OBJECTIVES: The purpose of this study was to validate a diabetes resource consumption index (DRCI). METHODS: The data for this study was collected from the Southern Arizona Veterans Affairs Health care System. A random split study sample was created from the computerized medical record of veterans with diabetes. Model estimation was done on the first sample (n = 367) and validation using the other 50% of the sample (n = 367). A Fisher's z-statistic was used to assess the degree of correspondence between the predicted and actual values obtained from the derivation versus the validation samples. The construct validity of the DRCI was assessed by comparing it to only demographics, a comorbidity-index, and the revised Chronic Disease Score (CDS). Wilcoxon matched-pairs signed-rank test was used to determine differences between the median squared residual scores between the various risk-adjustment models. RESULTS: The correlation between actual and predicted costs between the derivation and validation samples was not statistically different for the three predicted outcomes. Age and sex accounted for 0.8% and 0.1% of the variance in total and ambulatory cost. The comorbidity index and the CDS individually explained approximately 6% - 10% of the variance in total and ambulatory cost, respectively. The DRCI explained 6% - 8% of the variance in total and ambulatory costs, and did significantly (p <0.05) better than only demographics. The added variance explained by the incorporation of the comorbidity index or CDS accounted for 5% - 8% of the variance in total and ambulatory costs, respectively. CONCLUSIONS: The predictive validity of the DRCI is equivalent to that of the CDS. When the DRCI was used along with the CDS, up to eight percent of variability in costs and utilization were explained. This may suggest that the DRCI and the CDS may be explaining different dimensions of a subject's severity of diabetes.