The proposed five-year study focuses upon one of the major methodological questions in the mental health services research field- how to assess the relationship between service system changes and client outcomes? Several recent mental health services research demonstrations, including the RWI/PCMI and the Ft. Bragg projects, have reported findings showing little or no relationship between system improvements and client outcomes. This has generated a lot of debate about whether these results should be interpreted as a theory failure or an implementation failure, or both. The core idea behind the proposed study is that this situation may well result from a methods failure, i.e., the methods employed to date for assessing this relationship are inadequate suggesting that any conclusions about the ineffectiveness of system-level interventions are premature. This study will use network autocorrelation methods to estimate these effects for a variety of client level outcomes in a data set from a major research demonstration. The ACCESS Demonstration for homeless mentally ill persons with co-occurring substance abuse disorders. The ACCESS data set consists of system and client-level measures at three points in time (1994, 1`996, and 1998) for 18 sites in 16 cities. Network autocorrelation models will be applied to this data set cross-sectionally and longitudinally for a range of client outcome variables (e.g., continuity, symptoms, quality of life, level of functioning, and social support). Findings will be considered with regard to their implications for better ways of assessing system-client impacts in future mental health services research studies.
