By utilizing this primary source data, we design our predictive models for use in the widest population possible and test them across multiple cohorts. These models can report the risks of non-adherence, medication discontinuation, brand switching, patient recruitment, readmission, and other custom metrics.