OBJECTIVES: To examine admissions within the VA facility among veterans with diabetes and the relationship between number of admissions and type of antidiabetic therapy. METHODS: Veterans with diabetes mellitus at a Texas Veterans Health care System were selected by identifying those with an ICD-9 code for diabetes from the outpatient file. Veterans who received continuous care at the VA from FY 2000 to 2002 were selected. Local Patient Treatment File and pharmacy file provided information on all VA admissions and prescription data. A published method on attributing hospital utilization to diabetes based on ICD-9 codes and codes for diagnostic-related groups from administrative databases were used. This method classifies admissions into 3 categories: clearly attributable to diabetes; probably attributable to diabetes; or not attributable to diabetes. Type of medication therapy was defined as the following: oral therapy all 3 years; oral and insulin; insulin only; escalating therapy; and other (i.e. insulin to oral or oral to no medication). RESULTS: Of the 2285 veterans included in the study, there were 1687 admissions at the VA facility over the 3-year period. 772 patients accounted for the admissions. The mean length of stay per admission was 5.34 days (+/- 5.01). Of the admissions, 106 (6.4%) were clearly attributable to diabetes; 578 (35.1%) were probably attributable to diabetes and 964 (58.5%) were not attributable to diabetes. Comparisons showed that mean rank for number of admissions for those on oral therapy only was significantly lower than the following: both insulin and oral (Z = -5.43, p <0.005); insulin only (Z = -4.36, p <0.005); escalating therapy (Z = -6.67, p <0.005) and other (Z = -7.01, p <0.005). CONCLUSIONS: Approximately 40% of admissions among veterans with diabetes were either clearly attributable or probably attributable to diabetes. Receiving oral therapy only was associated with fewer number of hospital admissions than those on more intense therapy.