OBJECTIVES: Adherence to diabetes medications in the Medicare population is low, which can greatly reduce CMS star ratings for managed care organizations (MCOs). Proactive identification of patients at risk for future non-adherence can provide MCOs with a selective cost-effective approach to implement adherence intervention programs. The study aims to develop and validate a risk assessment tool (Prescription Medication Adherence Prediction Tool [RxAPT]) to predict non-adherence to diabetes medications using Medicare claims data. METHODS: Claims data from 2012-2013 was used; data from previous year (baseline period) was used to predict adherence in the next year (follow-up period). Members 65 years and older with diabetes diagnosis, at least one prescription for diabetes medication, no insulin prescription, and continuously enrolled for both the years were included in the study. Adherence in the follow-up year was the study outcome, defined as proportion of days covered (PDC) ≥ 80%. A multiple logistic regression model was used to identify the final model using 70% of the data and risk scores were calculated using significant predictors from the model. The remaining 30% was used for cross-validation using split-sample method. Data from 2011-2012 was used to test the temporal validity of the tool. RESULTS: Total sample included 7028 patients. Seven significant predictors (all from pharmacy claims) were identified and used in the tool. Cross-validation statistics were as follows: C-statistics=0.74, Hosmer-Lemeshow goodness-of-fit p<0.05, sensitivity=0.71, specificity=0.66, positive prediction value=0.75, and negative prediction value=0.62. Temporal validation showed decrease in the sensitivity (0.69) and specificity (0.60) statistics. CONCLUSIONS: RxAPT shows promise as an effective tool to identify patients who are likely to become non-adherent to diabetes medications in the follow-up year. Further validation is needed before the tool can be implemented in a real world setting.