Multivariate Logistic regression analysis and prediction model construction of associated peritonitis in maintenance PD patients
Download as PDF
DOI: 10.25236/icetmr.2024.022
Author(s)
Wu Zhiwei, Lyu Ping, Yuan Ying, Zhu Qiang, Xie Jiayong
Corresponding Author
Xie Jiayong
Abstract
To explore the multivariate logistic regression analysis and prediction model construction of factors affecting the occurrence of associated peritonitis in maintenance peritoneal dialysis (PD) patients. A total of 265 patients undergoing maintenance PD from June 2018 to December 2023 were selected as the study subjects. During the treatment process, they were divided into an incidence group (n=142) and a non-incidence group (n=123) based on whether the patients developed associated peritonitis. Medical records of the two groups were reviewed, and univariate and multivariate Logistic regression analyses were carried out for factors influencing associated peritonitis in the patients. Nomogram model of associated peritonitis was established with R software; ROC curves were drawn to analyze the efficiency of the constructed prediction model (sensitivity and specificity). Univariate and multivariate logistic regression results showed that concomitant gastric diseases, PD age, hemoglobin (Hb), total protein (TP), and urea were independent influencing factors of the associated peritonitis in maintenance PD patients (P<0.05). A prediction model formula was obtained: Z=concomitant gastric diseases×1.907+PD age×(-0.012)+Hb×(-0.019)+TP×(-0.128)+urea×(-0.088)+Mg×1.78. A nomogram prediction model was constructed, which indicated that the score of concomitant gastric diseases was: 37.5, PD age: 27.5, Hb: 47.5; TP: 100; Urea: 70; Mg: 35. ROC curve research revealed that the nomogram prediction model constructed had a greater calibration curve slope (close to 1), and the curve area was 0.814 (95% CI: 0.7643-0.8645); Hosmer-Lemeshow (H-L) fitting χ2=4.528 and P=0.807. Associated peritonitis in maintenance PD patients records a high incidence and is influenced by many factors. Based on the multivariate results, a prediction model has been successfully constructed with high sensitivity and specificity.
Keywords
Maintenance peritoneal dialysis, Associated peritonitis, Multivariate Logistic regression analysis, Prediction model; Prediction efficiency