Introduction
Good employment is intrinsic to recovery from mental health conditions, improving quality of life, and empowering people to live independently [Reference Waddell and Burton1]. Lost employment represents the majority of costs of mental health conditions to society [Reference Christensen, Lim, Saha, Plana-Ripoll, Cannon and Presley2]. Supported employment (SE) programmes help people to enter or return to employment. Programmes vary but are characterised by a “place and train” approach with a focus on competitive employment as the goal with ongoing support to retain employment [Reference Fogelgren, Ornstein, Rödin and Thoursie3]. The individual placement and support (IPS) approach is one form of SE, underpinned by eight evidence-based principles, including employment specialists integrated into mental health services, playing an intermediate role to match jobseekers preferences with employers in a competitive job market, and initiating rapid search for competitive employment [Reference Bond4, Reference Dreher, Bond and Becker5]. Systematic reviews indicate SE interventions, particularly IPS, are highly effective in helping people with mental health conditions into work [Reference Brinchmann, Widding-Havneraas, Modini, Rinaldi, Moe and McDaid6, Reference Modini, Tan, Brinchmann, Wang, Killackey and Glozier7]. Despite this, policy decisions and implementation of evidenced-based nonpharmacological structured interventions can be slow (at best). This is the case for SE/IPS.
The economic case for SE/IPS is potentially strong as it has merit to reduce long-term disability related to illness and may catalyse policy decisions and implementation. Reviews have focused on the effectiveness of SE/IPS, rather than the economic case; only one systematic review of economic evaluations was identified and restricted to trial-based evaluations of IPS for people with severe mental illness (SMI) [Reference Zheng, Stern, Wafford and Kohli-Lynch8]. It identified seven studies published between 1998 and 2017, generally favouring IPS. Another review of supported employment for people with disabilities reported 6 economic evaluations, again all focused on IPS, the most recent of which was published in 2014 [Reference Nøkleby, Blaasvær and Berg9]. A narrative review looking at the social costs of expanding access to SE/IPS identified 27 studies on the costs of interventions, but only included 5 economic evaluations [Reference Salkever10]. Consequently, this review aims to comprehensively review evidence on economic evaluations of SE/IPS programmes.
Methods
A systematic review was registered with PROSPERO (CRD42020184359) to identify economic evaluations of SE/IPS programmes using a “place and train” approach to support working-age individuals to find and stay in employment in the competitive labour market. Traditional vocational rehabilitation and sheltered work programmes, as well as education, training, and initiatives to promote return to work after sick leave of less than 1 year were excluded.
Although the PROSPERO review covers people with all health conditions/disabilities, this article focuses solely on interventions for people with any mental health condition, regardless of severity, including learning disabilities but excluding dementia. We searched PubMed/MEDLINE, EMBASE, PsycINFO, CINAHL, Business Source Complete, IBSS, and EconLit. Because of the volume of records found, and examination of previous reviews which identified only one economic evaluation between 2000 and 2009 [Reference Dixon, Hoch, Clark, Bebout, Drake and McHugo11], one deviation from the protocol was to search for literature published between January 2009, rather than January 2000 to August 2021 (see search strategies in Supplementary S1). The multicountry EQOLISE study on supported employment had just been published in 2008 [Reference Burns and Catty12]; this had an economic evaluation fully embedded that we felt would act as a catalyst for later studies. Another deviation was that we were also unable to search two social science databases, PAIS International and ASSIA as their subscriptions lapsed. Reference lists of relevant papers were checked and Google Scholar searched. There were no language restrictions. Title/abstracts and full texts of papers were independently screened by two reviewers. Disagreements were resolved through discussion and, if necessary, input from a third reviewer.
The economic outcomes in identified studies were likely to be very heterogenous and in the protocol, we anticipated finding studies with many different outcomes, including cost-effectiveness studies that report incremental costs per additional unit of outcome (employment) achieved between two or more interventions, such as cost per job, or day of work. In this case, the amount policymakers are willing to pay for better outcomes is a value judgment varying across countries.
However, these cost-effectiveness studies are of limited value as their relative cost-effectiveness cannot be compared easily with other health-related interventions, e.g. investment in support for carers of people living with dementia. To overcome this, health economists often measure cost per quality-adjusted life year (QALY) gained, where a year spent in perfect quality of life represents 1 QALY. Quality of Life can be elicited in different ways, for instance asking study respondents to complete the EQ-5D [Reference Devlin and Brooks13], an instrument often used in health economic studies. A monetary value can also be placed on a QALY, but this varies across countries, reflecting differences in societal willingness to pay for a QALY. Other anticipated measures included cost-benefit analyses where both costs and outcomes are valued monetarily to generate net monetary benefits (NMBs), and return on investment (ROI) analyses where the cost of investing in SE/IPS is compared to the monetary value of additional employment and costs averted.
Extracted information included country, population, intervention, study design, economic evaluation type, timeframe, perspectives, key impacts, costs related to mental health and work, and summary economic findings. We used the Consolidate Health Economic Evaluation Reporting Standards (CHEERS) checklist to judge the quality of economic evaluation methodologies [Reference Husereau, Drummond, Petrou, Carswell, Moher and Greenberg14]. Recent new CICERO appraisal guidance on economic reviews highlights this as one recommended tool [Reference Mandrik, Severens, Bardach, Ghabri, Hamel and Mathes15]. We added one additional item to the existing checklist on whether labour market outcomes, as well as health outcomes, were reported. The checklist has 23 or 25 items depending on whether it was a single study or model-based economic evaluation; we allocated one point per item with scores as a percentage indicating strength of evidence. All costs in the text have been converted to purchasing power parity adjusted 2020 US Dollars, using the CCEMG – EPPI-Centre Cost Converter [16]. We report original currency values in tables.
Results
Figure 1 shows a PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) diagram with review results. From 40,015 references, 56 papers covering 54 economic studies were identified (see Supplementary S2 for a list of all studies). A total of 43 studies (79%) reported a positive economic case for SE/IPS with three (5%) being negative and eight (15%) inconclusive. Studies are very heterogenous; 13 used multiple evaluation methods. A total of 23 included a cost–benefit analysis, 11 cost per QALY gained, 5 ROI and 25 cost per employment outcome achieved, measured as job gained, hour, day, or week worked. Five were cost-consequence studies where multiple outcomes and costs are recorded but no synthesis is conducted. Overall, 28 studies examined the economic case for IPS, four IPS augmented by another intervention, and 24 other forms of SE. Studies ranged from under 10 to 173,000 participants, with time horizons from 6 weeks to 50 years. Other than one economic evaluation of a six-country trial in Europe [Reference Knapp, Patel, Curran, Latimer, Catty and Becker17], all were single-country studies. In total, 17 (31%) were set in the UK, 18 (33%) in the USA, and 8 (15%) in the Nordic countries.
Figure 1. PRISMA flow diagram.
Quality was variable; the mean quality score was 63% (range 30–96%) (See quality checks in Supplementary Table S1). We focus here on 41 studies with quality scores over 50%, including 24 with scores over 70% (Table 1). Ten studies in Table 1 reported cost per QALY gained, (8 favourable to SE/IPS), 16 NMBs (14 positive), 5 ROI (4 positive), and 20 cost per employment outcome (14 favourable, 5 inconclusive, 1 negative). Sixteen of the 24 higher-quality evaluations were linked to empirical studies, mainly randomised controlled trials (RCTs). The remainder synthesised information from multiple sources on costs and benefits of SE/IPS.
Table 1. Summary results for studies with quality scores of 50% and above.
Abbreviations: CCA, cost consequences analysis; CEA, cost-effectiveness analysis; Govt, government; HC, health care sector; LGovt, local government; MANSA, Manchester short assessment of quality of life; NMB, net monetary benefits; NS, not stated; OTH, other; QALYs, quality-adjusted life years; QLS, quality of life scale; RCT, randomised controlled trial; SOC, society; SW, sheltered workshop; WF, welfare sector; WTP, willingness to pay.
a Only significant differences for sectors reported.
Individual Placement and Support
Of the 28 IPS studies focused on individuals with SMI (Supplementary Table S2), 23 had quality scores above 50% (Table 2). Of these, 6 studies reported cost per QALY gained, (5 favourable to SE/IPS), 8 NMBs (all positive), 4 ROI (3 positive), and 13 cost per employment outcome (9 favourable, 3 inconclusive, 1 negative). In total, 18 studies included impacts on health outcomes and/or service utilisation, as well employment [Reference Knapp, Patel, Curran, Latimer, Catty and Becker17–Reference Hoffmann, Jäckel, Glauser, Mueser and Kupper22, Reference Howard, Heslin, Leese, McCrone, Rice and Jarrett24–Reference Mavranezouli, Megnin-Viggars, Cheema, Howlin, Baron-Cohen and Pilling26, Reference Rosenheck, Leslie, Sint, Lin, Robinson and Schooler30–Reference Stroupe, Jordan, Richman, Bond, Pogoda and Cao36, Reference van Stolk, Hofman, Hafner and Janta38, 39]. Table 2 shows 16 of these 23 studies had a short timeframe (maximum 2 years) [Reference Knapp, Patel, Curran, Latimer, Catty and Becker17–Reference Heslin, Howard, Leese, McCrone, Rice and Jarrett21, Reference Howard, Heslin, Leese, McCrone, Rice and Jarrett24, Reference Khalifa, Talbot, Barber, Schneider, Bird and Attfield25, 27–Reference Shi33, Reference Stroupe, Jordan, Richman, Bond, Pogoda and Cao36, Reference van Stolk, Hofman, Hafner and Janta38], but there is growing evidence on longer-term economic impacts [Reference Hoffmann, Jäckel, Glauser, Mueser and Kupper22, Reference Holmås, Monstad and Reme23, Reference Mavranezouli, Megnin-Viggars, Cheema, Howlin, Baron-Cohen and Pilling26, Reference Stant, Van Busschbach, Van Vugt and Michon34, Reference van Busschbach, Michon, van Vugt, Stant, MmvLC and van Erp35, Reference Szplit37, 39, Reference Whitworth40]. For example, one Norwegian RCT compared IPS to traditional vocational rehabilitation for 327 people with moderate to severe mental illness [Reference Holmås, Monstad and Reme23]. Using registry data on employment and earnings it followed individuals for 43 months, reporting sustained increased levels of employment, equivalent to two full-time months per participant. Health service utilisation was lower although this was not significant. The cost–benefit analysis was limited in detail but reported net-benefits of $27,670 per participant. Cost-savings were due to increased time in work and reduced use of traditional active labor market programmes.
Table 2. Detailed data extraction for IPS economic studies (quality scores above 50% only).
Some modelling analyses, synthesising evidence from multiple sources, have explored longer-term impacts. The Washington State Institute for Public Policy has generated economic models for many mental health interventions, including a 50-year model for IPS [39], reporting a benefit-to-cost ratio for IPS of 7.7:1.
A UK modelling study [Reference Whitworth40] examined potential costs and monetary benefits for up to 10 years, by varying assumptions about the degree of “modifiable” IPS fidelity items using two approaches, a “discrete” approach within secondary mental health services versus a “networked” approach, partnering with wider health, housing, debt, and employment services. Limiting IPS to 15 months maximum, caseloads per employment specialist varied between 20 and 30, with 12 scenarios showing a return on investment between 0.19 and 4.53 after 5 years and 0.32–7.47 after 10 years. By adopting a broader perspective in economic analyses, cost–benefits were greater and the longer the time horizon, the higher the expected returns.
Six IPS studies included QALY outcomes and the results were mainly positive. For example, the potential costs of IPS for people with autism were modelled over 8 years in the UK [Reference Mavranezouli, Megnin-Viggars, Cheema, Howlin, Baron-Cohen and Pilling26], with a cost per QALY of $9,231, well below the $28,964 (£20,000) to $43,042 (£30,000) threshold recommended by the National Institute for Health and Care Excellence, the body responsible for producing guidance on cost-effective interventions in England [57]. In a Danish RCT [Reference Christensen, Kruse, Hellstrom and Eplov18], IPS for people with SMI was less costly, with better quality of life measured using the EQ-5D. A Swedish RCT for affective disorders [Reference Saha, Bejerholm, Gerdtham and Jarl31], found IPS less costly and more effective using the MANSA instrument. A Canadian RCT found positive impacts on quality of life measured with the EQ-5D, but data were only available for 30% of participants [Reference Sambo32]. However, another Danish study using IPS for mood and anxiety disorders reported neither a significant change in quality of life nor in costs and was not considered cost-effective [Reference Hellström, Kruse, Christensen, Trap Wolf and Eplov20].
Nine economic evaluations alongside RCTs showed IPS was less costly as well as more effective than usual support, including traditional vocational rehabilitation. [Reference Knapp, Patel, Curran, Latimer, Catty and Becker17, Reference Christensen, Kruse, Hellstrom and Eplov18, Reference Heslin, Howard, Leese, McCrone, Rice and Jarrett21–Reference Khalifa, Talbot, Barber, Schneider, Bird and Attfield25, Reference Saha, Bejerholm, Gerdtham and Jarl31, Reference Shi33]. For example, a multicentre trial across six European cities reported IPS had lower mean health care costs at first 6-month follow up than controls, $9,414 versus $13,909 (Mean difference $5,462, 95% CI −$9,286, −$1,682) [Reference Knapp, Patel, Curran, Latimer, Catty and Becker17]. In other words, IPS can be considered as a dominant strategy, compared with usual care.
A 5-year trial in Switzerland reported a 132% return on investment from IPS, taking increased earnings and lower health costs for participants into account [Reference Hoffmann, Jäckel, Glauser, Mueser and Kupper22]. In addition, the 2-year Supported Work and Needs (SWAN) RCT in the UK [Reference Howard, Heslin, Leese, McCrone, Rice and Jarrett24] showed total costs for usual care were significantly higher by $4040, compared to IPS, given no significant difference in outcomes. In other words, IPS had significantly lower costs than treatment as usual with similar effects. This can also be interpreted as a good investment.
Only two IPS studies reported significantly higher health care costs for IPS participants. In the Netherlands, while health care costs were higher, the authors argued IPS could still be cost-effective given significantly improved employment outcomes [Reference van Busschbach, Michon, van Vugt, Stant, MmvLC and van Erp35]. There was also significantly increased use of outpatient mental health services for people with PTSD using IPS in the US over 18 months [Reference Stroupe, Jordan, Richman, Bond, Pogoda and Cao36], but improved work outcomes offset these increased costs.
IPS Plus Psychological Therapies
Meta-analyses have shown additional benefits of adding Cognitive Remediation (CR) and Cognitive Behavioural Therapy to vocational interventions [Reference Chan, Hirai and Tsoi58]. Four studies looked at IPS augmented with these interventions, all with quality scores above 50% (Table 3). Two reported cost per QALY gained, (both favourable to SE/IPS), 1 positive NMBs, and 3 cost per employment outcomes (2 favourable, 1 inconclusive). In three RCTs studies, CR in Denmark [Reference Christensen, Kruse, Hellstrom and Eplov18] and Japan [Reference Yamaguchi, Sato, Horio, Yoshida, Shimodaira and Taneda43] and/or CBT [Reference Reme, Grasdal, Løvvik, Lie and Øverland41] in Norway were added to IPS. Impacts on health and social care service use were reported in all studies. In Japan adding psychological therapies to IPS was dominant with lower health care service costs and better employment rates and cognitive functioning within 1 year [Reference Yamaguchi, Sato, Horio, Yoshida, Shimodaira and Taneda43]. The Danish study reported mental health care costs in the IPS plus CR group of $18,950 versus $25,205 (p = 0.0426) in the usual care group over 18 months, with better employment outcomes and a cost per QALY gained of $46,817 [Reference Christensen, Kruse, Hellstrom and Eplov18]. This would be considered cost-effective in Denmark, with an accepted societal willingness-to-pay threshold of $48,261 [Reference Christensen, Kruse, Hellstrom and Eplov18]. The Norwegian study also had lower health service costs, better quality of life, and positive net monetary benefits for participants on long-term disability benefits [Reference Reme, Grasdal, Løvvik, Lie and Øverland41]. However, a small UK feasibility RCT [Reference Schneider, Akhtar, Boycott, Guo, Latimer and Cao42] of IPS plus work-focused counseling CBT compared to IPS alone, was more costly and unlikely to be cost-effective.
Table 3. Detailed data extraction for IPS+ psychological therapies economic studies.
Other Supported Employment Interventions
Of the 24 non-IPS SE studies (Supplementary Table S3) 14 had quality scores above 50% (Table 4), including 8 whose populations included some people with learning disabilities [Reference Fogelgren, Ornstein, Rödin and Thoursie3, Reference Cimera44, Reference Cimera, Wehman, West and Burgess45, Reference Cimera47–Reference Cimera49, 53, Reference Tholén, Hultkrantz and Persson56]. Two studies in Table 4 reported cost per QALY gained, (one favourable), seven NMBs (five positive), one positive ROI, and four cost per employment outcome (three favourable and one inconclusive).
Table 4. Detailed data extraction for supported employment economic studies (quality scores above 50% only).
Only five SE studies looked at health outcomes and/or service utilisation [Reference Dattilo50, Reference Evensen, Wisloff, Lystad, Bull, Martinsen and Ueland51, Reference Schneider, Boyce, Johnson, Secker, Slade and Grove54, Reference Sultan-Taïeb, Villotti, Berbiche, Dewa, Desjardins and Fraccaroli55, Reference Drake, Skinner, Bond and Goldman59]. Four reported lower healthcare costs [Reference Dattilo50, Reference Evensen, Wisloff, Lystad, Bull, Martinsen and Ueland51, Reference Schneider, Boyce, Johnson, Secker, Slade and Grove54, Reference Drake, Skinner, Bond and Goldman59], for example, a Norwegian RCT comparing SE plus CBT/CR to usual care had better QALY gains with lower healthcare costs after 2 years [Reference Evensen, Wisloff, Lystad, Bull, Martinsen and Ueland51]. In contrast, a Canadian analysis reported significantly lower health care costs for people working in social firms, rather than in supported employment [Reference Sultan-Taïeb, Villotti, Berbiche, Dewa, Desjardins and Fraccaroli55]. It argued this may have been due to the therapeutic nature of social firm support.
Five out of seven cost–benefit analyses of SE interventions had positive net benefits [Reference Fogelgren, Ornstein, Rödin and Thoursie3, Reference Cimera47, Reference Cimera49, Reference Hagen52, Reference Tholén, Hultkrantz and Persson56], for instance, $6,627 to $40,012 per participant in Switzerland under different scenarios over 20 years [Reference Hagen52]. A US study on SE for autism, intellectual disabilities, and other mental health conditions, from a public purse perspective, had various benefit-to-cost ratios, depending the presence of comorbid conditions (without secondary diagnosis = 1.46 versus with secondary diagnosis = 1.49), with the highest net benefits of 2.2 shown for learning disabilities over 6 years [Reference Cimera49]. Another US study of SE for intellectual disabilities [Reference Cimera47] estimated that every $1 investment led to $1.21 of monetary benefits from taxes paid and programme costs forgone. However, there were wide geographical variations across states in benefit–cost ratios ranging from 0.63 in Illinois to 2.77 in Nebraska. In another US analysis, the monthly benefit–cost ratio for supported employment for people living mainly with mental health conditions or mild learning disabilities was at best 0.73:1 from a taxpayer perspective [Reference Cimera48].
Modelling long-term outcomes of a RCT of a Swedish SE service adherent to most IPS principles, but delivered by public employment services, was found to have positive monetary benefits [Reference Fogelgren, Ornstein, Rödin and Thoursie3]. However, uncertainty over long-term costs and sustaining employment meant the time to achieve positive benefits might be between seven and 12 years. Another Swedish intervention, similar to IPS but using employment internships as an intermediate step to employment, was targeted at pupils in their final school year [Reference Tholén, Hultkrantz and Persson56]. Registry data confirmed participation was associated with increased rates of future employment, generating positive net economic benefits after 7 years. In the US a long-term (up to 8.5 years) observational analysis of 112 people in SE or sheltered workshops also found the former had six-fold lower costs per dollar earned [Reference Cimera46].
Discussion
To our knowledge, this is the most comprehensive systematic review on economic evaluations of SE/IPS to date, covering modelling and trial-based studies for all mental health conditions. Our systematic review reveals 56 papers covering 54 economic evaluations examining the economic case for SE/IPS for people living with all mental health conditions. Other reviews revealed only a handful of studies, because of their narrower scope on SMI [Reference Zheng, Stern, Wafford and Kohli-Lynch8, Reference Kinoshita, Furukawa, Kinoshita, Honyashiki, Omori and Marshall60].
Our findings are consistent with commentaries looking at earlier studies in the 2000s and 1990s [Reference Salkever10, Reference Luciano, Drake, Bond, Becker, Carpenter-Song and Lord61]. SE/IPS, when well-implemented can lead to significantly improved work-related outcomes and/or reductions in welfare payments at least in the short term, which partially or even completely offset the costs of intervention. Well-designed RCTs also demonstrate cost-effectiveness from a healthcare perspective; the economic case can be strengthened further when multiple impacts across employment/welfare, health, and other sectors are considered.
In total, 43 (79%) of our included studies reported a positive economic case for intervention with just three (5%) being negative and eight (15%) inconclusive or dependent on subjective judgment. We assessed the quality of these economic studies using the CHEERS checklist; 41 had quality scores above 50%, including 24 with quality scores over 70%. Twelve studies that scored 70% or more on quality focused on IPS, were linked to RCTs, and concluded that it was cost-effective [Reference Knapp, Patel, Curran, Latimer, Catty and Becker17, Reference Christensen, Kruse, Hellstrom and Eplov18, Reference Hellström, Kruse, Christensen, Trap Wolf and Eplov20, Reference Khalifa, Talbot, Barber, Schneider, Bird and Attfield25, Reference Rosenheck, Leslie, Sint, Lin, Robinson and Schooler30, Reference Saha, Bejerholm, Gerdtham and Jarl31, Reference Shi33–Reference Stroupe, Jordan, Richman, Bond, Pogoda and Cao36, 41–43]. Only two IPS studies were from the USA [Reference Rosenheck, Leslie, Sint, Lin, Robinson and Schooler30, Reference Stroupe, Jordan, Richman, Bond, Pogoda and Cao36]; most IPS studies were European. Moreover, SE/IPS studies have grown particularly in the Nordic countries [Reference Fogelgren, Ornstein, Rödin and Thoursie3, Reference Christensen, Kruse, Hellstrom and Eplov18, Reference Hellström, Kruse, Christensen, Trap Wolf and Eplov20, Reference Holmås, Monstad and Reme23, Reference Saha, Bejerholm, Gerdtham and Jarl31, Reference Reme, Grasdal, Løvvik, Lie and Øverland41, Reference Evensen, Wisloff, Lystad, Bull, Martinsen and Ueland51, Reference Tholén, Hultkrantz and Persson56], where IPS services have been implemented and there is good access to welfare and health registry data to follow-up service users.
Implications for policy and practice
Despite evidence on the effectiveness of SE/IPS strategies helping people with mental health conditions to obtain competitive employment, the availability of services remains limited in many countries, where traditional “train and then place” services continue to dominate [Reference Mueser and SR62]. One barrier to the scale-up of services is fragmentation of funding and responsibility for SE/IPS services, leading to budgetary silos. As Table 1 indicates, usually only one sector, such as health, is responsible for funding services, whilst there is a perception that most of the documented economic gains accrue to other sectors, such as through reduced welfare payments, less need for active labour market programmes, and higher earnings. Another barrier is the lack of sustainability. SE/IPS programmes as alternatives to traditional vocational rehabilitation services may not have secure mainstream funding [Reference Radhakrishnan, McCrone, Lafortune, Everard, Fowler and Amos63, Reference Whitworth64].
Overcoming these challenges requires more intersectoral collaborations between health and welfare sectors. Potentially this could include pooled budgets and joint delivery contracts to overcome budgetary silo issues that impede implementation [Reference McDaid and Park65]. Economic evidence can help inform contracting arrangements across sectors, quantifying the economic costs and value of gains to different sectors, as well as the time required for positive economic gains to be achieved. As Table 1 indicates, 24 of the 41 studies had total benefits that covered the additional costs of SE/IPS programmes; societal, government, welfare, and health benefits alone would completely cover these costs for 11, 6, 6, and 8 programmes, respectively. By knowing more about which sectors pay and which sectors gain, it is possible to demonstrate that multiple sectors make gains from SE/IPS, or alternatively estimate the level of financial risk sharing needed across sectors to incentivise implementation. To do this requires more analysis of the impacts across key sectors, including health, as well as a clear indication in all studies of which budget holders are responsible for funding SE/IPS services.
Implications for future research
We put forward a number of suggestions for future research. These include more assessment of long-term cost-effectiveness, making use of registry data, and/or modelling. Cost-effectiveness of implementation of SE/IPS through evaluation under naturalistic conditions rather than just in trials, as well as more assessment of combined interventions and those targeted at specific risk groups is also needed. The interpretation of economic evaluation can be strengthened by more mixed methods research that looks at the context and fidelity under which SE/IPS are implemented.
While economic studies in this review have shown the superior cost-effectiveness of SE/IPS compared to usual care, only six looked at outcomes beyond 5 years. 75% of mental health conditions occur by age 24 [Reference Kessler, Amminger, Aguilar-Gaxiola, Alonso, Lee and Ustün66], adversely impacting future life chances and leading to long-term disability benefits [Reference Knudsen, Øverland, Hotopf and Mykletun67, Reference Mojtabai, Stuart, Hwang, Susukida, Eaton and Sampson68]. If SE/IPS has an effect on this trajectory over the life-course, it will strengthen the economic case further. This could be assessed by linking service participation with long-term social security and welfare data, as seen in Norway [Reference Holmås, Monstad and Reme23], to look at sustainability of outcomes, impacts on long-term earnings and need for welfare benefits.
Although we found some modelling analyses [Reference Mavranezouli, Megnin-Viggars, Cheema, Howlin, Baron-Cohen and Pilling26, 39, Reference Whitworth40, Reference Hagen52] looking at the effectiveness of SE/IPS under different conditions and estimating long-run cost-effectiveness, there is scope for more modelling work. They could extrapolate short-term findings over longer-time periods making different assumptions on long-term employment. Evidence could also be adapted in models to estimate costs and benefits of SE/IPS in comparable countries where no primary economic evaluation exists.
More economic evaluations as part of naturalistic implementation studies, as well as studies determining the economic value of augmenting SE/IPS with additional measures such as psychological therapies, or looking at approaches that are more or less targeted to specific populations, would help inform policy and practice. For example, in the UK, one ongoing evaluation explores whether IPS will reach and support more people with mental health needs, if it is part of primary health services rather than linked to specialist mental health services [Reference Whitworth64]. We also found no studies directly comparing cost-effectiveness by severity of illness or dual diagnosis, or studies on high-risk populations including refugees or those with adverse childhood experiences. Comparisons between programmes that support young people to enter higher education as an alternative/complement to immediate employment are missing.
The results of qualitative research, surveys, and process evaluations on implementation and contextual issues can be incorporated into economic evaluations, for instance, to look at how well programmes engage potential service-users. Referrals to SE/IPS services may be impacted by the attitudes of people in the health and employment sectors, as well as employers [Reference Vukadin, Schaafsma, Westerman, Michon and Anema69]. Different mechanisms to recruit and retain employment specialists, as well as differences in caseloads, could be considered [Reference Whitworth40]. Another gap concerns the impact of adherence to fidelity in service delivery on the cost-effectiveness of programmes; good but less than perfect fidelity has been shown as most cost-effective for first-episode psychosis services [Reference Radhakrishnan, McCrone, Lafortune, Everard, Fowler and Amos63]. Finally, another gap is the lack of economic studies in low-income countries, although costs of delivering SE/IPS have been examined in South Africa [Reference Engelbrecht, van Niekerk, Coetzee and Hajwani70].
Strengths and limitations
We believe our systematic review to be the most comprehensive to date, covering multiple databases in health, psychology, social science, and economics, plus additional searches of gray literature and citation snowballing. While there were no language restrictions, we did not search non-English language databases. So we may have missed literature, for instance in mainland China, where there is emerging evidence on SE/IPS services [Reference Zhang, Tsui, Lu, Yu, Tsang and Li71]. Although we assessed 24 economic evaluations as being high-quality, we may have undervalued the quality of some studies as no existing health economic appraisal checklist is a perfect fit for employment-related studies, where health is usually not the primary focus of evaluation, and studies often are cost–benefit analyses, less used by health economists [Reference Mandrik, Severens, Bardach, Ghabri, Hamel and Mathes15].
Some caution in the generalisability of findings across countries is also needed. Our results cover a very heterogeneous set of studies, using many different comparators, where “usual care” is not always be clearly defined. Country-specific factors, such as generosity of welfare safety nets and employment conditions, might impact on potential cost-effectiveness, although a recent systematic review indicated IPS can be effective, regardless of the generosity of country welfare support [Reference Brinchmann, Widding-Havneraas, Modini, Rinaldi, Moe and McDaid6].
Conclusion
Our review indicates considerable evidence on the cost-effectiveness of IPS and other types of SE in high-income countries, particularly in Europe and North America. The economic case, while strong, is conservative; the short duration of most studies means that long-term economic benefits of being in work, accrued over the life-course, were not captured. It is time to implement SE/IPS, it is a good investment for society.
Supplementary Materials
To view supplementary material for this article, please visit http://doi.org/10.1192/j.eurpsy.2022.2309.
Data Availability Statement
All the studies included in this article are available on the quoted databases (PubMed/MEDLINE, EMBASE, PsycINFO, CINAHL, Business Source Complete, IBSS, and EconLit). The search strategies for all databases are available in Supplementary S1 and all extracted data from the studies are available in Supplementary Tables S2 and S3 as well as in Table 3.
Author Contributions
A.-L.P., D.M.D., and A.M. conceived the study. A.-L.P., D.M.D., A.M., B.B., E.K., and M.R. drafted the original review protocol. D.M.D. and A.-L.P. ran the literature review, obtained full-text papers, screened abstracts and full texts, assessed quality, extracted the data, and wrote the first draft. M.R. and B.B. resolved disagreements on the review inclusion and exclusion of studies. All authors critically reviewed the manuscript for important intellectual content and approved the final submitted version.
Funding Statement
This analysis represents independent work funded by the Research Council of Norway and Centre for Work and Mental Health, Nordland Hospital Trust in Bodø, Norway. The views expressed are those of the authors and not necessarily those of the funders. The funders of the study had no role in study design, data collection, data analysis, interpretations, or writing of this manuscript.
Conflicts of interest
A.-L.P., A.M., B.B., M.R., and D.M.D. received funding from the Research Council of Norway (project number 227097 for IPS Bodø 1, 280589 for IPS Bodø 2, and 273665 for IPSNOR), and funding was also provided by Centre for Work and Mental Health, Nordland Hospital Trust in Bodø, Norway. E.K. and N.A.P.A. have no conflict of interests.
Open access
Comments
No Comments have been published for this article.