Keywords:Employment, costs, cost-effective, review, mental health problems, IPS

Abstract

There is a growing interest in supported employment within mental health services, reinforced by the importance placed on occupation by service users, and by the development of specialist inter-disciplinary teams with a remit to treat individuals holistically. One model of supported employment, Individual Placement and Support (IPS), has ample evidence for its effectiveness, and this is reviewed briefly here. However, information about the costs of IPS and their relation to outcomes is patchy and equivocal. This paper reviews some basic approaches to evaluating cost effectiveness, looks at what inferences can be drawn from previous studies, including those that span all disability groups, and describes the findings of costs studies focused on interventions for people with mental health problems. In relation to the latter, it summarises what is known, what is uncertain and areas for future research. It makes recommendations about how costs should be taken into account in the development of new employment interventions, and in the evaluation of existing schemes.

---

Aim

The aim of this paper is to describe and summarise evidence for the cost effectiveness of supported employment, highlighting the way forward for the development and evaluation of supported employment.
 

Background

Supported employment is a model of occupational intervention that has attracted considerable interest, and the literature on its effectiveness for people with severe mental health problems is strong.The emphasis placed on occupation in the UK National Service Framework (Department of Health, 1999) and the widespread development of specialist teams practicing assertive outreach, early intervention and crisis resolution highlight the need for greater knowledge about effective employment interventions. Several reviews of this literature have been published. The greatest volume of literature is about the Individual Placement and Support (IPS) form of SE, whose characteristics Bond et al. (1997) list as:

 1. Competitive employment

 2. Rapid job search

 3. Integrated with mental health care

 4. Responds to user preferences

 5 Continuous and comprehensive assessment

 6. Time-unlimited support

Bond, Becker et al. (2001) “do not view Individual Placement and Support as a distinct supported employment model. Instead it is intended as a standardization of supported employment principles … so that supported employment can be clearly described, scientifically studied and implemented in new communities.” The same group of researchers has produced a fidelity scale that assists in this standardization and measurement (Bond, Evans et al., 2000; Bond, Vogler et al., 2001).

Another model of SE operating in the  UK  is known as User Employment (sometimes referred to as Pathfinder after the trust where it began). This is employer-initiated, through a policy of positive recruitment of people with personal experience of mental health problems. Support is offered in the workplace following the IPS approach (Perkins et al., 1997; Perkins, 2001).

Little doubt remains about the effectiveness of IPS. Here, we summarise the evidence very briefly. Bond et al. (1997) compared IPS to other forms of support aimed at employment, including vocational rehabilitation, the standard state-federal supported employment programme, day treatment and sheltered workshops. Overall, the authors conclude that IPS proved more effective with respect to employment outcomes than the other models, but it did not seem to affect non-vocational outcomes (e.g. symptoms, self-esteem). These findings have stood the test of subsequent reviews (Drake et al., 1999; Bond, 2001; Mueser, Bond et al., 2001) as well as from the Cochrane review of vocational rehabilitation by Crowther et al. (2001).

The findings demonstrating the effectiveness of supported employment in terms of work outcomes also gain support from four studies published since the Cochrane review that are described below (Mueser, Salver et al., 2001; Lehman et al., 2001; Bond et al., 2001; Dixon et al., 2002).

The first, the Hartford,  Connecticut, RCT compared IPS to a psychiatric rehabilitation centre (PRC) and standard care (Mueser et al., 2001). This found, at two year follow-up that the IPS group were significantly more likely to obtain any paid work (75%) than those in standard care (54%) or those in the PRC (34%) (Chi squared, 52.71 22.53 respectively, df 2, p<.001).

A second RCT published by Lehman et al. (2001) is important because it was carried out with a population that is relatively disadvantaged. The sample of service users had high levels of psychosis (75%), ethnic minority status (75%), and current substance abuse (40%). It also differed from most previous IPS trials because it did not recruit the sample through induction groups, which may tend to generate self-selected samples of people who are positively motivated to work.

Nevertheless, it was found that the sample on IPS were more likely than those using standard psychosocial rehabilitation to attain employment (47/113 v 12/106,p<.0001; and more likely to be in open work (p<.001). In either group, for those people who achieved employment, hours worked and wages did not differ significantly. Overall rates of employment were relatively low even for IPS (42%) and very low for the comparison group (11%) which possibly reflects low levels of motivation in the participants.

Bond et al. (2001) perform secondary analysis on data from the RCT of IPS conducted in Washington, DC (Drake, McHugo et al., 1999). This study is innovative in its attention to effect size, which is an important topic in studies of interventions where clinical significance and statistical significance may not always coincide. The authors explore the non-vocational outcomes – self-esteem, quality of life and psychiatric symptoms, of four groups of service users: those who did a substantial amount of competitive work, those who did sheltered work, those who did a minimal amount of competitive work and those who did none.  Over 18 months, people in competitive employment had greater satisfaction with vocational services, finances and leisure activities when compared to the rest of the sample taken together. They also showed a greater improvement in self-esteem and psychiatric symptoms, none of which showed any improvement in people who did sheltered work or a minimal amount of open work. However, the ‘control’ group may have deteriorated through demoralisation or discouragement arising from failed employment, making the cause of the difference ambiguous. Two further inferences may be made from the analysis. The findings indicate that it is continued employment, rather than temporary exposure to employment, that has positive effects. Bond et al. also demonstrate that, contrary to the assumption of many clinicians and carers, working does not appear to lead to deterioration in psychiatric symptoms.

In a cost-effectiveness comparison over 18 , Dixon et al. (2002) compare IPS, which focused on competitive employment, with enhanced vocational rehabilitation (EVR), which offered paid training in sheltered settings as well as competitive employment, yields frustratingly equivocal results (Dixon et al., 2002). To overcome the technical difficulties of costs analysis (Gray et al., 1997), this study adopted the perspective of a single payer and used bootstrapping to generate estimates of incremental cost-effectiveness ratios (ICER; increase in effectiveness divided by increase in costs). In the context of this particular study, because the EVR group were also being paid, it was not possible to draw any firm conclusions in relation to the differential impact of IPS on total earnings from competitive and non-competitive sources. Yet employment outcomes were superior, and the authors conclude that “it is statistically highly likely that IPS both costs more and produces more competitive employment” (p 1123). This result highlights the importance of comprehensive evaluation of employment programmes. When an intervention is both more costly and more effective, a proper appraisal requires details about its social acceptability, and any indirect benefits and costs.

While IPS has considerable evidence in its favour when employment is the outcome of interest, there is only limited evidence of its effects on symptoms, self-esteem and other non-vocational outcomes. Questions remain, concerning the impact of working on mental health status and service use. We also need to compare the direct service costs of IPS to those of other mental health interventions, both those aimed at employment and those aimed at other aspects of recovery, to understand more fully the cost effectiveness of the model. Few studies of IPS report its immediate costs, and data on longer-term costs (as well as benefits) is even more limited.

Lacking evidence relating specifically to people with mental health problems, findings concerning supported employment for people with learning disabilities are included here. In the main body of this paper, available costs data are brought together in one place, drawing on reviews of costs studies, evidence from large-scale surveys in the United States, and, finally, costs studies of specific models of supported employment. First, three different approaches to costs analysis are set out, and some of their implications for the generation and interpretation of cost effectiveness information are discussed.

Perspectives on costs

The costs of occupational interventions can be assessed from the perspective of the individual service users, taxpayers or society at large. The simplified components of each of these perspectives are shown in Table 1. This is not a definitive list, but serves to highlight some differences between three approaches. In practice, there is scope for varying these elements according to what information is available as well as what is theoretically desirable. Studies also differ in the length of time over which they measure these elements (longitudinal versus cross-sectional) and in the ways of attaching costs to each of the elements listed.

The interplay between state-funded welfare benefits and individual earnings is clearly a determining factor in estimating the cost-effectiveness of occupational interventions at the level of the individual. By implication, the limit on earnings under certain benefit regimes also affects the taxpayer perspective. Because of the part played by tax and welfare incentives, empirical studies from one system can only be applied with great caution to another welfare context. It should be noted that most studies reviewed here were undertaken in the United States.
 
A cost-benefit ratio can be derived by dividing net benefits (gross benefits minus gross costs) by gross benefits plus gross costs, yielding a ratio between –1 and +1. It is more commonly derived by dividing gross monetary benefits by gross monetary costs. Using this method, cost-benefit ratios of less than one mean that costs exceed benefits, at least in monetary terms. A cost-benefit ratio of less than one needs to be judged in the light of any alternatives. As Thornton (1992) notes, such ratios are meant to inform policy makers who have other social agendas to balance against net costs. For instance, the cost-benefit ratio of a given employment intervention might reasonably be compared to that of day care for the same individual. A return of 80% on investment in employment might for instance be compared to negligible financial return on investment in day care.

Very few studies incorporate a full cost-effectiveness analysis including social costs and benefits, because this entails quantifying in monetary terms intangibles such as user satisfaction and quality of life. Since such potential benefits are seldom measured in costs studies, the evidence may underestimate the actual benefits, and hence the cost-effectiveness, of occupational interventions (Cimera, 2000).

Reviews of costs studies

A critical review by Cimera (2000) identified 21 costs studies of supported employment, nine published since 1991, all relating primarily to people with learning disability. He found that, at the individual level, the cost-benefit ratio is almost always positive, regardless of level of disability.
 
At the taxpayer level, by contrast, the evidence is equivocal. This may change over time, as the costs of the programmes go down and the earnings (and hence taxes paid) by the users go up. By approximately the fourth year, Cimera estimates that taxpayers obtain a net benefit. By comparison, Noble et al. (1991) hold that projections of increasing productivity and decreasing support use are unreliable. They advocate a more conservative approach by taking measurements only during the period in which costs are incurred. The assumption is that the costs of a mature programme that has attained long-term stability with a mix of new and old clients are in fact reliable.

It is important to note, however, that the cost effectiveness of the services reviewed by Cimera appears to be correlated with the level of disability of the clients with learning difficulties (McCaughrin, 1993; Lewis, 1993). ‘Individuals with higher IQs appear to benefit more monetarily from being enrolled in supported employment programmes than individuals with lower IQs’ (Cimera, 2000, p. 57). This casts doubt on how far results from studies dominated by this client group can be extrapolated to people with mental health problems.

It might be assumed that people with mental health problems on average would have higher levels of cognitive functioning. A further consideration is that people with fluctuating illness may periodically incur support costs and experience variations in productivity in a way that does not affect most people with learning disability. These considerations make it practically impossible to draw inferences from cost effectiveness data dominated by people with other disabilities, unless the sample is so large that separate analyses by disability group may be reliable. Below, some key surveys are reviewed.

Large-scale surveys

Several studies of the costs of employment programmes for specific US states have been published since 1991. Although the study of McCaughrin et al. in Illinois (1993) concerns people with intellectual impairment, it conforms to the pattern identified by Cimera (above): the client's cost-benefit ratio was positive, the taxpayer’s was negative in year 1 but (marginally) positive in year 5. The benefits to individuals with mild, as compared to moderate or severe, impairment were 4 times greater. Supported employment was consistently less costly than sheltered work. However, in the learning disability field, SE clients tended to experience greater loneliness on the job. "Efforts to decrease loneliness in competitive employment is an important consideration that needs to be addressed in future research on quality of life for individuals with mental retardation" (p47). There is not much evidence to indicate that this is also true for people with mental health problems.
Similarly,  Lewis et al. (1993) explored the costs of 41 programmes in Minnesota for people with learning disability and concluded that ‘the clearest case for monetary efficiency from a social perspective was found when individual SE programs were compared to all other training and employment options’ (p 83). As noted elsewhere, disability and cost were positively correlated.

Within the sample of 1,250 individuals studied by Noble et al. in New York state (1991), 246 clients were classed as 'mentally ill, psychotic'. The programme costs of people with this label were lower, on a par with people with visual impairment and traumatic brain injury - by contrast with the higher programme costs of people with cerebral palsy, deafness, epilepsy, autism, and learning disabilities. The variance explained by disability group was 19%. For people with psychosis, societal benefits were similar to the average for all clients ($4,468, as compared to $4,500). The taxpayer benefits for this group were also similar ($3,568 as compared to $3,895). Evidence of benefits in psychosocial functioning is unreliable due to small sample size (9). The costs of SE provision conservatively estimated were $6,360 per client (averaged across all disability groups), giving a cost-benefit ratio to society of 0.7 for people with mental health problems. It should be noted that 45% of the subjects of this study were not earning, and their successful placement in paid work could dramatically affect the costs of the programme.

In a state-wide study of the wage effects of supported employment in Michigan, Thompson (1992) used regression analysis to control for factors that might influence wage outcomes for people in SE, compared to people in sheltered work (SW). Only 53 people in the whole sample had mental health problems as a primary diagnosis, most had learning disabilities. Males were over-represented in both SE and SW samples, and younger people predominated in SE. The IQ of people entering SE tended to be higher. Controlling for prior wage levels, gender, race, number of disabilities, IQ and living arrangements, the analysis indicated that SE was associated with higher wages in 10 of the 11 quarters monitored (on average $1.85 ph in 1986-1990). It was also associated with a higher number of hours worked. One problem with this study is that there is likely to be a good deal of variability in work skills that has not been taken into account. Besides being a largely discredited indicator of intelligence, IQ is an inadequate indicator of functional ability: other factors, notably social skills, are arguably more likely to affect employability and hence earnings.

Rusch et al. (1993) included 45 people with mental health problems in their state-wide sample in Illinois, but this was only a small proportion (6%) of the 729 participants with all disabilities, so it is impossible to say whether the findings are applicable to the small sub-group. The state-wide employment programme included a range of provision, mainly as individual placements (53%) and in ‘enclaves’ or groups alongside people without disabilities (39%). Rusch et al.’s findings over time are in line with Cimera’s general inference: individuals benefited throughout the 4 year period, but the economic return to society only became positive in the fourth year, averaging 0.91 overall. The taxpayer perspective showed an average cost-benefit ratio of 0.77. The trend over time was clearly towards greater returns from societal and taxpayer perspectives, but longer-term projections are hampered by a number of uncertainties.

Thornton (1992) highlights the problem of ‘uncertainty’ in cost-benefit analysis of interventions that are novel and complex, and illustrates this with reference to supported employment research. The principal sources of uncertainty are the population served, the process of the intervention and the methods used to evaluate it. Reviewing the history of SE costs studies from the 1980s onwards,  Thornton  explains the favourable findings for SE over sheltered employment settings, as being due mainly to:
 

• -higher earnings in SE, which generate tax revenue;

• -support costs that are geared to need in SE settings, and usually decline over time, by contrast with the constant and unchanging costs of support in sheltered settings; and

• -savings to the extent that SE represents an alternative to day care or other provision.
Should any of these factors differ, for instance in another service delivery model or under a different system of tax and benefits, the cost effectiveness of supported employment would be called into question. We therefore turn now to examine studies of individual interventions, taking particular account of their earnings, support inputs and comparator groups.


Studies of specific interventions

Sciarappa et al. (1993) followed 19 people with mental health problems over a year’s participation in SE (choose-get-keep model). Ten (53%) had a bipolar disorder, two (11%) had current substance abuse and 16 (84%) were ‘taking psychotropic medication’ and all had some work experience (averaging 12 months in previous 5 years). They collected actual costs data over one year prior to enrolment (retrospectively) and during the year on the programme. The results showed increased employment and decreased service use, including a reduction in supported housing. The final benefit-cost ratio came to 0.89. This is judged by the authors to be relatively small, and is attributed to the small size of the programme and limited time frame of the study.
 
There was no control group, so these findings are merely indicative. Moreover, this was a highly selected group of subjects, characterised by their desire to work in a university, and a disproportionate number of people with bipolar disorder, so they are not representative of a typical community mental health caseload. Nevertheless, the finding that supported housing use declined highlights the importance of costing accommodation as well as discrete health and social care services. In the UK, housing typically represents 80% or more of total community care costs, and it is vital to include this element in comparisons in order to make reliable comparisons between programmes that might have an impact on living arrangements, including hospitalisation, as one means of meeting needs for accommodation (Beecham, 1995).

In a natural experiment of transition from day provision to supported employment in two sites, Clark et al. (1996) measured costs and vocational outcomes at three time points between 1989-90 and 1993. Following conversion of the first site from day programme to IPS, case management input increased at both sites, and at site 2 (the ‘control’) outpatient services increased. This may be due to substitution, perhaps driven by clients' needs not met in SE, or it may be affected by providers trying to recapture revenue lost from day care. Taking community costs separately, there was a significant drop following conversion at both sites, suggesting that IPS was taking the place of other community mental health services. However, a general trend towards lower overall costs indicates that programme change was not the only causal factor. There was a negative correlation between work and use of services. The authors conclude that vocational opportunities can be improved significantly without increasing costs. Differences between sites in cost reductions highlight the impact of service context and client characteristics on cost effects. It is important to note that, without closing day care, the total costs would have appeared less favourable. As for negative effects, hospital and crisis intervention remained relatively unchanged, suggesting the conversion had little effect on relapse rates, and, while the 27 regular attendees interviewed recognised several positive outcomes, they reported loss of social opportunities as the main negative effect (Torrey et al., 1995), although this was quickly remedied through the introduction of social groups.

Clark et al. (1998) analyse the New Hampshire study of IPS versus group skills training (GST; a combination of pre-vocational training and supported employment after job placement) from the three perspectives summarised above (Table 5). While they did not include ‘hotel’ accommodation costs, staff costs associated with supported accommodation were included.

At the individual level, while IPS clients earned more, GST clients had more income from benefits, so that the net economic benefits did not differ significantly. The two programmes produce different levels of benefit, depending on the perspective from which they are analysed. IPS had higher benefits from the societal and taxpayer perspectives and lower benefits from the individual perspective, but the differences in net benefits were not significantly different between any of the three perspectives. This study, like many costs analyses, can be criticised because the sample size was too small to detect meaningful differences, given the huge variation usually found in costs.

Over an 18-month follow-up, people using the programme also used more state resources. On average, there was a significant increase in benefit income, causing costs to the state/governement, but not a significant increase in income earned. Analysis of earnings for 137 clients showed that being on benefits reduced earnings significantly (p<.01), as did being on GST. Having a work history increased earnings (p<.05). This analysis controlled for diagnosis, education, site of treatment and baseline severity of mental health problems. It therefore indicates fairly persuasively that being on benefits tends to limit earnings. Much more data would be required to understand the dynamics of this relationship. In addition, such findings would be highly specific to the welfare entitlements and disincentives in operation in a given context (Turton, 2001). The important message to learn from these findings is to proceed with caution, since increasing earnings to clients can also increase costs to the state.

There is evidence from New Hampshire that users of SE with mental health problems appear to make fewer demands on community health care provision while in a SE programme, but this may be related to how local provision is organised (supply-side factors). This finding is not supported by the study of Dixon et al. (2002), who found higher costs for IPS, together with better competitive employment outcomes. The difference between IPS and enhanced vocational rehabilitation in that study depends on the value placed on competitive, as compared to sheltered, employment.

Latimer (2001) undertook a detailed review of three RCTs and five non-randomised studies, and concurs with  Clark’s inference from the New Hampshire studies of IPS that the effect of supported employment on costs depends largely on the service context. The introduction of supported employment into a service where no substitution can be made is likely to increase service costs. Given that direct savings seem unlikely to compensate for the investment in supported employment, Latimer states that such investments may be justified in terms of increasing social inclusion (‘community integration’).

In a large-scale two-site study of two approaches to employment support in England, Johnson et al. (2001) surveyed all clients of a provider agency (Shaw Trust) who were on the government-sponsored Supported Placement Scheme (SPS) and the newer Personal Advisor Scheme (PAS) for disabled people. Both can be seen as forms of Supported Employment. Briefly, SPS was not time-limited and paid a wage subsidy to employers, while PAS was time limited and paid the job broker for training and placement. Both were run under contract to the Department for Work and Pensions by stand-alone providers, voluntary and statutory. We are not aware of any SPS or PAS schemes that operated as an integrated part of community mental health services.

Only a small proportion of respondents to this survey had a mental health problem (7% of PAS clients and 1% of SPS clients) so no analyses were conducted on this sub-group. Overall, cost benefit ratios of PAS were favourable to taxpayer (2.74) and societal (1.12) perspectives, but due to lost benefits and taxes paid, unfavourable to clients (0.47). By contrast, SPS was favourable to client (6.97) and society (1.53) but not so for taxpayers (0.45). These findings were robust to sensitivity testing. Qualitative data also presented.

These findings may be criticised because they do not have a control group, and as such do not allow for the possibility that people might have gone to work without PAS and SPS. The effect of this would be to reduce the relative benefits of the schemes. Another important consideration is the short-term scope of the study, longer-term savings are not known. The novelty of PAS introduces a bias compared to SPS, with the possibility that PAS costs (and benefits) never achieved stability. In this study, it appears that no control was taken of the variable lengths of time that people had been in the schemes. Prior to the publication of these findings, policy developments meant that PAS ceased to exist as such but was ‘rolled out’ as the New Deal for Disabled People  Job Broker scheme, while SPS was phased out altogether and partly replaced by a time-limited placement and support programme known as Workstep.

For our purpose, the fact that these findings are averaged over different disability groups is a severe limitation. It does however highlight the low use made of such programmes by people with mental health problems, and reinforce earlier findings. When Beyer et al. (1996) explored the cost-effectiveness of SPS compared to sheltered workshops, they found that only 6% of their representative sample of SPS and sheltered workshops had mental health problems.

Summary of the evidence

What is known

Since 1991, considerable evidence has been generated concerning the costs and benefits of supported employment as compared to vocational training or sheltered employment. However, this evidence is highly generalised, being drawn largely from state-wide evaluations of global programmes in the US. In both the US and the UK, it applies predominantly to the largest consumers of SE; people with learning disability.
The general findings noted by Cimera, that individuals benefit to a greater degree than the state or society, at least in the short term, is largely supported by the evidence from studies of IPS for people with mental health problems (Clark et al., 1998; Bond et al. 1995). In addition, both of these studies found that people in SE earned significantly more than people who underwent pre-vocational training (earnings evidence summarised by Crowther et al., 2001, Table 2).

At the global (state or national) programme level, Supported Employment tends to be more favourably evaluated than sheltered work or training, and this is because of its superior employment outcomes, which generate greater tax revenue.

What is uncertain

The costs of SE provision are difficult to describe, because of the variation in models, but they appear in most cases to be less than sheltered workshops or day care. However, this is partly due to assumptions about the tapering off of support over time, which may not apply equally to people with mental health problems.
The longer-term impact of SE is still not known. For example, a small increase in hospital use might only be detected in very large samples, and could affect costs disproportionately. Nor do we know if SE has costs implications for housing or education. Perhaps most importantly, there are a priori reasons to expect that their support-use profile will differ considerably from people whose disabilities do not fluctuate.

There are clearly structural factors imposed by welfare benefit systems that constrain earnings and thereby influence benefits from an individual perspective. There may also be structural factors in how mental health systems operate that also affect use of other services by people employment, and hence the variation in costs. For example, a reduction in use of community mental health provision that only operated during working hours might not be due to a decrease in need of that provision, and might be storing up trouble in the longer term.

Future directions for research

The range of benefits measured in most of the costs research reviewed is largely limited to employment outcomes: job tenure and earnings. A few studies take hospitalisation into account. More work needs to be done to quantify ‘intangible’ benefits, which might include satisfaction, quality of life, social inclusion, physical health and independence in activities of daily living. To this end, Kirsh advocates “a shift in focus from individual pathology to the dynamic inter-relationships among the work, the people and the environment in understanding the process and outcomes of work integration for mental health consumers” (2000, p 29). In so doing, the value placed by society on social inclusion would need to be estimated. Given that there is no firm evidence that IPS is more cost-effective than some less inclusive approaches, a social consensus that places greater value on competitive employment that on sheltered work would tip the balance in favour of IPS.
It is likely that the unit cost of supported employment can be reduced by methods such as large caseloads, or retiring people from a programme after a certain period if they have not succeeded in obtaining work (Clark, personal communication). Similarly, the impact on the ‘tapering off’ of support of diagnostic group, severity of impairment, mode of intervention and funding system has yet to be investigated.

There is a strong consensus on the best methodological approaches to the complex task of costing employment interventions. The recommendations of Clark and Bond (1996) are still valid. They list seven key actions for costs research in vocational rehabilitation: measure costs and benefits comprehensively; define the perspective for viewing costs and benefits; use appropriate comparison groups; control for the effects of other mental health treatments on outcomes; examine long-term costs; assess mature programmes as well as new ones; and explore the influence of client characteristics, welfare incentives and other variables (p 233).

Longitudinal data collection should be undertaken for newly-established programmes, and large-scale, cross-sectional data collection for established, robust, programmes. It is vital to include all services used in the costs comparisons, not merely the costs associated with employment programmes or mental health services, to monitor the possibility of cost-shifting. These costs should include accommodation as well as health, education and social care services. Recent advances in health economics indicate that the application of cost-effectiveness acceptability curves (Fenwick et al., 2001) may augment the information that can be derived from imperfect data.

The potential of pooling real-world data pertaining to people with mental health problems from larger data-sets spanning a wider range of disabilities for the purpose of secondary analysis should be investigated. This might be a relatively inexpensive way of exploring further the predictors of SE costs and outcomes. Econometric modelling of such data could investigate, among other things, the level of investment in supported employment programmes that is likely to optimise benefits for individuals, taxpayers and society at large.

We have found no published evidence that such cost-benefit comparisons have been made explicitly between supported employment and alternative, non-employment interventions. The costs and benefits of supported employment have yet to be compared to alternative interventions to judge its relative efficacy, taking individual and societal perspectives into account. In mental health care, these alternatives could include (for instance) individual psychotherapy, group work, day care, art therapy, physiotherapy or no specific intervention.

Conclusion

Supported employment systems are each a product of a particular constellation of labour market conditions, tax structures, social security, education and social welfare provision. It is risky to make assumptions about the cost effectiveness of a given programme on the basis of findings from similar programmes undertaken in different contexts. In most cases, the costs to the taxpayer of supported employment exceed the economic benefits, at least in the short term. Bearing in mind the possibility that the alternatives might provide benefits which employment cannot, supported employment is likely to compare favourably for several reasons. It is socially inclusive, which can be taken as a societal benefit that derives from employment interventions but not from most alternatives. Moreover, employment generates some financial ‘return on investment’, and this has been shown to be a benefit from the individual’s perspective in most cases. This income also offsets the costs to the taxpayer. Since most other mental health interventions listed here are unlikely to make the service user economically better off, much less the taxpayer, the evidence for supported employment over alternative interventions is strong, but still not conclusive. The uncertainty arises from the variability in costs of implementing such support. It is likely that the ultimate criterion of cost effectiveness will be the value that decision-makers place on the greater social inclusion promoted by some models of supported employment.
 

---

 References

Bedell, J.R., Draving, D., Parrish, A., Gervey, R. & Guastadisegni, P. (1998). A description and comparison of experiences of people with mental disorders in supported employment and paid prevocational training. Psychiatric Rehabilitation   Jo urnal, 21,  279-283.

Beecham, J. (1995). Collecting and estimating costs. In: M.R.J. Knapp (Ed) The Economic Evaluation of Mental Health Care (61-82). Arena, Aldershot.

Beyer, S., Goodere, L. and Kilsby, M. (1996). The Costs and Benefits of Supported Employment Agencies, Welsh Centre for Learning Disabilities Applied Research, Cardiff.

Bond, G.R., Becker, D.R., Drake, R.E., Rapp, C.A., Meisler, N., Lehman, A. F., Bell, M.D. & Blyler, C.R. (2001a). Implementing supported employment as an evidence-based practice. Psychiatric Services, 52, 313-322.

Bond, G.R., Drake, R.E., Mueser, K.T., & Becker, D.R. (1997). An Update on Supported Employment for People with Mental Illness. Psychiatric Services, 48, 335-346.

Bond, G.R., Evans, L., Salyers, M., Williams, J., & Kim, H.-W. (2000). Measurement of fidelity in psychiatric rehabilitation. Mental Health Services Research, 2, 75-87.

Bond, G.R., Resnick, S.G., Drake, R.E., Xie, H.Y., McHugo, G.J. & Bebout, R. R. (2001). Does competitive employment improve nonvocational outcomes for people with severe mental illness? Journal of Consulting and Clinical Psychology, 69, 489-501.

Bond, G. R., Vogler, K. M., Resnick, S. G., Evans, L., Drake, R. E. & Becker, D. (2001b). Dimensions of supported employment: Factor structure of the IPS fidelity scale. Journal of Mental Health, 10, 383-393.

Bond, G., Dietzen, L., McGrew, J. & Miller, L. (1995). Accelerating entry into supported employment for persons with severe psychiatric disabilities. Rehabilitation Psychology, 40, 91-111.

Cimera, R. E. (2000). The Cost-efficiency of Supported Employment Programs: A Literature Review. Journal of Vocational Rehabilitation, 14, 51-61.

Clark, R.E. & Bond, G.R. (1996). Costs and benefits of vocational programs for people with serious mental illness. In: M. Moscarelli, A. Rupp and N. Sartorius (eds) Handbook of Mental Health Economics and Health Policy, Vol. 1, John Wiley and Sons Ltd.

Clark, R.E., Bush, P., Becker, D. & Drake, R.E. (1996) A Cost-Effectiveness Comparison of Supported Employment and Rehabilitative Day Treatment. Administration And Policy In Mental Health, 24, 63-77.

Clark, R.E., Xie, H., Becker, D.R., & Drake, R.E. (1998). Benefits and costs of supported employment from three perspectives, Journal of Behavioral Health Services and Research, 25, 22-34.

Crowther, R., Marshall, M., Bond, G. R. & Huxley, P. (2002). Vocational Rehabilitation for People with Severe Mental Illness (Cochrane Review), The Cochrane Libarary, 2002.

Department of Health (1999) A National Service Framework: Mental Health.London: Department of Health.

Dixon, L., Hoch, J., Clark, R., Bebout, R., Drake, R., McHugo, G., Becker, D. (2002) Cost-effectiveness of two vocational rehabilitation programs for persons with severe mental illness, Psychiatric Services, 53, 1118-1124.

Drake, R.E., Becker, D.R., Clark, R.E., & Mueser, K.T. (1999). Research on the Individual Placement and Support Model of Supported Employment. Psychiatric Quarterly, 70, 289-301.

Drake, R.E., McHugo, G.J., Bebout, R.R., Becker, D.R., Harris, M., Bond, G. R. & Quimby, E. (1999). A randomized clinical trial of supported employment for inner- city patients with severe mental disorders. Archives of General Psychiatry, 56, 627-633.

Drake, R.E., McHugo, G., Becker, D., Anthony, W. & Clark, R.E. (1996). The New Hampshire Study of Supported Employment for People with Severe Mental Illness: A Longitudinal Analysis. Journal of Consulting and Clinical Psychology, 64, 391-399.

Gray, A, Marshall, M, Lockwood, A, Morris, J. (1997) Problems in constructing economic evaluations alongside clinical trials. British Journal of Psychiatry, 170, 47-52.

Hallam, A. & Schneider, J. (1999). Sheltered Work Schemes for People with Mental Health Problems: Service Use and Costs. Journal of Mental Health 8, 171-186.

Heyman, A., Turton, N.& Schneider, J. (2002). Occupational outcomes: from evidence to implementation. An expert topic paper commissioned by the Department of Health. Not publicly available. Centre for Applied Social Studies, University of Durham.

Johnson, S., Mitchell, D. & Avis, M. (2002). An Analysis of the Costs and Benefits of The New Deal For Disabled People Personal Adviser Service and the Supported Placement Scheme In Shaw Trust Clients. Report prepared for the Shaw Trust. Nottingham: University of Nottingham.

Kirsh, B. (2000). Factors associated with employment for mental health consumers. Psychiatric Rehabilitation Journal, 24, 13-21.

Latimer, E. A. (2001). Economic impacts of supported employment for persons with severe mental illness. Canadian Journal Of Psychiatry, 46, 496-505.

Lehman, A.F., Goldberg, R., McNary, S., Dixon, L., Postrado, L. & Osher, F. (2001). Improving employment outcomes for persons with schizophrenia and other severe and persistent mental illnesses. Schizophrenia Research, 49, 273-274.

Lewis, D., Johnson, D., Bruininks, H.&Kallsen, L. (1993). Accounting for Costs of Habilitation Training, Sheltered Workshops and Supported Employment. Education and Training in Mental Retardation and Developmental Disabilities, 28, 75-89.

McCaughrin, W., Ellis, W., Rusch, F. & Heal, L. (1993). Cost-effectiveness of Supported Employment. Mental Retardation, 31, 41-48.

Mueser, K. T., Bond, G. R. & Drake, R. E. (2001). Community-Based Treatment of Schizophrenia and Other Severe Mental Disorders: Treatment Outcomes? Medscape Mental Health, 6, 27pp <http://www.medscape.com>

Mueser, K. T., Salyers, M. P. & Mueser, P. R. (2001). A prospective analysis of work in schizophrenia. Schizophrenia Bulletin, 27, 281-296.

Noble, J., Conley, R., Banerjee, S., and Goodman, S. (1991). Supported Employment in New YorkState: A Comparison of Benefits and Costs. Journal of Disability Policy Studies, 2, 39-74.

Perkins, R. (2001) User Employment Programme Progress Report, South West London and St. George's Mental Health NHS Trust, London, pp. 66.

Perkins, R., Buckfield, R., and Choy, D. (1997) Access to employment: A supported employment project to enable mental health service users to obtain jobs within mental health service teams. Journal of Mental Health, 6, 307-318.

Rusch, F., Conley, R., and McCaughrin, W. (1993). Benefit-Cost Analysis of Supported Employment in Illinois. Journal of Rehabilitation, 59, 31-36.

Sciarappa, K., Rogers, E. S., MacDonald-Wilson, K. & Danley, K. (1993). A Benefit-Cost Analysis of a Supported Employment Models for Persons with Psychiatric Disabilities, University of Massachusetts, Massachusetts.

Thornton, C.V. (1992). Uncertainty in Benefit-Cost Analysis of Supported Employment. Journal of Vocational Rehabilitation, 2, 62-72.

Torrey, W.C., Becker, D.R. & Drake, R.E. (1995). Supported Employment: II Consumer, Family and Staff Reactions to a Program Change. Psychiatric Rehabilitation Journal, 18, 69-73.

Turton, N. G. (2001). Welfare Benefits and Work Disincentives. Journal of Mental Health, 10, 285-300.

Vogler, K.M. & Bond, G.R. (1998). Fidelity of an employment program for persons with severe mental illness. Rehabilitation Psychology, 43, 186-186.

__________________________________________________________________________________________________





Copyright © 2003, Hampstead Psychological Associates, Ltd - A Subsidiary of Southern Development Group, SA.
All Rights Reserved.   A Private Non-Profit Agency for the good of all, published in the UK & Honduras.