Construction of a risk prediction model for peristomal moisture-associated skin damage based on a decision tree method

ZHANG Dongju; GUO Siqin; WANG Wei

doi:10.12102/j.issn.2095-8668.2025.21.007

Abstract：Objective:To construct a risk prediction model for peristomal moisture-associated skin damage(PMASD)based on a decision tree method.Methods:Clinical data from patients who underwent colostomy surgery at our hospital between June 2022 and June 2024 were collected.Univariate and multivariate Logistic regression were used to analyze the risk factors for PMASD in patients with colostomy.Based on the identified risk factors,a decision tree model was constructed to predict the risk of PMASD in patients with colostomy.The predictive value of the risk prediction models constructed using Logistic regression and decision tree algorithms for PMASD was compared.Results:Among 220 patients who underwent colostomy surgery,75 developed PMASD,with an incidence rate of 34.09%.Multivariate Logistic regression analysis showed that stoma site,stoma height,stoma leakage,exposure to radiotherapy/chemotherapy,and stoma self-care ability were influencing factors for PMASD in patients with colostomies(P<0.05).A decision tree model was constructed based on these risk factors.The model consisted of four layers and 11 nodes,with stoma leakage,exposure to radiotherapy/chemotherapy,stoma self-care ability,stoma height,and stoma location as classification nodes.Stoma leakage was the most important predictor.Receiver operating characteristic(ROC)curves showed that the area under the ROC curve(AUC)for predicting PMASD in patients with colostomies was 0.847 for the decision tree model,while the AUC for predicting PMASD in patients with colostomies was 0.844 for the Logistic regression model.There was no statistically significant difference between the two models(P>0.05).Conclusions:Ileostomy,stoma height<1.4 cm,stoma leakage,exposure to radiotherapy/chemotherapy,and poor stoma self-care are risk factors for PMASD in patients with colostomies.The constructed decision tree model has a high predictive value.