Abstract: Objective To construct a risk prediction model for hypoglycemia within 6 months in the elderly patients with type 2 diabetes mellitus (T2DM). Methods Convenience sampling was used to select 277 elderly patients with T2DM in Affiliated Geriatric Hospital of Nanjing Medical University from February to August 2022. According to whether hypoglycemia occurred within 6 months of follow-up, the patients were divided into hypoglycemia group (69 cases) and non-hypoglycemia group (208 cases). LASSO regression and multivariate Logistic regression analysis were used to determine the influencing factors, and a nomogram prediction model was constructed and evaluated. Results 24.91% (69/277) of the patients presented with hypoglycemia within 6 months. History of hypoglycemia (OR=6.989, 95%CI: 3.671-13.844, P<0.001) was the risk factor, while high MMSE score (OR=0.890, 95%CI: 0.820-0.959, P=0.003) and triglyceride ≥1.7 mmol/L (OR=0.268, 95%CI: 0.083-0.708, P=0.014) were the protective factors for hypoglycemia in the elderly patients with T2DM within 6 months. The area under the receiver operating characteristic curve (AUC) of the model was 0.801 (95%CI: 0.742-0.860), with a sensitivity of 0.841, a specificity of 0.673, and an accuracy of 71.48%. In the internal validation, the corrected AUC was 0.783, indicating that the model had good discrimination. The Hosmer-Lemeshow goodness of fit test showed that the model had a good calibration (χ²=10.81,P=0.212). The results of clinical decision curve analysis showed that when the prediction threshold of the model was between 0.02 and 0.69, the net clinical benefit level of the patients was the highest. Conclusions The constructed model has good discrimination, calibration and clinical applicability, which can provide a basis for medical staff to identify high-risk group of hypoglycemia and take preventive intervention.

Key words: aged, type 2 diabetes mellitus, hypoglycemia, assessment of cognitive function, prediction model