Construction and Verification of Clinical Prediction Models of Peritoneal Dialysis-Associated Peritonitis

Abstract

Objective: To establish the clinical predict models(CPMs) to estimate the risk of incidence of peritoneal dialysis-associated peritonitis (PDAP) and guide clinical practice. Methods: The clinical, psychological, sleep and mental data of 392 PD patients were retrospectively collected. Single factor analysis, LASSO Cox regression analysis and logistic regression analysis were used to establish the CPMs of PDAP. The CPMs was validated according to the Area Under Curve (AUC), Calibration Curve and Hosmer-Lemeshow(H-L) Test, visualized by Nomogram and Decision Curve Analysis (DCA) was used to verify the clinical efficacy of the CPMs of PDAP. Results: Among the 392 followed up PD patients, 69 had PDAP, accounting for 17.60%. There were 120 occurrences of PDAP, 33 patients repeated occurrences, and the incidence of PDAP was 0.31 episodes/patient·year. The risk factors affecting the occurrence of PDAP were serum potassium ([OR] 0.407, 95%[CI] 0.235-0.704; P=0.001); serum albumin ([OR] 0.890, 95%[CI] 0.827-0.958; P=0.002); triglyceride ([OR] 1.653, 95%[CI] 1.246-2.192; P<0.001); MMSE ([OR] 0.740, 95%[CI] 0.635- 0.862; P<0.001); PD patients with catheter and tunnel exit infection ([OR] 47.552, 95%[CI] 7.130- 317.112; P<0.001); PD patients with DM ([OR] 2.961, 95%[CI] 1.116-7.861; P=0.029). The CPMs of PDAP was constructed and visualized with 6 variables. The AUC of the CPMs was 0.856 (95%[CI] 0.782-0.927) and 0.863 (95%[CI] 0.752-0.973), respectively. The Calibration Curves in both the modeling group and the verification group are close to a straight line with a slope of 1, indicating that the model has good prediction ability. The DCA of the modeling group and the verification group all showed a threshold between 5% and 98%, and the DCA curve is at the upper right of the reference line that means the decision curve also had good clinical practicability. Conclusions: This study developed and validated a new CPMs for predicting the occurrence of PDAP.