A Simple Score Model to Assess Prediabetes Risk Status Based on the Medical Examination Data

Abstract

Objectives: We aimed to build a risk score model to screen out the patients at high-risk status so as to prevent or delay the conversion of prediabetes to diabetes.

Methods: The population were divided into 2 groups: 1 was an exploratory population, and the other was a validation population. All the data were extracted from the electronic medical examination datasets in the School Hospital of Harbin Institute of Technology, Harbin, China. A binary logistic regression model was used to screen out the risk factors, and the associated risk factors were categorized into 3 levels to create the prediabetes score model. We divided the total score into 4 risk categories: low, middle, high and extremely high risk. We also tested the performance of our prediabetes risk score model.

Results: Age, body mass indexes, histories of hypertension, family histories of diabetes, diastolic blood pressure levels and triglyceride levels were screened out as independent risk factors in order to build the risk score model. The area under the curve (AUC) of the prediabetes risk score model was 0.748 (95% CI, 0.720 to 0.777), and the AUC for the validation population reached 0.713 (95% CI, 0.686 to 0.740). Low, middle, high and extremely high risk statuses for prediabetes were associated with a total score of 0 to 3, 4 to 6, 7 to 10 and 11 to 12.

Conclusions: Our prediabetes score model can be used easily and understood by doctors and other related users to assess prediabetes risk status. The intervention program, designed based on our prediabetes score model, is likely to prevent or delay the conversion of prediabetes to diabetes.

Keywords: catégorisation des risques; données des examens médicaux; medical examination data; modèle de cotation du risque; prediabetes; prédiabète; risk categorization; risk score model.