administrative policies, and resources on substance abuse treatment program outcomes. Schildhaus et al. (2000b) examined the stability of staff and ownership/administration in treatment facilities, along with staffing patterns and the orientation of programs toward outcome goals. They found that a substantial number of treatment facilities (about one-third) underwent changes in ownership or administration during the 1990-4 period, and staff turnover averaged about 23 percent. In regression models of drug use after treatment, however, just two measures of program characteristics (treatment modality and facility revenue per patient) were included, in addition to client characteristics and treatment variables (length of treatment, completion of treatment, etc.). They found weak effects of the program variables in most models, although they acknowledged that both better measures (additional information) and statistical methods (such as multilevel modeling) might improve our understanding of the relationships among program, treatment, and individual characteristics.
This literature review shows that many researchers, working with different sets of data and research questions, have come to similar conclusions about the need for better measures and the importance of using a statistical methodology such as multilevel modeling that will facilitate the analysis of interrelationships and cross-level effects among variables measuring program, treatment, and individual characteristics. In initial analyses of the NTIES data using multilevel methods, Orwin and Ellis (2000) in fact found more interaction effects than main effects among program (or service delivery unit) variables and client characteristics on patient outcomes. They comment that their findings confirm that “in substance abuse treatment, the question of ‘what works?’ is more productively specified as ‘what works for whom, and in what setting?’” (viii)
A PUBLIC MANAGEMENT APPROACH TO THE ANALYSIS OF ORGANIZATION