Predicting hypoglycemia in elderly inpatients with type 2 diabetes: the ADOCHBIU model

Hypoglycemic episodes cause varying degrees of damage in the functional system of elderly inpatients with type 2 diabetes mellitus (T2DM). The purpose of the study is to construct a nomogram prediction model for the risk of hypoglycemia in elderly inpatients with T2DM and to evaluate the predictive performance of the model.

From August 2022 to April 2023, 546 elderly inpatients with T2DM were recruited in seven tertiary-level general hospitals in Beijing and Inner Mongolia province, China. Medical history and clinical data of the inpatients were collected with a self-designed questionnaire, with follow up on the occurrence of hypoglycemia within one week. Factors related to the occurrence of hypoglycemia were screened using regularized logistic analysis(r-LR), and a nomogram prediction visual model of hypoglycemia was constructed. AUROC, Hosmer-Lemeshow, and DCA were used to analyze the prediction performance of the model.

The incidence of hypoglycemia of elderly inpatients with T2DM was 41.21% (225/546). The risk prediction model included 8 predictors as follows(named ADOCHBIU): duration of diabetes (OR=2.276, 95%CI 2.097˜2.469), urinary microalbumin(OR=0.864, 95%CI 0.798˜0.935), oral hypoglycemic agents (OR=1.345, 95%CI 1.243˜1.452), cognitive impairment (OR=1.226, 95%CI 1.178˜1.276), insulin usage (OR=1.002, 95%CI 0.948˜1.060), hypertension (OR=1.113, 95%CI 1.103˜1.124), blood glucose monitoring (OR=1.909, 95%CI 1.791˜2.036), and abdominal circumference (OR=2.998, 95%CI 2.972˜3.024). The AUROC of the prediction model was 0.871, with sensitivity of 0.889 and specificity of 0.737, which indicated that the nomogram model has good discrimination. The Hosmer-Lemeshow was χ ² = 2.147 (P=0.75), which meant that the prediction model is well calibrated. DCA curve is consistently higher than all the positive line and all the negative line, which indicated that the nomogram prediction model has good clinical utility.

The nomogram hypoglycemia prediction model constructed in this study had good prediction effect. It is used for early detection of high-risk individuals with hypoglycemia in elderly inpatients with T2DM, so as to take targeted measures to prevent hypoglycemia.

ChiCTR2200062277. Registered on 31 July 2022.