肠造口周围皮肤损伤风险预测模型的系统综述

A Systematic Review of Risk Prediction Models for Skin Damage around Enterostomy Sites

摘要: 目的：系统评价肠造口患者周围皮肤损伤的风险预测模型，分析相关模型的构建方法、临床预测性能及预测因子，为临床决策提供参考。方法：本研究系统检索PubMed、Web of Science、Embase、中国生物医学文献数据库、中国知网、万方、维普等数据库，全面获取肠造口患者周围皮肤损伤风险预测模型的相关研究文献，检索时限为建库至2026年3月。采用预测模型偏倚风险评价工具(PROBAST)评价纳入研究的偏倚风险，从研究对象、预测因子、结局和统计分析4个领域进行评估。结果：共纳入12篇文献，模型的构建方法有Logistic回归模型和决策树2类，样本量范围120~617例，工作特征曲线下面积为0.803~0.914。预测因子出现频率最高的为造口部位，其次为造口高度、放/化疗史和底盘贴合情况。结论：纳入的模型预测效能较好，整体偏倚风险较高，模型性能可能被高估，临床适用性有待进一步验证。未来研究应遵循PROBAST和TRIPOD指南，避免单因素筛选变量，确保样本量充足，并开展外部验证。

Abstract: Objective: To conduct a systematic review of risk prediction models for peristomal skin damage in patients with enterostomy, analyze the construction methods, clinical predictive performance, and predictors of relevant models, and provide a reference for clinical decision-making. Methods: A systematic search was performed in PubMed, Web of Science, Embase, Chinese Biomedical Literature Database, CNKI, Wanfang, and VIP databases from inception to March 2026 to comprehensively retrieve studies on risk prediction models for peristomal skin damage in enterostomy patients. The risk of bias of included studies was assessed using the Prediction model Risk of Bias Assessment Tool (PROBAST), evaluating four domains: participants, predictors, outcomes, and analysis. Results: A total of 12 studies were included. The modeling methods comprised Logistic regression and decision tree. The sample size ranged from 120 to 617, and the area under the curve ranged from 0.803 to 0.914. The most frequently identified predictor was stoma site, followed by stoma height, history of radiotherapy/chemotherapy, and flange fit. Conclusions: The included models demonstrate good predictive performance but overall high risk of bias. Model performance may be overestimated, and clinical applicability requires further validation. Future studies should follow the PROBAST and TRIPOD guidelines, avoid univariate screening for variable selection, ensure adequate sample size, and perform external validation.