Abstract:
Objective To explore the risk factors of type 2 diabetic retinopathy, establish and verify a visual tool to assist clinical prediction of type 2 diabetic retinopathy patients.
Methods The clinical data of 1172 type 2 diabetes patients with fundus examination were retrospectively analyzed.Using logistic regression analysis, statistically significant variables were screened out, and model and nomogram were established.The classification performance, accuracy and clinical practicality of the prediction model were evaluated using the area under the ROC curve (AUC), correction curve and decision curve analysis (DCA).
Results Age, long course of diabetes, high level of glycosylated hemoglobin (HbA1c), low level of C-peptide 1 hour after meal, positive urinary protein and large coefficient of variation of red blood cell distribution width were risk factors for retinopathy in type 2 diabetes (P < 0.05 to P < 0.01).AUC in training set was 0.772, AUC in validation set was 0.784, and Hosmer-Lemeshow goodness of fit test showed that the goodness of fit was acceptable (P>0.05).DCA showed that the optional threshold probability range for this model was large and relatively safe.
Conclusions The nomogram risk prediction model has good performance, which has indicated the risk indicators for type 2 diabetic retinopathy and can provide basis for early diagnosis and prevention of type 2 diabetic retinopathy.