Background

Schizophrenia is a mental disorder present in 1% of the population, representing an important source of costs for the health system and society. In 2013, the cost of schizophrenia was $155.7 billion USD in the United States (USA), 38% of which was associated with unemployment, 35% of which was related to lost productivity due to care, and only 24% to direct costs related to health care (Cloutier et al., 2016). It is therefore important to redirect the provision of care services and treatment focused on the control of symptoms towards the promotion of outcomes linked to functional performance of those living with these disorders (Cloutier et al., 2016; Chong et al., 2016).

Employment is a goal of functionality and recovery in people with schizophrenia, linked to an improvement in the quality of life (Bouwmans, de Sonneville, Mulder, & Hakkaart-van Roijen, 2015; Luciano, Bond, & Drake, 2014; van Rijn, Carlier, Schuring, & Burdorf, 2016), social functioning (Charzynska, Kucharska, & Mortimer, 2015), increased self-esteem, improvement of psychiatric symptoms, decrease in the number of hospitalizations, as well as the response to pharmacological treatment (Luciano et al., 2014). Despite the fact that 70% of health service users diagnosed with schizophrenia have expressed an interest to work, the evidence shows that this recovery goal is far from being reached and that, in addition to the low employment rate, they are at risk of living in a street situation, as well as suffering from stigma or victimization (Drake & Whitley, 2014).

Within work rehabilitation models, the Individual Placement and Support (IPS) model has stood out due to its effectiveness in obtaining competitive employment, defined as the condition of being hired in an activity or service in exchange of a salary or payment. The IPS model has solid evidence due to the methodological rigor of its application. It consists of eight postulates: 1. the objective is to obtain competitive employment; 2. it is available to anyone with a serious mental illness (SMI) wih a desire to work; 3. the job search begins when joining the program; 4. potential jobs are chosen according to the preferences of the participant; 5. employment specialists develop relationships with employers, on the basis of patient preferences; 6. the support provided in the program is unlimited in time and continues after the employment is obtained; 7. the program is integrated into the mental health care services; and 8. personalized counseling is provided (Becker & Drake, 1993; Drake, Bond, & Becker, 2012).

The IPS model achieves success rates of over 50% of competitive employment in people with SMI in six-month periods (Drake et al., 1999; Drake & Becker, 1996). However, it has proven to be less optimal in job tenure, since 50% fail to maintain employment at follow-up beyond six months (Bond, Drake, Mueser, & Becker, 1997; Tsang, Chan, Wong, & Liberman, 2009). Accordingly, augmented versions have been developed in which IPS is complemented by other psychosocial interventions (Gold et al., 2006; McGurk, Mueser, & Pascaris, 2005; Mueser et al., 2005; Wallace & Tauber, 2004). Systematic reviews and recent meta-analysis on supported employment in people with SMI (Carmona, Gomez-Benito, Huedo-Medina, & Rojo, 2017; Kinoshita et al., 2013; Modini et al., 2016; Suijkerbuijk et al., 2017) do not evaluate the effectiveness of augmented IPS interventions or stratify their results by diagnosis. This is important due to the functional limitations entailed by schizophrenia (Harvey et al., 2012), as well as the accompanying stigma compared with other SMI, which leads to lower employment rates.

Given that the planning of rehabilitation and reinsertion in mental health services requires more effective interventions, this research sought to evaluate the effectiveness of augmented IPS interventions, designed to obtain competitive employment in people with schizophrenia and other psychoses.

Method

We conducted a systematic review of the literature published in English between 2007 and 2017 in the electronic databases of Medline (PubMed), ScienceDirect, and EBSCO, on IPS-based interventions that added other psychosocial interventions, with the aim of increasing the effect on obtaining or maintaining competitive employment in people with schizophrenia and other psychoses.

Randomized clinical trials (RCTs) of people diagnosed with psychosis, mainly schizophrenia (in more than 60% of the total sample), who were over 18 years of age, were included. The interventions evaluated were those based on IPS in combination with another psychosocial intervention, designed to obtain or maintain competitive employment (this combination was called “augmented IPS”). The main outcome was obtaining and maintaining competitive employment, while secondary outcomes included the number of hours and weeks worked, as well as the salary in dollars (or any local currency).

Search methods for the identification of studies

First, an electronic search that combined MeSH terms and descriptors was carried out. Afterwards, potentially important studies were identified through the references in the articles identified in the electronic search. We attempted to contact the authors of those articles in which more information was required.

Selection of studies and data extraction

The titles of the articles obtained in the search were reviewed and the abstracts were subsequently analyzed to evaluate inclusion and exclusion criteria. This review was conducted by two independent researchers and, in the event of disagreement, a third party intervened. A kappa coefficient of 89% was obtained. Subsequently, full articles were reviewed and main characteristics and outcomes were extracted in duplicate.

Bias risk evaluation

The evaluation was done by three raters and the risk of selection, performance, detection, attrition, and notification bias was established on the basis of the Cochrane Handbook for Systematic Reviews of Interventions.

Statistical analysis

Using competitive employment rates as a dichotomous variable, risk ratios were calculated while means and standard deviations (SD) were reported for the continuous variables (hours, work weeks, and salary). A meta-analysis was carried out using the random effects method, which expands the confidence intervals for the average effect of the intervention and makes statistical significance more conservative (Higgins & Green, 2011). Heterogeneity was calculated using the Cochrane’s Q test and the I² statistic, where a value below 25% denotes low, 50% moderate, and 75% high heterogeneity (Higgins & Green, 2011). The publication bias was evaluated through visual inspection of the funnel plot. Lastly, stratified analyses were performed by diagnosis, comparing the meta-analysis between those with schizophrenia versus first psychotic episode, as well as among the clinical interventions that were more homogeneous according to the raters. The statistical analyses were carried out using Stata version 14 and Review Manager 5 (Revman 5).

Results

Description of studies

Search results

A total of 476 articles were identified, 44 titles were selected, and full text of 23 articles were reviewed by three raters. Lastly, 12 articles comparing 10 augmented IPS interventions with another intervention (IPS, modified IPS, or usual treatment) were selected ( Figure 1 ).

view

Studies included

Twelve studies that met the inclusion criteria were reviewed, although four of them were follow-up studies by Killackey, Jackson, and McGorry (2008), and Baksheev, Allott, Jackson, McGorry, and Killackey (2012) in Australia; Au et al. (2015), and Tsang, Bell, Cheung, Tam, and Yeung (2016) in China. It is worth mentioning that the raters identified similarities in the results shown in the articles by Tsang et al. (2009), and Tsang, Fung, Leung, Li, and Cheung (2010), as a result of which an attempt was made to contact the author. However, in the absence of response, it was decided to consider them as individual studies due to the differences in the total sample, the characteristics of the participants, and the recruitment periods. Therefore, in the end, 10 augmented IPS interventions were described.

Designs

All the studies were randomized clinical trials (RCTs), four of which were multicentric: one by McGurk, Mueser, Feldman, Wolfe, and Pascaris (2007), conducted at two rehabilitation centers in Brooklyn; one by Tsang et al. (2009), carried out at three community mental health units and three-day hospitals; one by Craig et al. (2014), undertaken at four units in England; one by Au et al. (2015) and its follow-up by Tsang et al. (2016), conducted at two day clinics in China.

Place and time

Most of the studies were done in the USA (n = 5), followed in frequency by China (n = 3), Australia (n = 1), and England (n = 1). The follow-up time ranged from six to 39 months, with follow-up rates varying from 89% at 11 months by Tsang et al. (2016), to 76% for the intervention group and 63% for the control group at 39 months in the article by Tsang et al. (2010). The 100% follow-up at 24 months in the article by Bell, Zito, Greig and Wexler (2008) is worthy of note.

Augmented IPS interventions

The set of interventions evaluated was diverse and included the following psychosocial interventions: cognitive remediation, social skills, motivational interventions, job skills, and IPS in addition to usual care. Cognitive remediation strategies were used in five studies, two led by McGurk (McGurk, Mueser, DeRosa, & Wolfe, 2009; McGurk et al., 2007), two by Bell (Bell, Choi, Dyer, & Wexler, 2014; Bell et al., 2008), and one by Au et al. (2015) and its follow-up by Tsang et al. (2016). For the last study, two types of cognitive remediation softwares were used: the Strong Arm System and the Captain´s Log, in which participants could have received up to 72 hours of training. In turn, in those by McGurk et al. (2009; 2007), Cogpack software was used for two to three weeks and up to six months. Bell et al. (2008) used CogRehab and Sci-Learn, adding PositScience’s Brainfitness and Insight software (Bell et al., 2014). Likewise, the two studies by Tsang et al. (2009; 2010) combined IPS and social skills. The three remaining interventions were: a motivational intervention (Craig et al., 2014), one with the addition of job skills (Glynn et al., 2017), and finally one that included IPS plus usual treatment, which included individual attention and referral to external vocational agencies (Killackey et al., 2008).

Control group interventions

Nine articles included IPS-based interventions, five of which were augmented IPS interventions (Au et al., 2015; Bell et al., 2014; Bell et al., 2008; McGurk et al., 2009; Tsang et al., 2009); four standard IPS interventions (Craig et al., 2014; Glynn et al., 2017; McGurk et al., 2007; Tsang et al., 2010), the remainder being based on usual treatment, which included individual care and referral to external vocational agencies (Killackey et al., 2008).

IPS fidelity scale score

Given the variability between interventions derived from the standard IPS model, Bond, Peterson, Becker, and Drake (2012) designed a scale that assesses the degree of adherence to the IPS model standards. This was done on the fact that the implementation of programs to which there is high adherence will achieve similar outcomes to those obtained in controlled studies that established the effectiveness of the model.

Most RCTs (n = 8) included in this review report the score obtained and describe it as adequate (> 80%), with only two articles failing to report this (Au et al., 2015; Bell et al., 2014). A general description of the studies reviewed is given in Table 1 .

view

Evaluation of bias in the studies included

Inter-rater consensus shows a low risk of selection bias, since most of the studies reported random generation of the sequence, carried out mainly through statistical software. There is no clear risk of allocation concealment bias, although neither this nor the blinding of participants and staff is likely to be achieved given the psychosocial nature of the interventions. On the other hand, there is a high risk of incomplete data at follow-up, identified in five studies (Craig et al., 2014; Glynn et al., 2017; McGurk et al., 2009; McGurk et al., 2007; Tsang et al., 2010). The majority showed a low risk of notification bias, and a low risk in relation to publication bias, on the basis of visual inspection of the funnel plot ( Figure 2 ).

view

Population

The total sample was 1 073 participants, with 485 in the augmented IPS intervention groups and 588 in the comparison groups. The average age of those with a schizophrenia diagnosis ranged from 33.5 to 45.5 years. Meanwhile, in the first psychotic episode, the average age was between 21.3 and 25 years. Over half the sample was male, with the exception of the studies by Bell et al. (2014) and its follow-up by Tsang et al. (2010), where they accounted for between 44.8% and 48%, respectively.

Description of job outcomes

Percentages of competitive jobs

Job outcomes are described in Table 2 . Four studies reported statistically significant differences in favor of the intervention groups (IG), with percentages ranging from 33% vs. 12% (Tsang et al., 2010) and up to 69.6% vs.14.3% (McGurk et al., 2007). Bell et al. (2014) only indicated a higher percentage in the subgroup with low community functioning (IG = 49% vs. CG = 20%). Four studies (Au et al., 2015; Bell et al., 2008; Glynn et al., 2017; McGurk et al., 2009) reported higher percentages of competitive employment in control groups (CG).

view

Hours of work per week

Only two studies reported statistically significant differences in favor of the intervention groups: the one by Killackey et al. (2008) in Australia (IG = 33.9 [SD 15.5], CG = 22.5 [SD 10.6], p = .006) and Bell et al. (2008) in the USA. (IG = 59.2 [SD 111.1], CG = 35.4 [SD 81.1], p < .05). In the study by Bell et al. (2014), differences were only shown in the subgroup of low community performance (IG = 230.4 [SD 399.7] vs. CG = 145.3 [SD 441.1], p < .01). No differences were reported in three studies (Craig et al., 2014; Glynn et al., 2017; McGurk et al., 2009) and this variable was not evaluated in four studies (Au et al., 2015; McGurk et al., 2007; Tsang et al., 2009; Tsang et al., 2010).

Weeks worked

Three studies reported more weeks worked in the augmented IPS groups: the one by McGurk et al. (2007) (IG = 27 [SD 37.2], CG = 5.4 [SD 21.1], p < .001), with a follow-up rate of 50% at 26 months; the one by Killackey et al. (2008) (IG = 8.6 [SD 9.2], CG = 3.8 [SD 10.1], p < .05), analyzed during the time of the intervention (6 months); and the study by Tsang et al. (2009), in a 15-month follow-up (IG = 23.8 [SD 17.5], CG based on IPS:12.3 [SD 16], CG vocational rehabilitation services: 1.1 [SD 4.5], p < .001). Four studies reported no effect in favor of the intervention group (Au et al., 2015; Craig et al., 2014; Glynn et al., 2017; McGurk et al., 2009).

Salary

Two studies reported salary differences in favor of augmented IPS interventions: the one by McGurk et al. (2007) (IG = $5 320.2 [SD 8723.2], CG = $530.2 [SD 2105.9, p < .001]); and the one by Killackey et al. (2008) ([IG = $3 526 [SD 4016], CG = $2 865 [SD 9886], p < .05). Five studies found no statistically significant differences between the intervention groups (Au et al., 2015; Glynn et al., 2017; McGurk et al., 2009; Tsang et al., 2009; Tsang et al., 2010) and another three were not evaluated (Bell et al., 2014; Bell et al., 2008; Craig et al., 2014).

Other comparisons

Given the clinical and age heterogeneity between those with schizophrenia and those with a first psychotic episode, the result of the meta-analysis is presented by random effects in each diagnostic category. Figure 3 shows that, among people with schizophrenia, there is no effect in favor of augmented IPS interventions compared to control groups (RR = 1.37, 95% CI [.97, 1.95]). Conversely, the effect does occur in the first psychotic episode group (RR = 3.35, 95% CI [1.23, 9.13]). However, it should be noted that only two studies were included in this comparison ( Figure 3 ). On the other hand, a meta-analysis was performed among the interventions based on cognitive remediation (n = 5), but no statistically significant effect was observed in favor of augmented IPS interventions compared to control interventions (RR = 1.33, 95% CI [.85, 2.10]).

view

Discussion and conclusion

The analysis of augmented IPS interventions fails to show greater effectiveness compared with usual care interventions in people with schizophrenia. However, it should be considered that in most studies, the comparison groups had interventions based on the standard IPS model. Accordingly, it cannot be concluded that the augmented intervention is a better alternative to a standard IPS model designed to obtain competitive employment. On the other hand, there seems to be an effect in favor of interventions aimed at young people with a first psychotic episode. However, the small number of studies suggests that this result should be viewed with caution. Nevertheless, it is reasonable to consider that early interventions will mean greater long-term benefits and that, given the clinical and sociodemographic characteristics of this age group, the treatment goal should include outcomes that are not only job-related, but also academic.

Previous meta-analysis (Carmona et al., 2017; Kinoshita et al., 2013; Modini et al., 2016; Suijkerbuijk et al., 2017) have consistently reported the positive effect of standard IPS interventions in work outcomes. However, very few of them have evaluated the effect of augmented IPS interventions, given that the heterogeneity of the interventions makes it difficult to compare. At the same time, stratified analysis by diagnosis have not been reported. Regarding the negative results of the present study, it should be recalled that the analysis technique used (random effects model) makes more conservative estimates, meaning that the positive effect disappears more than it would with a fixed effects model.

One of the most important limitations of this meta-analysis is related to the heterogeneity of augmented IPS interventions. Despite the fact that most of them used cognitive remediation, this was carried out through various computer programs and with different duration schemes, which makes it difficult to standarize the process and results. It is also necessary to consider losses in long-term follow-up, even in studies that achieved high rates of competitive employment. For example, in the work of McGurk et al. (2007), the follow-up rate at 26 months was 50%. Also, it was not taken into account conditions such as severity and stability of the condition, the role of other relevant variables, baseline cognitive status measured through standardized tests (such as MATRICS cognitive assessment battery), the role of the family (Gold et al., 2006), or any other conditions that could be possible effect modifiers. For example, the meta-analysis conducted by Metcalfe, Drake, and Bond (2018), in which they evaluated the impact of local factors on the likelihood that IPS recipients would obtain competitive employment, compared to those who received alternative vocational services. The authors determined that changes in the Gross Domestic Product, local unemployment, labour union rates, labour benefits, level of compensation of disability benefits, as well as efforts to integrate people with disabilities into the workforce, can serve as barriers or facilitators of the effectiveness of the IPS.

Regarding the applicability of the augmented IPS, it is necessary to consider the context of the interventions, since they are usually spaces where the effectiveness of the IPS model has already been proven and, therefore, when another psychosocial intervention is added, this is done in order to increase its effectiveness, particularly in the long-term. It should also be taken into account that both IPS and the augmented IPS models depend partly on the conviction, faith, and trust of the mental health providers in their application. Consequently, motivational interventions (e.g., motivational interview) as a model for resolving the ambivalence of providers, has shown a considerable effect (Craig et al., 2014).

The fact that the interventions are more effective in the first psychotic episode proves the importance of early interventions. However, in these age groups, alternatives involving a combination with vocational interventions, where the person determines whether the goal is competitive work or the resumption of study, should also be taken into account. Moreover, it should be pointed out that since not all of those who have a first psychotic episode will develop schizophrenia, early interventions in this population will obviously have a better outcome. In first psychotic episode, there are studies evaluating augmented IPS interventions that incorporate vocational components. However, they were excluded from the review since the objetive of this review is focused on competitive employment outcomes (Baksheev et al., 2012; Rinaldi, Perkins, McNeil, Hickman, & Singh, 2010; Rosenheck et al., 2017).

Suggestions to be considered in future research include cost-effectiveness studies on augmented IPS interventions, as well as the study of effect modifiers, in order to determine which characteristics of the intervention or the participants are associated with the greatest effect, since certain conditions such as cognitive impairment (the main complication in severe psychotic disorders) could require attention prior to reincorporation into school or work.

At the same time, there is a need to redirect the mental health budget, since it is currently applied predominantly to the management of acute psychotic episodes. It is therefore necessary to determine the extent to which investment in rehabilitation interventions provides social and economic benefits for families, institutions, and society in general. Although the effectiveness of the implementation of complex interventions in routine mental health services has been proven-such as the Social Security Disability Insurance (SSDI) intervention, which provides supported employment, medication management, and health services, in addition to full health insurance coverage (without out-of-pocket expenses) (Drake et al., 2013), in the Latin American context-, there is a delay in the implementation of these interventions, meaning that it is necessary to involve government institutions and the private sector to open up sensitized work spaces.

It is essential to reduce discrimination in the workplace, since this can determine whether employment has positive effects in terms of self-stigma and stress among people with mental illness (Rüsch et al., 2014). It is therefore necessary to implement interventions to reduce discrimination at work and improve the resources of people with schizophrenia to increase the positive effects of employment (Rüsch et al., 2014). It is also neccesary to identify the skills and attitudes of health personnel and organizational and health system barriers is also necessary (Mueser, Drake, & Bond, 2016).

The IPS model requires an integrated, well-trained and proactive team, which spends most of its time providing personalized support for people with severe mental illness in seeking, obtaining, and maintaining employment. It is necessary to reorient the skills of mental health professionals (medical doctors, psychologists, nurses, and social workers) in order to establish employment as a central objective of rehabilitation and to assess results continuously and critically.

This review adds value to the study of IPS-based interventions and underscores the heterogeneity of augmented IPS interventions and their effects on work outcomes. It also shows that standard IPS interventions are the current basis for the promotion of competitive employment in people with schizophrenia and other psychoses, and that they should be carried out in an early manner. The recovery process should prioritize work and vocational outcomes in the usual mental health care services. This entails the restructure of services, with a new attitude towards the promotion of a mental health system oriented towards the functional recovery of people living with severe mental disorders.