The construction and application of risk prediction model for elderly diabetic retinopathy: a cross-sectional study

Background To analyze the independent risk factors of retinopathy in elderly patients with type 2 diabetes mellitus and to establish and evaluate a nomogram prediction model. Methods 1045 elderly diabetic patients from the Diabetes Complications Data Set of the National Population Health Data Archive(PHDA) between December 2016 and December 2021 were included. Univariate analysis, correlation analysis and LASSO regression were used to screen out the influencing variables of diabetic retinopathy in elderly. The independent risk factors of elderly diabetic retinopathy were found by the Logistic regression model, and then the nomogram prediction model was established. The ROC curve, Hosmer-Lemeshow test, calibration curve, and decision analysis curve were used to evaluate the prediction model. Results Diastolic blood pressure, BMI, fatty liver disease, nephropathy, glycosylated hemoglobin, and blood urea were independent risk factors for retinopathy in elderly diabetic patients. The AUC value of the nomogram prediction model was 0.783(95% CI: 0.755–0.810). The Hosmer-Lemeshow test suggested that the model fit was good. The decision analysis curve (DCA) showed that the threshold of the prediction model was higher in the probability range of 0.12–0.86. Conclusion This study established a nomogram prediction model for elderly diabetic retinopathy based on diastolic blood pressure, BMI, fatty liver disease, nephropathy, glycosylated hemoglobin, and blood urea. It has high discrimination and consistency and clinical benefits. It can be used to predict the risk of retinopathy in elderly diabetic patients.