The epidemiology of cardiovascular disease (CVD) complications in people with diabetes is changing with the increasing prevalence of presentations other than coronary heart disease (CHD) (e.g., heart failure (HF), cardiomyopathy (CM)). Existing CVD risk estimators such as the Framingham Risk Score (FRS), SCORE, and UKPDS Risk Engine primarily assess CHD risk. Our goal was to assess the prediction accuracy of HF and CM from these risk equations.

We evaluated FRS, SCORE, and UKPDS to predict 1-year risk of CHD, HF, and CM from electronic health records of 469 diabetes patients seen at Emory Healthcare. Only adults without a prior history of CVD at baseline were included. A patient was denoted as a case if s/he had a subsequent ICD-9 code of HF, CHD, or CM at the next visit within a year, and if not, s/he was denoted as a control. We calculated the C-statistic of each score for each disease type.

Our population had an average age of 71 with 47% females and 58% Caucasian. CHD, HF, and CM annual incidence rates were 54%, 44%, and 38%, respectively. Across FRS, SCORE, and UKPDS, CHD had the highest C-statistic (0.589-0.608) with HF the next highest (0.547-0.578). CM (0.369-0.426) performed worse than chance. Moreover, UKPDS performed the best across the three CVD complications. These results suggest that CVD risk estimators need to be calibrated for variations in disease presentation.


J.C. Ho: None. L.R. Staimez: None. K. Narayan: None. R.L. Simpson: None. V.S. Hertzberg: None.


National Institute of Diabetes and Digestive and Kidney Diseases (P30DK111024); National Library of Medicine (1K01LM012924-01)

Readers may use this article as long as the work is properly cited, the use is educational and not for profit, and the work is not altered. More information is available at