OBJECTIVES: The study examined relationships between specific treatment elements and their costs and ten outcome measures using data from a longitudinal outcome study of a Veterans Affairs program for homeless mentally ill veterans. METHODS: Baseline and outcome data over an eight- month period were analyzed for 406 homeless veterans with psychiatric and substance use disorders who were treated in VA's Homeless Chronically Mentally Ill Veterans Program. Multivariate techniques were used to examine the relationship between ten measures of outcome and six treatment elements: program entry via community outreach, the number of contacts with program clinicians, the number of referrals for other services, duration of program involvement, number of days of residential treatment, and increased public support payments. RESULTS: Each of the six treatment elements was significantly related to improvement on at least one of the ten outcome measures. The number of clinical contacts with program staff and the number of days in residential treatment were associated with improvement in the greatest number of outcome domains. However, improvement associated with residential treatment was far more costly than improvement related to other treatment elements. CONCLUSION: This study provides evidence of the effectiveness of a multimodal approach to the treatment of homeless mentally ill persons. However, results indicate that special attention should be paid to to differences in the cost of improvement associated with various treatment elements.