of both treatment settings was also evident in 12-month outcomes in both randomized and self-selecting patients (McKay et al., 1994).

Similar findings have been shown in field studies of private substance abuse treatment programs treating primarily cocaine and cocaine-plus-alcohol-dependent patients (McLellan et al., 1993a; Pettinati et al., 1998). In all of these studies, patients who were assigned to one of several outpatient treatment programs, were less likely to complete treatment than those assigned to the inpatient programs; but those who did complete treatment showed equal levels of improvement and outcome in the inpatient and outpatient settings. It is important to note that virtually all studies of this type have shown greater engagement and retention of patients in inpatient settings.

There have been at least two attempts to formalize clinical decision processes regarding who should, and should not be assigned to inpatient and outpatient settings of care (Cleveland Criteria; American Society of Addiction Medicine [ASAM] Criteria). McKay et al. (1992) failed to show evidence for the predictive validity of the Cleveland placement criteria at least when applied to the assignment of alcohol and drug dependent patients to day hospital or inpatient care. That is, patients who met the Cleveland Criteria for inpatient treatment did not have worse outcomes than those who met criteria for day hospital only when both groups received day hospital treatment. If the Cleveland Criteria had been valid, those who "needed inpatient treatment" but did not receive it should have had poorer outcomes than those who were appropriately "matched" to day hospital. In a similar study evaluating the psychosocial predictors from the ASAM criteria, McKay et al. (1997b) did find at least partial support for the predictive validity of these placement variables. That is, among patients who "needed inpatient treatment" as defined by the psychosocial elements of the ASAM criteria, those who were randomly assigned to outpatient care did show somewhat worse abstinence rates and generally poorer social outcomes than those who were randomly assigned to inpatient rehabilitation. The retrospective nature of this study made it impossible to complete a full evaluation of these criteria.

The most recent versions of the ASAM criteria have attempted to make very fine grained decisions regarding placements to levels of care defined by the amount and quality of medical supervision and monitoring. Research is needed to determine the predictive validity of these finer distinctions and whether placements to settings and modalities with "more medical supervision" actually receive more medical contact or services than placements that are not expected to receive such services.

Length of Treatment/Compliance with Treatment—Perhaps the most robust and pervasive indicator of favorable posttreatment outcome in all