Investigation and application of a diabetes risk score in predicting incident diabetes in male senile people in Beijing

Abstract

Objective To develop a diabetes risk score (DRS)to predict the risk of development of incident diabetes in male senile people in Beijing. Methods DRS was developed basing on a test group including a cohort of 1 370 individuals aged 48-87 years without diabetes at baseline, followed for 10 years by Logistic regression and validated on a value group including a cohort of 340 individuals aged 43-88 years without diabetes at baseline. Results The model with the highest area under the ROC curve ( AUC ) included age, hypertension,history of hyperglycemia, body mass index, fasting plasma glucose, triglycerides, and high-density lipoproteincholesterol (HDL-C). DRS was developed basing on this model with a range from 0 to 12 and an optimal cut-off of 4. AUC were respectively 0. 726 ( 95% CI0. 692-0. 759 ) and 0. 765 ( 95% CI0. 691-0. 839 ) in test group and validation group. The sum score value ≥4 had sensitivity of 65.3% and 68. 1%, specificity of 70. 0% and 64.8%, positive predictive value of 37.0% and 23.2%, negative predictive value of 88.2% and 94. 1%.Conclusion The DRS, derived from clinical information combined with plasma glucose and lipids, is an effective tool to predict incident diabetes. Key words: Diabetes risk score; Diabetes mellitus; Receiver-operating characteristic curve; Cohort study