Abstract
Keywords
Introduction
Work is vital in promoting recovery for people with mental illness since it provides a social identity, a personal sense of achievement and, most importantly, is associated with reduced psychiatric symptoms and hospital admissions (Rinaldi & Perkins, 2007). However, employment rates especially for people with severe mental illness (SMI) declined over the past decades, e.g., ranging from 4% to 27% in the UK (Perkins & Rinaldi, 2002; O’Brien et al., 2000). Unemployment among people with SMI is a major barrier to a successful integration in the community and significantly interferes with this clientele’s life satisfaction (Mallik, Reeves, & Dellario, 1998; Brown, Cosgrove, & DeSelm, 1997).
Extensive evidence demonstrates that people with SMI want to work and can function productively within competitive jobs, e.g. 53% to 61% of unemployed patients suffering from schizophrenia express an interest in working (Marwaha & Johnson, 2005; Cook & Razzano, 2000; Mueser, Salyers, & Mueser, 2001). Competitive employment is often the most favored long-term goal by mental health service users, followed by education, voluntary work and training (Secker, Grove, & Seebohm, 2001).
If the majority of people with SMI want to work, why do their employment rates remain relatively low? Evidence suggests that individual placement and support (IPS) programs can be effective in helping people with mental illness attain and retain competitive employment (Becker & Drake, 2001; Drake, Becker, & Bond, 2003). However, even with optimal support, rates for competitive employment for people with SMI remain moderate, e.g. for clients with schizophrenia about 60% (Bond, 2004). This rate is attributed to a variety of personal, societal, system and program factors (McQuilken et al., 2003).
The identification of factors outside the Mental Health Vocational Rehabilitation (MHVR) service system that may be relevant in determining the employment success of various demographic and diagnostic subgroups is crucial in informing future effective interventions. Therefore, recent research in vocational rehabilitation focused on identifying potential barriers which hinder people with mental illness from entering or returning to competitive employment (Marwaha & Johnson, 2004; Koletsi et al., 2009). Employers’ attitudes, threats to benefits, lack of skills or work experience and health problems are the barriers most frequently mentioned by mental health service users (Secker & Gelling, 2006).
Research investigating the impact of demographic and clinical factors on employment outcomes for people with mental illness is inconsistent (Wewiorski & Fabian, 2004; Cook & Razzano, 2000). With regards to age, Mechanic, Bilder, and McAlpine (2002) showed that among people with any mental illness in the U.S., employment was most likely in the youngest age group and less likely in older clients. In contrast, for people with no mental illness, employment was least likely for the 18–24 years group and most likely for the 25–44 years group. This suggests that while there is a reduced likelihood of employment with increasing age in the general population, this trend seems to occur at an earlier age and to a greater extent for people with mental illness (Burke-Miller et al., 2006).
Findings about the impact of ethnicity on employment outcomes for people with mental illness are also mixed or even contradictory (Wewiorski & Fabian, 2004). One study reported ethnicity not to be significantly associated with work attainment in France and the UK, while in Germany people born outside of Germany were more likely to work than people born in Germany (Marwaha et al., 2007). Conversely, a meta-analysis has found that across various studies Whites were significantly more likely to enter paid employment than non-Whites (Wewiorski & Fabian, 2004). This perhaps seems more plausible, given that even in the general labor market ethnic minorities face racial disparities and inequality (US Department of Labor, 2004).
Research investigating the impact of clinical diagnosis on employment outcomes for people with mental illness has produced fairly consistent results. The majority of studies show that people with schizophrenia versus another mental illness are less likely to be employed (Wewiorski & Fabian, 2004; Marwaha & Johnson, 2004; Cook & Razzano, 2000).
Our study had two major objectives: First, we wanted to validate and extend the above mentioned barriers to work for people with mental illness. To the best of our knowledge, our study is the first to consider specific differences in self-perceived barriers to work for relevant demographic and clinical subgroups. Second, we aimed to establish associations between age, ethnicity, diagnosis and employment outcomes for people with mental illness and investigate whether self-perceived barriers to work were related to actual employment outcomes.
Methods
Study design
A two-stage cross-sectional study was conducted at Career Management Service (CMS) at South London and Maudsley (SLaM) NHS Foundation Trust. CMS provides vocational support for people with mental illness (mainly schizophrenia, depression, bipolar disorder, anxiety). Clients are unemployed and seek to enter or return to work.
Stage 1: Data previously collected in unstructured interviews with CMS clients yielded a list of barriers to employment which in the first stage of this study, was compared to the scientific literature. Secondly, we investigated whether clinical and demographic subgroups of people with SMI differed in their reports of perceived barriers to work.
Stage 2: Here we examined whether and how demographic (age, ethnicity) and clinical factors (diagnosis) are associated with particular vocational outcomes, i.e., education & training, non-competitive employment and competitive employment. We further tested whether CMS’s users reported barriers to work had an actual impact on employment outcomes.
This investigation was conducted in accordance with the Helsinki declaration. Since the present study was a re-audit of anonymized and archived data of CMS, no ethical approval was needed.
Procedure
Data was obtained from the SLaM Vocational Output Database (VOD), which is managed by CMS. All clients attended at least two sessions of Individual Career Management (ICM) support. The career coach entered the following information for each client respectively: date of entry, birth date, ethnicity, and three main barriers that a client reported to have prevented him or her from entering or returning to employment. Later, information was recorded whether the client had entered education/ training and/or non-competitive (unpaid, voluntary) employment and/or competitive (paid) employment following the provision of ICM at CMS. Participants’ employment status and clinical diagnosis was taken from their internal CMS records.
Sample characteristics
Participants were N = 279 adults who were recruited during a three year period caseload of CMS at SLaM NHS Foundation Trust, based in South London. Average age was 38.08 years, ranging from 18 to 62. All received treatment for mental illness by an adult community mental health service (CMHT) or psychological therapy service (IAPT) at the SLaM NHS Foundation Trust. Initially, 754 clients were recruited. However, for our study N = 279 were included based on three criteria: (1) Diagnosis of a mental and behavioural disorder according to the ICD 10; (2) Unemployment or long-term absence due to sickness; (3) Goal of returning to work. Clients were excluded for the following two criteria: (1) Diagnosis of a physical disability according to the ICD-10; (2) Unspecified diagnosis (missing information).
Measures
Self-perceived barriers to work
The most important out of three barriers mentioned by CMS users were identified. Clients’ reports on perceived barriers to work individual response were classified with labels of pre-identified barriers to work in the academic literature. Accordingly, individual responses were classified into eight different types of barriers to work, namely job search/application process issues, health problems, lack of skills, lack of prior work history, low confidence, work-related stress, social and structural disadvantages and economic resources. A detailed overview of this step is depicted in Appendix 1. Individual responses were then coded regarding to these eight barriers, which in turn were coded as 1 for the barrier of interest (e.g. health problems = 1) and as 0 for the rest of the barriers as a reference group.
Age
First, participants’ age was applied as a continuous variable. Secondly, the sample was split into two age groups using the mean age (M = 38.08) as a midpoint. Clients were divided into a reference group of younger age (18–39) (coded as 0) and a comparison group of older age (39–62 years) (coded as 1).
Diagnosis
Appendix 2 gives a detailed overview as to which types of diagnoses were initially obtained. Because disorders varied amongst schizophrenia, mood and anxiety disorders, substance use, and personality disorders, we collapsed the diagnoses since some of the subgroups contained too little participants. Two groups were formed: clients with a diagnosis of schizophrenia versus those with other kinds of diagnoses. This is consistent to previous practice of similar studies which used the same approach (Marwaha & Johnson, 2004; Cook & Razzano, 2000). A reference group of ‘other diagnoses’ (coded as 0) and a comparison group of clients with a diagnosis of schizophrenia (coded as 1) were established.
Ethnicity
Appendix 2 also depicts the various ethnic groups observed. Owing to the limited numbers, we divided the sample into a reference group of Whites (coded as 0) and a comparison group of ethnic minorities (coded as 1).
Employment outcomes
Employment outcomes included entering either education/training and/or entering non-competitive work and/or entering or returning to competitive work. For each outcome, the reference group of no vocational activity was coded as 0, whereas the comparison group of those with vocational activity was coded as 1.
Statistical analysis
Stage 1
First, frequencies of clients’ reported barriers to work were established. Then, cross tabulations using chi-squared tests were applied to examine differences in reported barriers to work among demographic and clinical subgroups.
Stage 2
Descriptive analyses were conducted to describe the sample’s demographics and to examine clients’ employment outcomes. Then, bivariate associations between socio-demographic and clinical characteristics for employment outcomes were examined, using univariate or unadjusted logistic regression models. Relationships of reported barriers to work and employment outcomes were examined identically. Next, multivariate logistic regression models were established to predict respective employment outcomes using all of the barriers found to be significant in the corresponding unadjusted models in combination with age, ethnicity and diagnosis. Thereby, unadjusted and adjusted associations between age, ethnicity, diagnosis, self-perceived barriers to work and employment outcomes were examined for changes in effect size of the predictor variables. All analyses were conducted using SPSS (SPSS Inc., Chicago, IL) while all statistical tests reported were significant to the 5% error rate.
Results
Stage 1: Reports of self-perceived barriers to work by people with mental illness
The most frequently reported barrier to work circled around ‘job search/ application process issues’ (cf., Table 1 and Appendix 1 for a detailed overview of individual responses). Here, participants (24.4%) mentioned being unsure about training options, being unsure of how to complete an application or of how to write a CV. The second most frequently reported barrier to work (16.5%) was ‘health problems’, which incorporated things like medication side effects or symptoms of the mental illness itself. Then, two types of barriers, namely ‘lack of skills’ and ’prior work history’, were observed. 15.8% of participants indicated that they lacked relevant skills to gain employment such as basic literacy, numeracy or IT skills. On the other hand, 15.8% of CMS users referred to their (lack of) ‘prior work history’ as a barrier to work which included, for example, lack of experience in a chosen area as well as long-term unemployment.
Next, we tested associations between reported barriers to work and participants’ age, diagnosis, and ethnicity. Chi-squared tests indicated that there was no difference between older and younger participants in the barriers they perceived to hinder them from obtaining employment (cf., Table 2). Although we observed that older participants mentioned ‘health problems’ and ‘lack of prior work history’ more frequently as an important barrier to work than their younger counterparts, these differences were not significant. With regards to diagnosis, participants with schizophrenia mentioned ‘lack of prior work history’ significantly more often as a barrier to work than participants with another diagnosis. Moreover, no differences in terms of the barriers mentioned were found between participants of an ethnic minority group and white people (cf., Table 2).
Stage 2: Sample characteristics and employment outcomes
Table 3 reports statistics of the independent variables age, ethnicity, diagnosis and the three different employment outcomes as the dependent variables. Our sample comprised 38% of white people and 62% of an ethnic minority group. Around half of participants (51%) had a serious mental disorder such as schizophrenia, whereas the other half (49%) suffered from a range of common mental health problems (for further details refer to Appendix 2). Age was normally distributed. Participants were divided into older (39–62 years) and younger (18–38 years) participants. Following the provision of ICM at CMS, 52% of participants started an education or training course, 24% went into non-competitive employment (e.g. voluntary work) and 10% went into competitive employment. However, apart from these numbers, some participants engaged in more than one activity, whereas 34% did not engage in any form of vocational activity at all.
Univariate and multivariate logistic regression models for different employment outcomes
Employment outcome ‘Education/ Training’
In the unadjusted models, ethnicity and diagnosis were strongly associated with participating in education/training following ICM (cf., Table 4). People of an ethnic minority group compared to white people (p = 0.006) and those with a diagnosis of schizophrenia as opposed to those with other diagnoses (p = 0.02) were significantly more likely to enter education/training. No difference was observed with regards to age. Only one out of eight barriers to work as reported by CMS users was significantly associated with participating in education/training. Those who mentioned ‘work-related stress’ as a barrier were significantly less likely to enter education/training (p = 0.023).
For the adjusted, multivariate models, we found that participants of an ethnic minority group were significantly more likely to enter education/training than white people (p = 0.02). The association between diagnosis and education/training was not significant in the adjusted model. Accordingly, there was only a trend for participants with a diagnosis of schizophrenia to be more likely to enter education/training (p = 0.11). The relationship of the barrier ‘work stress’ and education/training remained significant (p = 0.04). The logistic model, overall, fit the data to a highly significant level (Deviance = 15.72, d.f.=4, p = 0.003) and explained between 5.5% (Cox & Snell R2) and 7% (Nagelkerke R2) of the variation in education/training.
Employment outcome ‘Non-competitive employment’
In the unadjusted models (cf., Table 4), age, ethnicity and diagnosis were found to be unrelated to the outcome of entering non-competitive employment. In terms of barriers to work, ‘health problems’ (p = 0.06) and ‘social and structural disadvantages’ (p = 0.05) showed substantial, however, insignificant relationships with non-competitive employment.
For the adjusted models, we found that participants who mentioned ‘health problems’ as a barrier to work were significantly less likely to be non-competitively employed (p = 0.05). Participants who noted ‘social and structural disadvantages’ as a barrier to work were significantly more likely to enter non-competitive employment (p = 0.04). While age continued to be unrelated in the fully adjusted model, we found a significant impact of ethnicity on non-competitive employment. Accordingly, people of an ethnic minority group were less likely to be non-competitively employed than white people (p = 0.04). Moreover, a trend was found that people with schizophrenia were more likely to enter unpaid employment than people with a different diagnosis (p = 0.07). This logistic model, overall, fit the data to a significant level (Deviance = 13.68, d.f.=5, p = 0.02) and explained between 5% (Cox & Snell R2) and 7% (Nagelkerke R2) of the variation in non-competitive employment.
Employment outcome ‘Competitive employment’
In the unadjusted models, age was found to be strongly associated with entering or returning to competitive employment (p = 0.02). In contrast, diagnosis and ethnicity were not related to competitive employment. With regards to reported barriers to work, a trend was found for ‘health problems’ to be associated with a lower likelihood of competitive employment (p = 0.09) (cf., Table 4).
After adjusting for the other variables, we found that older participants were significantly less likely to be competitively employed than their younger counterparts (p = 0.02). The association between the barrier ‘health problems’ and competitive employment was fully attenuated. This logistic model, overall, fit the data to a highly significant level (Deviance = 13.88, d.f.=4, p = 0.008) and explained between 5% (Cox & Snell R2) and 10% (Nagelkerke R2) of the variation in competitive employment.
Discussion
Our two-stage study aimed to identify barriers to work for people with mental illness, and investigated demographic and diagnosis-related determinants of vocational outcomes. Our findings contribute to the current evidence base on mental health service users and vocational service in various ways:
First, our participants’ reported major barriers to work were job search/application issues, health problems, lack of skills and lack of prior work history. This is mostly congruent with those identified in previous studies (Secker et al., 2001; Secker & Gelling, 2006). However, in contrast to existing literature our group (given the context of our study) did not report a lack of support in the process of finding work (Secker & Gelling, 2006; Bassett, Lloyd, & Bassett, 2001). Our findings also corroborate that health problems involving the symptoms of mental illness, substance abuse, physical disabilities and medication side effects are a major obstacle to employment (Blank, Harries, & Reynolds, 2011; Johannesen et al., 2007). A lack of skills has consistently been reported to be a key barrier to work for people with mental health problems, however, information on what kind of skills are lacking is sparse (Secker et al., 2001; Secker & Gelling, 2006). Appendix 1 shows that especially IT and literacy skills as well as specific educational qualifications were lacking amongst our population. Particularly, long-term unemployment and lack of work experience were perceived to hinder work attainment (Russinova et al., 2002; Mueser et al., 2001).
Low self-esteem such as fear of disclosing mental illness to employers as well as being anxious about new experiences were also identified (cf., Appendix 1). Because people with mental illness who obtain employment report better self-perceptions of potential barriers to work such as in attitudes towards their mental illness and work itself, respective interventions should place greater emphasis on fostering such self-perceptions since they are integral to the employment process (Cunningham, Wolbert, & Brockmeier, 2000; Johannesen et al., 2007). Furthermore, several obstacles to the act of working itself were mentioned (cf., Appendix 1). This included not being used to having a daily routine, punctuality problems and poor adjustments made at work. Participants also mentioned economic resources as a barrier, including the need to care for others (Secker & Gelling, 2006).
Two unexpected findings deserve further consideration: Stigma and feared loss of benefits did not seem to be perceived as major barriers to work which is not consistent to previous research (Schulze & Angermeyer, 2003; McQuilken et al., 2003). However, the importance of these barriers to work may have faded for our sample since it includes only participants who want to work and believe that employment is an achievable outcome (Boyce et al., 2008).
Secondly, our study aimed to investigate whether there are differences in self-perceived barriers to work for demographic and clinical subgroups of people with mental illness. This is important given that similar studies predominantly investigated white participants (Blank et al., 2011). However, no significant differences in terms of barriers mentioned between white people and those of an ethnic minority group as well as between older and younger participants were found. Notwithstanding, participants with a diagnosis of schizophrenia were significantly more likely to cite lack of prior work history as a barrier to employment. This is not surprising, given the early onset of schizophrenia which often interrupts secondary or tertiary education for young adults due to periods of hospitalizations, relapse and recovery (Bassett et al., 2001). Thus, affected individuals often lack relevant qualifications in order to meet the requirements for a particular job (Kessler et al., 2005). However, the particularly disabling nature of schizophrenia itself which leads to severe disruptions in social and cognitive functioning, might in part limit patients’ capacity to work as well (Rössler et al., 2005).
Third, our findings on the impact of demographic and clinical factors as well as of self-perceived barriers to work on employment outcomes for people with mental illness add upon previous evidence on determinants relevant to attain competitive employment (Butler et al., 2010; Wewiorski & Fabian, 2004). The particular strength of our study is that it also differentiated for the risks to enter non-competitive employment and education/training. This is important since vocational goals often vary among people with mental illness since many deliberately prefer education/training or non-competitive employment (e.g. voluntary work) over competitive work.
Our findings show that younger participants had a greater chance of being competitively employed, thus, confirming previously observed negative effects of older age on vocational outcomes for people with mental illness (Butler et al., 2010; Wewiorski & Fabian, 2004). This implies a need for services with specialized vocational attention to this group (Burke-Miller et al., 2006). It has been shown that middle-aged and older people with schizophrenia can benefit substantially from supported employment programs (Twamley, Jeste, & Lehman, 2003; Twamley et al., 2012).
Fourth, contrary to previous results, we did not find that people with schizophrenia demonstrated less favorable outcomes in competitive employment than people with other diagnoses (Marwaha & Johnson, 2004; Cook & Razzano, 2000). Our findings suggest that diagnosis alone might not be very useful as a predictor of competitive employment (Jones, Perkins, & Born, 2001; Henry et al., 2001). Diagnosis may be more important in determining employment outcomes for people with mental illness who do not present to a vocational rehabilitation service (Michon et al., 2005). Recent studies suggest that negative symptoms as well as neurocognitive functioning rather than the diagnosis itself are predictive of employment status in people with schizophrenia (Tsang et al., 2010; Rosenheck et al., 2014; Razzano, Cook et al., 2005).
With regards to self-perceived barriers to work, we observed a trend that participants who reported health problems were less likely to enter or return to competitive employment. This is in line with Mechanic et al. (2002) who found that better perceived health was related to higher rates of employment among people with serious or any mental illness.
In terms of non-competitive or unpaid employment, people of an ethnic minority group were significantly less likely to enter unpaid employment. Furthermore, participants with schizophrenia tended to be more likely to enter non-competitive employment. However, some people with schizophrenia might prefer part-time or voluntary work over competitive employment because often they assume being incapable to cope with full-time jobs (Marwaha & Johnson, 2005).
Our participants who cited social and structural disadvantages such as being on disability benefits were significantly more likely to enter non-competitive employment (Rosenheck et al., 2014; Becker & Drake, 2001). Moreover, participants who reported health problems were significantly less likely to enter non-competitive employment which supports the argument that poor health (or at least if perceived as poor) hinders any type of vocational activity (Rosenheck et al., 2014).
People of an ethnic minority group were significantly more likely to enter education/training than white people. Since they were also significantly less likely to enter non-competitive employment, this might perhaps reflect a preference of ethnic minority clients to achieve paid employment as the ultimate goal via means of education/training to escape a lack of resources and poverty (Kim, 2002).
Limitations of the study
Several limitations of this study should be acknowledged. First, due to the cross-sectional design no causal inferences can be made. With regards to demographic and clinical factors and employment outcome, the direction of the relationship is nevertheless clear. Second, potential confounding variables such as benefit status, prior work history or educational attainment were not measured (Taylor, 2001; McQuilken et al., 2003; Mechanic et al., 2002). Third, since the whole sample was drawn from one UK-based database, external validity might be limited. Our sample exclusively contained people expressing the wish to work while neglecting those who do not desire to work. Both groups might have different perceptions on barriers to employment, e.g. worries about losing disability benefits (McQuilken et al., 2003). In our study, feared loss of benefits was therefore not a major concern among participants. Samples drawn from vocational rehabilitation services often predominantly include clients with extensive histories of mental illness and associated care which might not be representative of people with mental illness in general (Cook & Burke, 2002). Information on barriers to work came from self-reports which are prone to bias related to under- or over-reporting due to feelings of shame or because participants wanted to justify their unemployment status (Mechanic et al., 2002). Future replications may apply standardized instruments such as the Barriers to Employment and Coping Efficacy Scale (BECES) (Corbière, Laisne, & Mercier, 2000). However, our results were consistent with the academic literature which supports their validity (Secker et al., 2001).
Implications for further research and health service practice
Our findings suggest that clinical and demographic subgroups of people with mental disorders all have similar perceptions of barriers to work. Participants’ health problems interfered with any kind of vocational activity and social and structural disadvantages (e.g. disability benefits) were associated with entering non-competitive employment. Our results on demographic and clinical factors may inform future investigations on the varying impact of age, ethnicity and diagnosis on different types of vocational activities: particularly with regards to older age groups, ethnic minorities and people with schizophrenia.
While diagnosis is associated with employment in general, its impact might be overshadowed by demographic factors which might be relatively more important when it comes to attaining competitive employment. This hypothesis finds support in a study by Rosenheck et al. (2014) which found that people with schizophrenia who were competitively employed did not differ on clinical measures from schizophrenics in non-competitive employment, while being black was found to be negatively related to competitive employment only. Our findings reflect well-known inequities in the general population, thus, implying the need for clinical practice as well as social policies to place greater emphasis on supporting these demographic subgroups of people with mental illness.
Conclusion
In order to enhance the efficacy of supported employment interventions, our study aimed to extend previous findings on barriers to work for people with mental illness. While vocational rehabilitation service users perceived a variety of barriers to work, preliminary support for an actual impact of those on employment outcomes could only be found for health problems and social and structural disadvantages. When it comes to competitive employment as opposed to other types of employment, demographic factors seem to have a bigger impact than diagnosis. This shows that vocational rehabilitation for people with mental illness is influenced by factors beyond clinical impairment which generally affect the labor market (Cook & Burke, 2002). Therefore, our findings have implications for clinical practice as well as for those involved in vocational rehabilitation research as factors outside of supported employment programs might interfere with a successful employment outcome.
Conflict of interest
The authors declare that they have no conflict of interest.
Footnotes
Appendices
Demographic and clinical characteristics of N = 279 CMS users
| Original sample characteristics | N | % |
| Ethnicity | ||
| |
|
|
| |
|
|
| Mixed | 11 | 3.9 |
| Asian or Asian British | 13 | 4.6 |
| Black or Black British | 133 | 47.7 |
| Others | 15 | 5.4 |
| Diagnosis | ||
| |
|
|
| |
|
|
| Affective Disorder | 50 | 17.9 |
| Bipolar Disorder | 26 | 9.3 |
| Anxiety Disorder | 12 | 4.3 |
| Substance Misuse Disorder | 10 | 3.6 |
| Personality Disorder | 10 | 3.6 |
| PTSD | 12 | 4.3 |
| Others | 17 | 6.0 |
Acknowledgments
We would like to thank Gillian Wiscarson for critically reviewing the manuscript for intellectual content as well as for expert consultations on the subject and methods.