OBJECTIVE: To develop a predictive risk-assessment model for asthmatics of a Medicaid managed care population using claims data. The model weighs the predictive value of commonly collected variables to identify members at risk for hospitalizations and emergency room utilization. METHODS: For this retrospective cohort study, asthmatics continuously enrolled from January 1, 1998 through December 31, 1998 were identified. Asthmatics were defined as members with: (1) at least one medical claim with an ICD-9 code (493.00-493.9) for asthma (between 18-65 years of age) or (2) at least one prescription for any asthma-related drug for patients between 18-44 years of age. Patients receiving beta2-agonists must have had at least 2 prescriptions during the 1-year study period. Patients were defined as high- or low-risk based on asthma-related resource use for a 6-month period (7/1/98-12/31/98). Claims from the prior six months were used to obtain predictive variables. A multivariate predictive model was developed using classification and regression tree analyses. RESULTS: We identified 5299 asthmatics of which 13% (694/5299) were classified as high-risk. The model identified high-risk patients with a sensitivity 91%, specificity 27%, positive predictive value 30%. We assigned a higher weight for misclassifying high-risk verses low-risk recipients. The sensitivity of the model outperformed individual predictive variables with the exception of primary care visits. Variables identified as having relative importance include (in order of importance): ER utilization, hospitalizations, number of oral steroid prescriptions, number of asthma prescriptions, and number of primary care visits. CONCLUSION: We conclude that claims data can identify high-risk patients with a high sensitivity, so as to allow for timely intervention. The limitation, however, is poor specificity.