When conducting economic analyses using administrative claims, it is important to account for the impact of comorbidities on healthcare costs. The Chronic Disease Score (CDS) is a useful risk-adjustment tool for use in pharmacy claims-based research, however the published version does not include important new drugs and the weights were developed from smaller, geographically restricted databases. OBJECTIVES: The purpose of this study is to: (1) update the number and types of drugs included in the CDS, (2) develop new weights based on a large, multi-region pharmacy claims database, and (3) evaluate the revised CDS in explaining variations in and predicting drug expenditures. METHODS: Analyses were conducted using a leading PBM s research sample that contains prescription claims from 3.9 million beneficiaries. Continuously eligibles for 1997 and 1998 were included. A mapping algorithm was developed to link each of 27 diseases with their corresponding medication classes using Standard Therapeutic Class and/or Hierarchical Ingredient Code List codes. A weight system was developed based on relative costs derived using multivariate regression. The second-year drug costs were predicted based on the first-year demographic and CDS scores. To assess the accuracy of the prediction models, a random split-sample method was applied. RESULTS: The revised CDS was moderately correlated with age (r=0.42) and had a high year to year correlation of 0.83. It performed significantly better in explaining variations in and predicting costs than the demographic model. Adjusted R-Square for fitting was fourfold higher: 0.65 vs. 0.15 for log-transformed costs (0.32 vs. 0.06 for actual dollars). CONCLUSION: The revised CDS is better than demographics in explaining the variations in prescription drug costs and predicting the future costs, however further validation work is required. We plan to compare the revised CDS with the old version to determine the incremental improvement in predicting both drug and medical cost.