OBJECTIVE: Health districts have been established as part of the decentralization of responsibility within the Italian National Health Service. A major challenge is to assure that appropriate financing is provided to meet the needs of the population. Risk adjustment models are being developed that can be used for districts’ resource allocation, planning and evaluation activities. METHODS: Pharmaceutical, hospital, and demographic data from 2000 and 2001 have been assembled for the entire population of Emilia Romagna, a large northern Italian region (4 million). Pharmaceutical and hospital tariffs were a proxy for costs. Morbidity indicators based upon pharmacy and hospital data were developed for risk adjustment. Prospective risk adjustment models were fit. We tested several models of increasing complexity, taking advantage of the predictive power of pharmacy- and hospital-based diagnostic groups. Our final adjuster was based upon a combination of the pharmacy and hospital groupings. We considered fairness across administrative units, as equity was a key policy goal. RESULTS: The pharmacy cost model predicts 25.8% of the variation in pharmacy costs. Our hospital cost model predicts 10.1% of variation in prospective hospital costs. Predictive accuracy for pharmacy cost models were improved by information from the hospital data; and were more stable for those who used health services in year 1, and better for those who used hospital and pharmacy services compared to those who did not have any service use. For the pharmacy model predictive accuracy by district ranged from from 0.91 to 1.10; for the hospital cost model, predictive accuracy by district ranged from .93 to 1.13. CONCLUSIONS: We demonstrate that risk adjustment models using pharmacy data to identify individual morbidity are good predictors of future year costs. Regional and district health managers can use these models for planning specific interventions and for evaluating patterns of pharmaceutical and hospital use.