This study compares the efficacy and cost-effectiveness of two models of services for persons with severe and persistent mental illness, assertive community treatment and standard case management. A randomized experimental research design is used to evaluate outcomes and costs. A secondary objective is to carry forward the exploration of linkages between case management program ingredients and outcomes through use of a theory-driven evaluation approach. Study subjects included persons with a history of severe, persistent mental illness and failure to benefit from officebased treatment indicated by high inpatient treatment use. One hundred-fifty persons evidencing these criteria were randomly assigned for outreach efforts. The first twenty-five persons engaged by each of the program constituted the study groups. Multiple were measured, including inpatient treatment and crisis service use symptomatology functioning, quality of life, and service satisfaction and cost-effectiveness. The theorydriven evaluation approach examined five outcome mediator variables including program structure, service, frequency and intensity, medication compliance, and strength of client-case manager working relationship. Multivariate analyses indicated no significant difference between programs for only one outcome indicator. Standard case management evidenced stronger client-case manager working relationships than assertive community treatment. Regression analyses indicated strength of the client-case manager working relationship is a significant predictor of positive change in inpatient treatment and crisis service use functional status, quality of life, medication compliance, and service satisfaction. Study findings indicate evidence of efficacy and costeffectiveness may not be sufficient to support the movement of assertive community treatment into standard practice. Further research is needed for a direct comparison of the efficacy and cost-effectiveness of the intensive case management and assertive community treatment program models, and to test which program ingredients are empirically linked to outcomes. The theory-driven approach would be well suited to that research objective.