication among those working in caring occupations, interaction with sex indicates that female informal caregivers recorded a higher likelihood of psychotropic medication (OR ¼ 1.26, 1.06-1.50).Lone parenthood was associated with higher levels of psychotropic medication among associate professionals (OR ¼ 1.25, 1.12-1.39),administrative workers (OR ¼ 1.18, 1.12-1.26),caring.occupations (OR ¼ 1.20, 1.11-1.29)and process and machine operatives (OR ¼ 1.29, 1.14-1.46);while female lone parents in professional occupations also recorded a higher likelihood of medication (OR ¼ 1.33, 1.06-1.68).

Our findings
This study provides the first estimates of mental health problems among workers across all occupation types based on population wide administrative data.We examined both chronic self-reported mental health problems and receipt of any prescribed psychotropic medication over a 12-month period.Almost a fifth of workers had any psychotropic medication prescription.Interestingly, a much lower percentage reported experiencing a chronic mental disorder.We found that self-reported chronic mental ill health was more prevalent among workers in lower paid occupations, while "public-facing" occupations such as those working in caring, leisure and other personal services had the highest rates of medication, after controlling for other factors, including indicators of SEC.We also noted lower risk of psychotropic medication among skilled workers in fully adjusted analysis.In analysis of all workers, consistent findings across both outcomes include higher risk among females, lower risk associated with working greater hours, and lower risk among workers with dependent children.
Inconsistencies are evident, however, with respect to other indicators of family demands: following adjustment for other social determinants and the presence of other health conditions, informal caregivers were less likely to report a mental health problem but more likely to be in receipt of psychotropic medication; while lone parents were no more likely to self-report a mental condition but had higher risk of medication.Interaction analysis stratified by occupation underlines the need for a gender and occupation-based approach to supporting the mental health needs of workers.
Our findings show a considerable difference between the prevalence of self-reported mental health provided by the Census and prevalence indicated by the receipt of psychotropic medications.Several factors may explain this discrepancy.First, the Census question requires persistence over 12 months rather than short-term emotional or mental health problems that are likely to form a substantial subpopulation of those receiving medication.Second, people with a mental illness may be overrepresented as nonrespondents on the Census.Third, the stigma of mental illness and/or misrecognition may have contributed to low rates of self-report.
Conversely, psychotropic medication may be used for physical health problems, tricyclic anti-depressants for back pain, for example (Staiger et al., 2003) or hypnotics and anxiolytics for transient conditions, such as sleep disorders (Ohayon and Lader, 2002).Despite these caveats, our estimate of 18% based on psychotropic medication is in line with previously reported figures on workers (McManus et al., 2016).Given the greater uncertainties of self-reported chronic mental conditions obtained from Census data, our medication-based findings may provide a more reliable indication of "current" need among the working population.
The remainder of our discussion therefore considers the context and implications of our findings based on receipt of medication.We found higher risk in public-facing occupations, such as sales, administrative or caring roles, all often characterised by low autonomy, higher job insecurity and effort-reward imbalance and linked to poorer mental health outcomes (Clausen et al., 2021;Meltzer et al., 2010;Niedhammer et al., 2006).This broad group includes childcare providers and caregiving occupations associated with high physical and emotional demands.Adverse mental health outcomes among care workers have been evidenced in earlier studies, which highlight the impact of working conditions, role stress and burnout (Acker, 2012;Curran et al., 2021;Peterson et al., 2008;Rantonen et al., 2017).Reduced risk among workers with skilled trade occupations is consistent with evidence on the latent benefits of working, including higher levels of autonomy and greater job satisfaction among workers who use their skills in their everyday work (Jahoda, 1988;Warr, 1987).
Higher likelihood of mental health problems among females is consistent with previous research on workers and mental health (Inoue et al., 2010;Stansfeld et al., 2011) and reflective of wider mental health epidemiology (Boyd et al., 2015).We also found that the likelihood of mental health problems decreased as working hours increased compared to those who worked 15 hours or less.However, those on part-time working are more likely to be in poor health and disability, and on welfare support; factors associated with mental health problems.Unfortunately, we were unable to examine the impact of shift work, which has been shown to have detrimental mental health effects if undertaken for a prolonged period (Bara & Arber, 2009).
Our study also sought to explore the association of family demands with mental health problems among workers.Individuals who manage multiple roles in relation to childcare and family caregiving may be at increased risk of stress and adverse mental health outcomes due to competing responsibilities of family and work demands (Chandola et al., 2004;Melchior et al., 2007;Oomens et al., 2007).Our analysis points to elevated risk of psychotropic medication among informal caregivers, as well as lone parents, reflecting the role strain perspective alluded to by Oomens et al. (2007).Our study, however, found a protective main effect for the presence of dependent children (for both selfreported CMD and psychotropic medication), which aligns with the role accumulation perspective (Oomens et al., 2007), which proposes that multiple social roles generally have a positive effect on mental health.
Further interaction analysis stratified by occupation underlines the need for a gender and occupation-based approach to supporting the mental health needs of workers and particular consideration of risk of mental health difficulties among workers with family responsibilities.Our analysis highlights the needs of lone parents across a range of occupations, and of informal caregivers in more skilled/professional occupation types.Interaction analysis suggests higher risk of psychotropic medication among females with dependent children in higher paid occupations.Finally, female informal caregivers working in "caring, leisure and other personal service" occupations had higher risk of medication, underlining mental health needs of females who care for others in both an informal and formal capacity.

Strengths and limitations
In this study, we extend the evidence base related to occupations and mental health by providing the first estimates of mental health problems among workers across occupation types based on population wide administrative data.Thus, the availability of a large population-based dataset allowed a nuanced analysis of the risk of mental health problems across occupations and the association of family demands, controlling for other key characteristics.While evidence on the validity of self-reported mental disorders from Census data is limited, earlier studies on self-report of chronic conditions point to reasonable accuracy (Chapman, 2012;Lee, 2000).Nonetheless, we have previously noted several factors that may contribute to low rates of self-report.We also acknowledge limitations relating to our second indicator based on prescriptions for psychotropic medication.Estimates are based on analysis of EPD, an administrative data source, detailing prescriptions dispensed only.Details of the conditions that these medications were prescribed for was not available and the impact of non-psychiatric prescribing on estimates has also previously been considered.

Conclusion
The United Kingdom (UK) is facing a major mental health at work challenge, with one in six workers experiencing a mental health problem at any one time, costing businesses an estimated £42-£45 billion annually due to sick leave, reduced productivity and staff turnover (Deloitte, 2020).The government commissioned "Thriving at Work" review (Department for Work and Pensions and Department for Health and Social Care, 2017) outlines core standards that should be adopted by all UK companies, including implementation of a mental health at work plan.Employers and managers have a moral obligation, as well as economic motivation, to foster mental wellbeing in the workplace and should develop tailored work plans that consider occupation specific risks and needs of their employees, based on work/family dynamics and other risk factors.
Our study, based on population wide administrative data, shows that mental ill-health rates, both in terms of selfreport and psychotropic medication, vary substantially across major and sub-major and occupation categories.The risk factors are complex, and we provide insight into the association of occupation type, sex, and family demands.Future development of mental health at work plans should take cognisance of both occupation specific mental health risk and wider family circumstances to support workers' mental wellbeing most effectively.

Acknowledgments
The Administrative Data Research Network takes privacy protection very seriously.All information that directly identifies individuals will be removed from the datasets by trusted third parties, before researchers get to see it.All researchers using the Network are trained and accredited to use sensitive date safely and ethically, they will only access the data via a secure environment, and all of their findings will be vetted to ensure they adhere to the strictest confidentiality standards .The help provided by the staff of the Administrative Data Research Network Northern Ireland (ADRC-NI) and the Northern Ireland Statistics and Research Agency (NISRA) Research Support Unit is acknowledged.The ADRC-NI is funded by the Economic and Research Council (ESRC).The authors alone are responsible for the interpretation of the data and any views or opinions presented are solely those of the author and do not necessarily represent those of the ADRC-NI.The Census and Honest Broker Service/ Business Services Organisation data has been supplied for the sole purpose of this project.The support and guidance of steering committee members representing MindWise, Carers NI, Action Mental Health, the NI Department for Communities, Irish Congress of Trade Unions, NI Equality Commission is also acknowledged.

Table 3 (
self-report) and Table4(psychotropic medication) present results stratified by the major occupational

Table 2 .
Logistic regression models showing the association of major SOC, working hours and household demands with self-reported mental disorder and any 12-month psychotropic medication use.
a Models adjusted for age, marital status, locale of residence, highest level of educational qualification, car ownership, property value and other self-reported chronic conditions.

Table 3 .
Logistic regression models showing the association of family responsibilities (and interaction with sex) with self-reported mental disorder, stratified by major standard occupational classifications.All models are fully adjusted for age, marital status, locale of residence, highest level of educational qualification, car ownership, property value, working hours and other self-reported chronic conditions.

Table 4 .
Logistic regression models showing the association of family responsibilities (and interaction with sex) with receiving any twelve month psychotropic medication prescription, stratified by major standard occupa-All models are fully adjusted for age, marital status, locale of residence, highest level of educational qualification, car ownership, property value, working hours and other self-reported chronic conditions.