The prevention of type-2 diabetes (T2DM) has become imperative to stem the rising rates of this disease, particularly among black Americans; however, the at-risk pool is large and a clinically meaningful metric for risk stratification to guide interventions remains a challenge. The objective of this analysis is to predict diabetes risk using the full-information continuous analysis from nationally sampled data from white and black American adults ≥ 45 years.


The population-based cohort, the REasons for Geographic and Racial Differences in Stroke (REGARDS) (2003-2007), was observed through 2013-2016. A sex and race stratified cardiometabolic score using data regularly assessed in the primary care setting to assess the association between T2DM in a sample of 12,043 black and white men and women using a series of logistic regressions, accounting for blood glucose, blood pressure, HDL-cholesterol, waist circumference or body mass index, triglycerides, and age. Discrimination was assessed with area under the receiver operating characteristic curves (AUCs) and C-statistics.


In the REGARDS cohort, there were 1,602 incident cases of diabetes. The final clinically meaningful model did not include interactions, as these did not significantly improve AUC, and contained: age, sex, race, BMI, triglycerides, HDL, blood pressure and blood glucose. Using BhGLM methods and the REGARDS dataset with available predictors, the final BhGLM model outperformed scores developed by Framingham (AUC = 0.76), the American Diabetes Association (AUC = 0.68), and ATP-III criteria (AUC = 0.77).


A BhGLM model using full-information continuous dada has high model discrimination using available clinical information, and can be used to quantify race- and sex-specific diabetes risk providing a new powerful predictive tool. This tool can be used for diabetes prevention efforts by allowing clinicians to target high risk individuals in a manner that could be used to optimize outcomes