In the U.S., diversion has increasingly become one of the most utilized alternatives to detention of delinquent youth. Programs providing diversion can vary greatly. Variations in program design make it difficult to evaluate the effectiveness of program outcomes. Utilizing hierarchical linear modeling, this study examines variations in outcome for ten program sites of the New York State MH/JJ Diversion Project. Program and youth predictors were evaluated on two outcomes: out-of- community placement and recidivism. At the individual level, significant mental health and substance abuse problems, age, prior placements, and use of wraparound funds were predictive of youth placements, while significant substance abuse problems were predictive of recidivism. Program variations were found to have a significant impact on youth outcomes. Specifically, sites providing direct (or "in house") care had significantly reduced rates of placement. Study results and implications for future research are discussed.