Use of administrative data to risk adjust amputation rates in a national cohort of medicare-enrolled veterans with diabetes.
Academic Article
Overview
abstract
BACKGROUND: A reduction in diabetes-related lower extremity amputations is a national health care priority. OBJECTIVE: To develop a risk adjustment model for total amputation rates, using claims data. RESEARCH DESIGN: A retrospective longitudinal cohort analysis of veteran clinical users of the Veterans Health Administration (VHA)--Veterans with diabetes who were Medicare nonhealth maintenance organization enrolled in 1997 or 1998. Baseline risks ascertained in 1997 to 1998 were used to adjust Veterans Integrated Service Networks (VISN) amputation rates in 1999. MEASURES: Individual-level amputation outcome in VHA and private hospitals in 1999; VISN-level amputation rates adjusted for age, gender, race, foot risk factors, and macro- and microvascular complications; and rankings of 22 VISNs on amputation rates. RESULTS: A total of 218,528 patients incurred 3077 (14.1 per 1000) amputations in 1999, with 10.6 to 18.0 amputations per 1000 across 22 VISNs. Age, gender, race, prior amputation, infections, ulcers, peripheral vascular disease, and vascular complications were significant independent predictors of amputation (R = 0.20); demographic variables accounted for < 1% of the variance. The C statistic of the final model was 0.83. VISN rankings using age-, gender-, and race-adjusted rates were not substantially altered compared with rankings using the full risk-adjusted model (Spearman rank correlation, 0.85). CONCLUSION: Addition of foot risk and comorbidity variables increased the discrimination of a predictive model for total amputations in an elderly, largely male population of veterans with diabetes compared with use of demographic data alone. The authors suggest that this model be validated in other settings with availability of individual-level claims data.
