Construction of a Nomogram Prediction Model for the Risk of Poor Vision After Vitrectomy for Diabetic Retinopathy

YANG Xin; DU Wenwen; HAN Junjun

doi:10.3969/j.issn.1004-437X.2026.10.018

Abstract：Objective To explore the influencing factors of poor vision after pars plana vitrectomy(PPV)for diabetic retinopathy(DR)and construct a risk nomogram prediction model.Methods The clinical data of 222 patients with DR who underwent PPV in the Henan Provincial People's Hospital from June 2020 to June 2022 were retrospectively collected,and they were randomly divided into development set(148 cases)and test set(74 cases)by computer-generated random number table with a ratio of 2:1.All patients were followed up for 12 months after surgery,and the patients with good vision in the development set were included in the good group,and those with poor vision were included in the poor group.The clinical data of the good group and poor group in the development set were compared.Cox regression model was used to analyze the influencing factors of poor vision after PPV,and the regression equation and prediction nomogram model were established.Receiver operating characteristic(ROC)curve,calibration curve were used to evaluate the efficacy of the risk nomogram prediction model.Results Twelve months after PPV,incidence of poor vision in 222 patients with DR,of which 21.62％in the development set and 22.97％in the test set.Cox regression analysis showed that hypertension,preoperative glycated hemoglobin(HbA1c),stage Ⅵ DR,preoperative macular edema,preoperative injection of anti-vascular endothelial growth factor(VEGF)drugs,postoperative blood glucose control was suboptimal were influencing factors of poor vision after PPV for DR(P＜0.05).Cox regression equation was constructed,nomogram model predicted poor vision after PPV in development set with area under curve(AUC),sensitivity,specificity of 0.938,96.77％,87.06％,respectively,Nomogram model predicted poor vision after PPV in test set with AUC,sensitivity,specificity of 0.920,94.44％,86.42％,respectively.Hosmer-Lemeshow test showed that there were no significant differences in the calibration curve between the development set and test set(x2=0.621,P=0.132,x2=0.590,P=0.154).Bootstrap method internal validation showed that the C-index of development set and test set were 0.943(95％CI:0.891-0.995)and 0.930(95％CI:0.876-0.984),respectively.Conclusion Hypertension,preoperative HbA1c,stage Ⅵ DR,preoperative macular edema,preoperative injection of anti-VEGF drugs,postoperative blood glucose control is suboptimal are influencing factors of poor vision after PPV for DR,and the nomogram model was established to predict this vision.