Judith A. Cook, Ph.D., et al. (1) presented an illuminating reanalysis of data from the Employment Implementation Demonstration Program, perhaps the largest study of supportive employment ever conducted. In their article, the authors shifted the analytic focus from a comparison of supported employment and control interventions (2) to a comparison of programs with greater integration of psychiatric and employment services and those with less integration. This is evidently a crucial shift because the difference in employment rates in the original analysis (55% for those with supported employment and 34% for comparison subjects—a 1.6-fold difference) was almost double in the newer analysis (58% employed in the high-integration programs versus 21% in the low-integration programs—a 2.8-fold difference).
Unfortunately, the explanation of the measure of integration was somewhat ambiguous. On p. 1950, a clearly defined four-level measure of integration was presented. It was subsequently explained that the measure was dichotomized at 50% or more, with integrated sites defined as those that met two or more of the four criteria. Given the hierarchical nature of the four-level measure (the higher levels seem to encompass the lower levels), the threshold for integration would seem to have been met if 1) common charts were used and 2) both types of service were provided by the same agency. This modest level of structural integration seems to have had a substantial impact on employment outcomes, and if replicable in other studies, it could be implemented widely. However, in footnote c to Table 2 and in the first paragraph of the discussion, all four integration criteria are listed, suggesting that perhaps all four had to be met for a program to be considered highly integrated. I inferred from the fact that two-thirds of the programs met the threshold for high integration that the looser criteria had been applied. Could the authors clarify what is the correct definition of the integration variable used in their analysis and whether they believe that such modest structural characteristics can, indeed, improve employment outcomes? If so, what is the role of the quality of supported employment itself? The importance of integration would have received stronger support if the authors had demonstrated a monotonal progressive improvement in employment outcomes as one moves up the five levels of integration on this measure. Was this observed?