Risk-Need-Responsivity-informed machine learning and geospatial analytics were used to evaluate corrections and behavioral health program outcomes. Using integrated data, the project compared pre-program, during-program, and combined study views to examine outcome prediction, risk identification, county-level program patterns, and decision support. Findings highlight predictive signals in held-out testing, the value of geographic information in selected models, and the importance of using human judgment, not automation, to guide earlier case review in community supervision settings and future evaluation efforts.