Factors influencing corneal edema after phacoemulsification combined with intraocular lens implantation in elderly diabetic cataract patients and establishment of a predictive model.

This study aimed to explore the factors influencing corneal edema after phacoemulsification combined with intraocular lens implantation in elderly diabetic cataract (DC) patients and to establish a predictive model. A retrospective analysis was conducted on 300 DC patients who underwent phacoemulsification combined with intraocular lens implantation in our hospital between June 2022 and October 2023. Patients were randomly divided into a modeling group (n = 210) and a validation group (n = 90) in a 7:3 ratio. The modeling group was further divided into corneal-edema group (n = 57) and non-corneal-dema group (n = 153) based on the occurrence of postoperative corneal edema. Univariate and binary logistics regression analyses were employed to identify the influencing factors of postoperative corneal edema in DC patients. A predictive model was constructed using SPSS, and the model's application value was evaluated using receiver operating characteristic curve, calibration curve, and decision curve analysis in R language. There were no statistically significant differences in terms of general information between the modeling group and validation group (P > .05). Significant differences were observed in the comparison of diabetes duration, preoperative endothelial cell density, postoperative corneal endothelial cell loss, hypertension, surgery duration, and age between the corneal-edema group and non-corneal-edema group (P < .05). The results of binary logistics regression analysis showed that diabetes duration, preoperative endothelial cell density, postoperative corneal endothelial cell loss, hypertension, surgery duration, and age were influencing factors for postoperative corneal edema in DC patients (P < .05). The calibration curves of the model in the modeling group and validation group exhibited a slope close to 1, showing good consistency between predicted and actual risks. Receiver operating characteristic analysis results indicated an area under the curve of 0.96 and a standard error of 0.021 (95% confidence interval: 0.872-0.994) in the modeling group and an area under the curve of 0.94 and a standard error of 0.024 (95% confidence interval: 0.867-0.966) in the validation group. Decision curve analysis curve analysis demonstrated that the model had good application efficacy and clinical benefits in both groups. The predictive model based on influencing factors demonstrated good predictive performance and warrants further validation.