OBJECTIVE: To identify the characteristics of cost-effective inpatient substance abuse treatment programs. DATA SOURCES/STUDY SETTING: A survey of program directors and cost and discharge data for study of 38,863 patients treated in 98 Veterans Affairs treatment programs. STUDY DESIGN: We used random-effects regression to find the effect of program and patient characteristics on cost and readmission rates. A treatment was defined as successful if the patient was not readmitted for psychiatric or substance abuse care within six months. PRINCIPAL FINDINGS: Treatment was more expensive when the program was smaller, or had a longer intended length of stay (LOS) or a higher ratio of staff to patients. Readmission was less likely when the program was smaller or had longer intended LOS; the staff to patient ratio had no significant effect. The average treatment cost $3,754 with a 75.0% chance of being effective, a cost-effectiveness ratio of $5,007 per treatment success. A 28-day treatment program was $860 more costly and 3.3% more effective than a 21-day program, an incremental cost-effectiveness of $26,450 per treatment success. Patient characteristics did not affect readmission rates in the same way they affected costs. Patients with a history of prior treatment were more likely to be readmitted but their subsequent stays were less costly. CONCLUSIONS: A 21-day limit on intended LOS would increase the cost-effectiveness of treatment programs. Consolidation of small programs would reduce cost, but would also reduce access to treatment. Reduction of the staff to patient ratio would increase the cost-effectiveness of the most intensively staffed programs.