[Construction and validation of a prediction model for treatment failure in peritoneal dialysis associated...

Objective: To construct and validate a prediction model for treatment failure in peritoneal dialysis associated peritonitis (PDAP). Methods: A total of 211 PDAP patients who were admitted to Jiangsu University Affiliated Hospital from January 1, 2013 to December 31, 2023 were selected as the modeling group and 72 PDAP patients who were admitted to Nanjing First Hospital from January 1, 2015 to December 31, 2019 were selected as the verification group. Based on the modeling group data, univariate analysis and least absolute shrinkage and selection operator (LASSO) regression were used to screen variables. Logistic regression was used to construct the model, and then a nomogram was developed. Receiver operating characteristic (ROC) curves, calibration curves and decision curves (DCA) were used to evaluate the model efficiency and clinical applicability. Results: A total of 283 PDAP patients (148 males and 135 females) aged (51.1±12.5) years were included. Based on the modeling group data, univariate analysis identified factors affecting the failure of PDAP treatment, including dialysis age, positive ascites culture, albumin, glucose and triglycerides. LASSO regression screened 5 non-zero coefficient indicators, including dialysis age, positive ascites culture, total bilirubin, albumin and triglycerides. Based on the Z-test results of the product under the curve (Z=2.258, P=0.024), the optimal LASSO regression screening variables were selected to establish the model. Multivariate LASSO-logistic regression showed that independent risk factors for PDAP treatment failure included dialysis age (OR=1.011, 95%CI: 1.001-1.022, P=0.034), positive ascites culture (OR=2.605, 95%CI: 1.082-6.272, P=0.033), total bilirubin (OR=1.336,95%CI: 1.143-1.563, P<0.001), albumin (OR=0.880, 95%CI: 0.822-0.943, P<0.001) and triglycerides (OR=1.678, 95%CI: 1.260-2.235, P<0.001). The area under the ROC curve of the model in the modeling group and validation group were 0.765 (95%CI: 0.698-0.833) and 0.794 (95%CI: 0.693-0.896), respectively. In addition, the two sets of calibration curves and decision curves indicated good model consistency and clinical practicality. Conclusions: The factors of PDAP treatment failure include dialysis age, positive ascites culture, total bilirubin, albumin and triglycerides. The prediction model established based on the above indicators had good prediction accuracy and clinical practicality.