key: cord-0843823-xhfoeo88 authors: Probyn, Katrin; Engedahl, Martin Stav; Rajendran, Dévan; Pincus, Tamar; Naeem, Khadija; Mistry, Dipesh; Underwood, Martin; Froud, Robert title: The effects of supported employment interventions in populations of people with conditions other than severe mental health: a systematic review date: 2021-12-09 journal: Prim Health Care Res Dev DOI: 10.1017/s1463423621000827 sha: dad364a9347b354b02127341e6138cf105b960ba doc_id: 843823 cord_uid: xhfoeo88 AIM: To assess the effectiveness of supported employment interventions for improving competitive employment in populations of people with conditions other than only severe mental illness. BACKGROUND: Supported employment interventions have been extensively tested in severe mental illness populations. These approaches may be beneficial outside of these populations. METHODS: We searched PubMed, Embase, CINAHL, PsycInfo, Web of Science, Scopus, JSTOR, PEDro, OTSeeker, and NIOSHTIC for trials including unemployed people with any condition and including severe mental illness if combined with other co-morbidities or other specific circumstances (e.g., homelessness). We excluded trials where inclusion was based on severe mental illness alone. Two reviewers independently assessed risk of bias (RoB v2.0) and four reviewers extracted data. We assessed rates of competitive employment as compared to traditional vocational rehabilitation or waiting list/services as usual. FINDINGS: Ten randomised controlled trials (913 participants) were included. Supported employment was more effective than control interventions for improving competitive employment in seven trials: in people with affective disorders [risk ratio (RR) 10.61 (1.49, 75.38)]; mental disorders and justice involvement [RR 4.44 (1.36,14.46)]; veterans with posttraumatic stress disorder (PTSD) [RR 2.73 (1.64, 4.54)]; formerly incarcerated veterans [RR 2.17 (1.09, 4.33)]; people receiving methadone treatment [RR 11.5 (1.62, 81.8)]; veterans with spinal cord injury at 12 months [RR 2.46 (1.16, 5.22)] and at 24 months [RR 2.81 (1.98, 7.37)]; and young people not in employment, education, or training [RR 5.90 (1.91–18.19)]. Three trials did not show significant benefits from supported employment: populations of workers with musculoskeletal injuries [RR 1.38 (1.00, 1.89)]; substance abuse [RR 1.85 (0.65, 5.41)]; and formerly homeless people with mental illness [RR 1.55 (0.76, 3.15)]. Supported employment interventions may be beneficial to people from more diverse populations than those with severe mental illness alone. Defining competitive employment and increasing (and standardising) measurement of non-vocational outcomes may help to improve research in this area. Aim: To assess the effectiveness of supported employment interventions for improving competitive employment in populations of people with conditions other than only severe mental illness. Background: Supported employment interventions have been extensively tested in severe mental illness populations. These approaches may be beneficial outside of these populations. Methods: We searched PubMed, Embase, CINAHL, PsycInfo, Web of Science, Scopus, JSTOR, PEDro, OTSeeker, and NIOSHTIC for trials including unemployed people with any condition and including severe mental illness if combined with other co-morbidities or other specific circumstances (e.g., homelessness). We excluded trials where inclusion was based on severe mental illness alone. Two reviewers independently assessed risk of bias (RoB v2.0) and four reviewers extracted data. We assessed rates of competitive employment as compared to traditional vocational rehabilitation or waiting list/services as usual. Findings: Ten randomised controlled trials (913 participants) were included. Supported employment was more effective than control interventions for improving competitive employment in seven trials: in people with affective disorders [risk ratio (RR) 10. 61 (1.49, 75.38) ]; mental disorders and justice involvement [RR 4.44 (1.36, 14.46) ]; veterans with posttraumatic stress disorder (PTSD) [RR 2.73 (1.64, 4.54) ]; formerly incarcerated veterans [RR 2.17 (1.09, 4.33) ]; people receiving methadone treatment [RR 11.5 (1.62, 81.8)] ; veterans with spinal cord injury at 12 months [RR 2.46 (1.16, 5.22) ] and at 24 months [RR 2.81 (1.98, 7.37) ]; and young people not in employment, education, or training ]. Three trials did not show significant benefits from supported employment: populations of workers with musculoskeletal injuries [RR 1.38 (1.00, 1.89)]; substance abuse [RR 1.85 (0.65, 5.41) ]; and formerly homeless people with mental illness [RR 1.55 (0.76, 3.15) ]. Supported employment interventions may be beneficial to people from more diverse populations than those with severe mental illness alone. Defining competitive employment and increasing (and standardising) measurement of nonvocational outcomes may help to improve research in this area. Conventional traditional vocational rehabilitation (TVR) tends to follow a train-and-place model; often involving pre-employment training, testing, or counselling to prepare individuals for employment, and often involving sheltered employment (Bond et al., 2012) . Supported employment interventions emphasise a place-and-train approach and placement of individuals (who have health problems but would like to work) in real-world work settings, then providing the support that is needed (Bond, 2004) . Individual placement and support (IPS) is a standardised model of supported employment that has been developed for people with severe mental illness, which is defined as schizophrenia or schizophrenia-like disorder, bipolar disorder, or major depression with psychotic features (Drake et al., 1999; Crowther et al., 2001; Kinoshita et al., 2013) . IPS was developed in the USA in the 1990s. It is based on eight principles and aims to secure competitive employment. The focus of the intervention is on providing preference-based supported employment for those who want to work, integration of services, the provision of benefits counselling, a rapid job search, and 'time-unlimited' support (Drake et al., 1999; Drake, 2012b; Becker et al., 2019) . Supported employment more generally operates around the central concept that people who want to work can be placed in a job they want and then receive appropriate support. IPS and supported employment in populations involving severe mental illness alone have been extensively evaluated (Kinoshita et al., 2013; Crowther et al., 2001; Modini et al., 2016; Marshall et al., 2014; Suijkerbuijk et al., 2017) . This individualised approach is gaining support and beginning to be implemented in new populations other than those with severe mental health conditions (Drake, 2012a; 2012b) . However, it is not clear how effective such interventions are in these populations. Our aims were to assess the effectiveness of any type of supported employment intervention (including IPS) for improving rates of competitive employment in populations of people for whom severe mental illness is not their problem or not their only problem; to describe and report secondary outcomes measured; to summarise definitions of competitive employment used; and to extract summary details of tested interventions. We prospectively registered the systematic review with the International Prospective Register of Systematic Reviews: PROSPERO (NIHR, 2017) and followed the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines for reporting of systematic reviews (Moher, 2009 ). The review was part of a larger study (Return to work with Individualised Supported Employment -RISE), which was funded by Versus Arthritis, and was designed to determine the feasibility of delivering an individualised supported work placement intervention to people with chronic pain. The early phases of the present review informed the intervention development process and the choice of outcome measures used in the RISE intervention (Froud et al., 2020) . We searched for peer-reviewed randomised controlled trial (RCT) reports in PubMed, Embase, CINAHL, PsycInfo, Web of Science, Scopus, JSTOR, PEDro, OTSeeker, and NIOSHTIC from database inception to March, 2020. We based our search on MeSH indexing terms and free text terms relating to supported employment intervention. An example search string is included in Appendix 1. We supplemented our search with backwards citation tracking. Only trials published in English were included. We excluded grey literature and conference proceedings. Papers reporting RCTs were eligible for inclusion if they met the following criteria: (1) They assessed effectiveness of a supported employment intervention compared to TVR waiting list/services as usual; (2) The intervention aim was to obtain and maintain competitive employment; (3) Study participants were unemployed at the beginning of the trial; and (4) Rates of obtaining competitive employment were measured and reported as an outcome. Papers describing trials of supported employment intervention for participants with any problem other than severe mental health illness alone were included. Pilot studies were included if they reported smaller scale RCTs and included an objective of estimating effectiveness parameters. Conversely, feasibility studies that did not conduct a small-scale RCT and estimate effectiveness were excluded. For those with severe mental illness, we included RCTs where an additional co-morbidity or a materially different population was involved compared to severe mental illness alone; for example, severe mental illness and homelessness. Studies were excluded if they were not RCTs, not published in a peer-reviewed journal, or the target population did not comprise adults of working age. The search results were managed using EPPI reviewer 4 software (EPPI-Centre at the Social Science Research Unit of the UCL Institute of Education, University of London, London, UK). Two of four reviewers (KP, KN, MSE, or DR) independently screened all records by title and abstract. Disagreements were resolved through discussion and, if necessary, with a third reviewer acting as an arbitrator. Articles that could not be excluded from titles and abstracts sifting were retrieved as full text and assessed independently against inclusion criteria. Two reviewers independently extracted data. For each study, data on the country of the trial, intervention and control arm, structure and delivery of interventions, fidelity to any relevant instrument [e.g., an edition of the Supported Employment Fidelity Manual (Becker et al., 2008) ], competitive employment rates, definitions of competitive employment used, secondary vocational and non-vocational outcomes, and length of follow-up time were abstracted. Two reviewers (MSE, DR) independently assessed the risk of bias using the Cochrane Risk of Bias Tool v2.0 (Sterne et al., 2019) . Disagreements were resolved through discussion with a third acting as arbitrator if necessary. Our a priori primary outcome (PROSPERO 2017:CRD42017067586) for all comparisons of effectiveness was 'obtaining competitive employment'. We extracted numbers returning to work in intervention and control groups. We abstracted author-reported risk ratios (RRs) and 95% confidence intervals, or calculated these from reported rates in cases where RRs were not reported by authors. We planned a meta-synthesis, and a meta-regression of study characteristics on effects, in the case of sufficient homogeneityeither overall, or within strata of sub-populations. In the case of high population heterogeneity and the absence of any such strata, we did not meta-analyse and presented obtaining competitive employment RRs from studies, making a forest plot for illustrative purposes only, and without estimating any pooled effects. We note that random effects models assume that underlying effects follow a normal distribution (Higgins et al., 2021) . This assumption is not credible in case included studies span very different populations. P-values for between-group differences were extracted as reported by original authors, and authors' descriptive statistics were used to summarise secondary outcomes of included studies. Assessment of publication bias was determined through visual inspection of funnel plots. All analyses were done using Stata Version 15.1 (IBM, Washington). We included 13 articles describing 10 trials. Figure 1 shows a flow chart of the search process and included articles (Figure 1 , Flow chart). The 10 included trials (913 participants) used supported employment interventions across different populations (Table 1 ) (Bejerholm et al., 2017; Bond et al., 2015; Davis et al., 2012; LePage et al., 2016; Li-Tsang et al., 2008; Magura et al., 2007; Katrin Probyn et al. ; Sveinsdottir et al., 2020) . Each compared to some form of TVR/services as usual, as per inclusion criteria. One paper reporting follow-up data for one of the included trials and two available protocols for included trials were also included in our review material (Ottomanelli et al., 2014; Ottomanelli et al., 2009; Sveinsdottir et al., 2016) . Table 1 shows characteristics of the included trials. We judged that six trials used IPS as a supported employment intervention and authors of these trials reported assessing fidelity against a scale (Appendix Table 1 ) (Becker et al., 2001; Becker et al., 2015; Becker et al., 2008) . In four trials, we judged that other supported employment interventions were tested (Appendix Table 1 ). Two trials included participants with physical disabilities: one studied supported employment for veterans with spinal cord injury (Ottomanelli et al., 2009; Ottomanelli et al., 2012; Ottomanelli et al., 2014) and one for workers with musculoskeletal injuries (Li-Tsang et al., 2008) . Two trials included substance-misuse (methadone) patients (Magura et al., 2007; Lones et al., 2017) . In inclusion criteria, Li-Tsang et al. describe that participants must have been injured or on sick-leave for at least 6 months. However, participants are referred to as being unemployed in the Discussion section, and it is clear from Li-Tsang's description of methods that participants are being helped to find new jobs. The study was based in Hong Kong, where under the Employment Ordinance employees can only accrue employer-paid sickness for up to 120 days, for example <6 months (HK Labour Department, 2021). Thus, we reasoned that participants cannot have been in paid employment at the point of recruitment and judged that the trial met the inclusion criteria of this review. Lones et al. is a pilot study that featured a small-scale RCT and included as an objective in the estimation of effectiveness of the intervention on employment (Lones et al., 2017) . Bejerholm et al. (2017) included people with less debilitating mental illness and participants with affective disorders. Bond included people with severe mental illness alongside justice involvement . Two trials included veterans: one veterans with PTSD; (Davis et al., 2012) and one formerly incarcerated veterans with substance-use disorder and/or mental illness (LePage et al., 2016). Magura et al. (2007) included substance-misuse methadone patients. Poremski et al. (2015) included people with mental illness and who were previously homeless but had been recently housed. Finally, Sveinsdottir et al. (2020) included young adults with various social or health-related problems who were at risk of work disability. The reported caseloads across all included studies, per employment specialist, ranged from 15 to 35. We judged five studies (5/10; 50%) to have an overall 'low risk' of bias, which reflected the scores of 'low risk' across all five domains. Three (3/10; 30%) were judged to have an overall 'high risk' of bias, which reflected judgements of high risk in at least one of the five domains. Two (2/10; 20%) were judged to have 'some concerns', reflecting judgements of some concerns in at least one of the five domains in the absence of judgements of high risk in these domains ( Figure 2 ). All of the included trials assessed obtaining competitive employment as a primary outcome. The phrase 'competitive employment' was either elaborated on or more explicitly defined in eight of the trials (Appendix Table 2 ). Three of the trials gave details of how long a job must be held to qualify as meeting the criterion of competitive employment: two stated for a minimum of one day, and one for a period of four weeks for a minimum of 18 hours/week. Figure 3 , which is a forest plot for illustrative purposes only, shows that in 8 of the 11 (80%) included trial result rows (n.b. where one row is a two-year follow-up of the same trial), supported employment was more effective than control in returning people to work. Supported employment was more effective than TVR in people with affective disorders, where the outcome was competitive employment at 12 months [RR 10.61 (95% CI 1.49 to 75.38)] (Bejerholm et al., 2017) ; in people with severe mental illness and justice involvement where the outcome was competitive employment at 12 months (RR 4.44 (1.36 to 14.46)) ; in veterans with posttraumatic stress disorder (PTSD) where the outcome was competitive employment within 12 months [RR (Ottomanelli et al., 2012; Ottomanelli et al., 2014) ; and in young people not in employment, education, or training, where the outcome was any (expressed as weeks, days, or hours) competitive employment (minimum of 1 day) at 12 months [RR 5.90 (1.91 to 18.19) ] (Sveinsdottir et al., 2020) . Supported employment interventions were not shown to be more effective than control in returning people to work in three trials: one trial including workers with musculoskeletal injuries, where the outcome was continuous (Magura et al., 2007) ; and one trial including homeless people with mental illness who were recently housed, where the outcome was competitive employment of at least one day within a 30-day period [RR 1.55 (0.76 to 3.15)] (Poremski et al., 2015) . As heterogeneity was high across the different included populations, and the assumption of an underlying normal distribution of treatment effects not plausible, we did not do a metaanalysis. We note that the two papers by Ottomanelli, reporting 12 and 24 month results on the same participants, are both included in Figures 3 and 4 (since we did not meta-analyse). Figure 4 shows the comparison of percentages of participants returned to work in supported employment versus their corresponding control groups. (2020) Poremski (2015) Ottomanelli (2014) Ottomanelli (2012) Magura (2007) Lones (2017) Li−Tsang (2008) Le Page (2016) Davis (2012) Bond (2015) Bejerholm (2017 Table 1 for population details). † The authors state RTW outcomes were collected three weeks after the programme and note the four week 18-hour outcome is defined by the Employment Ordinance of Hong Kong. We presume therefore that employment could be started during the programme so the outcome could be achieved. Katrin Probyn et al. . Two studies assessed mean number of weeks worked during study period, reporting these as significantly favouring supported employment (affective disorders and veterans with PTSD). Two studies assessed the mean days worked during study period with both reporting a significant effect favouring supported employment (people with severe mental illness and justice involvement; and veterans with PTSD). Finally, one study reported obtaining any paid employment and any informal paid employment as an outcome in addition to obtaining competitive employment and reported significant effects in both cases for those receiving supported employment. Table 3 summarises non-vocational outcomes reported in the included studies. Again, not every domain in the table was reported in each study. Satisfaction with employment services was significantly greater than control for those receiving supported employment in a population of homeless people with mental illness. Quality of life was significantly improved over control for those receiving supported employment in workers with musculoskeletal injuries, and stress and anxiety was significantly lower in the same population group. Visual inspection of a funnel plot (not shown) may have suggested some asymmetry within an indication of publication bias in smaller trials; however, too few data points existed to draw any firm conclusion. Given we did not meta-analyse, no sensitivity analyses were done. Ten trials examined the effectiveness of supported employment interventions across populations of people with conditions or circumstances other than serious mental illness alone. Three of the trials were judged to have high RoB. Results suggest that supported employment may be effective outside populations defined by only severe mental illness. Supported employment interventions were more effective than control in returning people to work in trials including people with affective disorders; people with severe mental illness and justice involvement; veterans with PTSD; formerly incarcerated veterans; veterans with spinal cord injury; and in young people not in employment, education, or training. There was an evidence from a trial that supported employment interventions were effective in people receiving methadone treatment (high RoB); although in another trial there was no evidence that ITT, intention to treat population; ND, no data; PTSD, posttraumatic stress disorder; SMI, severe mental illness. *P-value as reported verbatim as per study reports (e.g., NS/P