【摘要】 目的 探讨糖尿病肾病（DN）患者透析期间症状负担的影响因素，并基于随机森林算法构建预测模型。方法 纳入2022年10月至2024年10月于我院接受透析治疗的257例DN患者作为研究对象，采用自我感受负担量表（SPBS）将其分为高负担组（181例）及低负担组（76例）。采用医院电子病历系统获取临床基线资料，利用单因素分析确定影响因素后，采用R语言软件和随机森林算法构建DN患者透析期间症状负担的预测模型，并利用受试者工作特征（ROC）曲线评估该模型对症状负担的预测效能。结果 257例DN透析患者中高负担比例为70.4%，低负担比例为29.6%。单因素结果显示，年龄、医疗支出方式、家庭人均月收入、高血压史、透析频率、疼痛恐惧水平、家庭关怀度是影响SPBS评分的影响因素，差异均有统计学意义（均P<0.05）。构建的随机森林预测模型显示影响因素的重要性排序依次为家庭关怀度、疼痛恐惧水平、透析频率、家庭人均月收入和年龄。随机森林模型的ROC曲线下面积（AUC）为0.91，最佳截断值0.50，对应的灵敏度为0.84，特异度为0.99。结论 本研究基于随机森林算法成功构建了DN透析患者症状负担的预测模型并筛选出年龄、家庭人均月收入、透析频率、疼痛恐惧水平和家庭关怀度等关键影响因素，该模型预测效能良好，提示应针对上述因素优化高危人群管理，提供个性化干预措施。更多还原

【Abstract】 Objective To investigate the influencing factors of symptom burden in Diabetic nephropathy(DN) patients during dialysis, and to construct a prediction model based on random forest algorithm. Methods A total of 257 diabetic nephropathy patients who received dialysis treatment in our hospital from October 2022 to October 2024 were included in the study, and were divided into high burden group( n =181) and low burden group(n=76) by the self-perceived burden scale(SPBS). After obtaining clinical baseline data from the hospital electronic medical record system and determining influencing factors by single factor analysis, R language software and random forest algorithm were used to construct a prediction model of symptom burden of DN patients during dialysis, and the receiver operator characteristic(ROC) curve was used to evaluate the prediction efficiency of this model on symptom burden. Results In 257 patients with DN dialysis, the proportion of high burden was 70.4% and the proportion of low burden was 29.6%. The single-factor results showed that age,medical expenditure pattern, family per capita monthly income, history of hypertension, frequency of dialysis,pain and fear level, and family caring degree were the influencing factors of the SPBS score, and the differences were statistically significant(all P <0.05). The stochastic forest prediction model showed that the importance of influencing factors was ranked as family caring degree, pain fear level, dialysis frequency, per capita monthly income and age. The Area Under Curve(AUC) of the random forest model is 0.91, and the sensitivity and specificity corresponding to the optimal cutoff value of 0.50 are 0.84 and 0.99. Conclusion This study successfully constructed a predictive model of symptom burden of DN dialysis patients based on a random forest algorithm, and screened out key influencing factors such as age, per capita monthly family income, dialysis frequency, pain fear level and family caring degree. The predictive efficacy of this model is good, suggesting that the management of high-risk groups should be optimized according to the above factors,and personalized intervention measures should be provided.更多还原