An RCT Evaluating the Effects of Skills Training and Medication Type on Work Outcomes Among Patients With Schizophrenia

The demographic composition of the sample at study entry and at point of employment are shown in Table 1 and Table 2, respectively. Demographic variables did not differ by randomly assigned medication or by work program, but differences by site emerged for several variables. Of the 107 participants who entered the study, 28 were hospitalized during study participation (IPS only, N=15; IPS plus WPFM, N=13). We did not find a relationship between employment and timing of hospitalization—seven participants who were hospitalized during the study never got a job, eight had a first hospitalization before obtaining a job, six had a first hospitalization after obtaining a job, and seven were hospitalized after ending their first job. Participants in New Hampshire were more likely to be hospitalized than those in Los Angeles (χ²=4.91, df=1, p=.027).

Of the 67 participants who obtained a job, 39 were in IPS only and 28 were in IPS plus WPFM, a nonsignificant difference. [A CONSORT chart is available as an online supplement to this article.] Kaplan-Meier survival curves provided estimates of mean±SE and median days to employment for IPS only (243.03±23.84; median=224) and for IPS plus WPFM (244.64±22.23; median=212). The group difference, after adjustment for site, symptoms, and medication, was not significant. Notably, the estimated times to first job based on the survival analysis are substantially longer than the actual number of days it took to find a job during the study period (Table 3). We present the “raw means” in Table 3 to permit comparison with other studies of IPS; the disparity between the means estimated in the survival analysis and the raw means reported in Table 3 is due to the censoring of observations in the survival analysis. Typical jobs obtained included parking attendant, library assistant, and cleaner; most were entry level and part-time.

Medication did not have a differential impact on clinical or side-effect outcomes, except over the first 12 months of the study, BMI increased by an estimated annual rate of .966 units among olanzapine participants but decreased by an estimated annual rate of .949 units among those randomly assigned to risperidone, a statistically significant medication × time interaction (F=8.82, df=1 and 847, p=.003). The effect of site on BMI (higher in New Hampshire) was significant (F=22.14, df=1 and 847, p<.001). Medication had no apparent impact on any employment outcome.

Among participants who obtained jobs, the participants in IPS plus WPFM had significantly higher total BPRS scores compared with participants in IPS only (t=2.92, df=62, p=.005); as a consequence, we included BPRS total score at the point of employment as a covariate in analyses of phase-2 work outcomes. BPRS and QLS scores were stable subsequent to obtaining a job, with no longitudinal trends or group × time interactions. The IPS only and IPS plus WPFM groups did not differ in rates of attrition after adjustment was made for site, medication, and baseline BPRS total score. Twenty-nine (74%) participants in IPS only and 19 (68%) participants in IPS plus WPFM provided 24-month data. Among those who obtained employment, the groups did not differ on job tenure or any other work outcome, after adjustment was made for medication, site, and phase-2 baseline BPRS symptoms (Table 3).

The above analyses were intent to treat. However, five (22%) participants in IPS plus WPFM attended zero classes. An as-treated analysis that excluded those participants did not reveal any further WPFM benefits. The mean number of skills training classes attended was 18.93±13.51, with a median of 18.50 classes and a range of 0 to 42. A survival analysis conducted with the participants in IPS plus WPFM showed no significant relationship between number of classes attended and job tenure after adjustment was made for medication, site, and symptoms. A subsequent analysis found no benefit from class attendance even among individuals in IPS plus WPFM who held their initial job at least two months and thus presumably would have been most able to generalize skills from the clinic to the work environment.

The primary finding was that clinic-based skills training showed no benefits compared with IPS alone on any work outcomes, despite careful attention to implementation and methodological issues. A secondary result was that type of second-generation antipsychotic medication had a differential effect on BMI but had no other associations with clinical or work outcomes among work-ready outpatients. The employment rate (63%) achieved here is comparable to the average (61%) of 11 IPS trials, which included individuals with a broader range of psychiatric diagnoses (21). Compared with the results of a review by Bond and colleagues (21), the mean annualized total of weeks worked over the two years of the trial was slightly less (47.31 weeks, or 23.65 weeks per year compared with 24.2 weeks per year reported by Bond and colleagues), our participants were less likely to work at least 20 hours per week when they did work (19% versus 43.6%), and they took longer to get a job (243.07 days [estimate from the Kaplan-Meier survival curve] compared with 137.6 days). These disparities may reflect the additional burdens associated with employment experienced by individuals with schizophrenia compared with persons with other psychiatric diagnoses (22).

The participants in IPS plus WPFM were offered structured classes that used a manualized format and were delivered by trained and well-supervised facilitators, to little apparent benefit. WPFM may not be tailored sufficiently to specific jobs, may not generalize to some work settings, or may be unnecessary in the context of IPS services. A parsimonious interpretation of the clinic training’s lack of effect is that job match, accommodations, supports, and other factors determine job tenure of people with serious psychiatric illnesses more than the worker’s job maintenance skills. In other words, successful job tenure results from a good person-environment fit and optimal supports. Consistent with this hypothesis, a recent naturalistic study of job tenure among adults with severe psychiatric illnesses receiving supported employment found that supervisor and coworker supports were the primary predictors of job retention (23). We lacked the data to explore this issue more fully here.

The practical challenges in delivering a clinic-based work support program at the same time that participants began a new job were noteworthy. Many participants struggled with the time and social demands of starting work and a new group simultaneously. The two other published studies that integrated skills training with supported employment dealt with the timing of the training differently. Wallace and Tauber (7) initiated the employment search and skills training simultaneously, meaning that some individuals participated in the modules although they did not yet have a job. Mueser and others (9) recruited people who had been successfully employed for up to two months prior to random assignment to WPFM or not, so some may have lost a job before benefiting from the training. The overall power of the IPS model may make it difficult to achieve further improvements based on skills training.

Several limitations of this study deserve note. First, only 63% of the participants obtained jobs. Although this number is in accord with other studies of IPS for people with schizophrenia (24), it reduced our power to find significant effects of clinic-based skills training. Statistical power was probably not a significant limiting factor in finding WPFM benefits, however, because our outcome data generally showed that the group that received IPS only did as well as or slightly better than the group that received IPS plus WPFM. Second, we randomly assigned participants to the work programs at study entry rather than at time of employment because, consistent with IPS principles, we wanted them to remain with a consistent employment specialist throughout the protocol. This early randomization may have inadvertently contributed to unbalanced groups and symptom nonequivalence at the point of employment. However, using covariate adjustment to control for symptom differences, we still did not find even a trend for a benefit from skills training.

Third, WPFM benefits were also likely constrained by limited opportunities to generalize skills to the work setting; a number of participants assigned to IPS plus WPFM either never came to the groups or continued to attend WPFM sessions even after they had lost their jobs, often very early in their group participation. Obviously, individuals who were not working during the time they participated in WPFM or did not attend the group had less opportunity to practice the skills at work. Thus we could not discern the optimal dose of WPFM (if there is one) from our results. Perhaps if resources have been available to provide transportation to the clinic, more WPFM participants might have attended groups. In addition, we lacked independent confirmation that individuals learned the skills taught in WPFM, although the employment specialists often noted that participants appeared to be using WPFM skills outside of the session. Finally, the requirement that participants agree to randomized assignment to a medication may have made this sample different from one found in a typical behavioral health clinic.

IPS is an effective intervention to help individuals with serious psychiatric illnesses obtain employment. Although concerns have been raised about job tenure among IPS recipients, our findings do not support the addition of clinic-based skills training to address the issue.

The authors thank the employment specialists and staff who worked on the project and Jim Mintz, Ph.D., and Robert P. Liberman, M.D., for their contributions in the development of the study.