目的 分析慢性心力衰竭患者（chronic heart failure,CHF）发生再入院的风险因素，构建预测模型并验证，为提出CHF再入院防治策略提供参考。方法 2020年9月至2021年4月，采用方便抽样法抽取山东省某三级甲等医院心内科就诊的CHF患者402例为研究对象，对其随访半年。将患者以7∶3随机分为建模组和验证组，使用随机生存森林算法筛选变量，多因素Cox比例风险回归构建CHF患者再入院风险预测模型，绘制列线图，通过时间依赖性受试者工作特征（receiver operating characteristic，ROC）曲线下面积（area under curve，AUC）、校准曲线和决策曲线分析模型的预测效能。结果 402例CHF患者6个月再入院率为37.31%。慢性病门诊参保、离院方式、1年内入院次数、营养控制状况评分、胱抑素C、红细胞平均宽度及血清高敏肌钙蛋白是CHF患者发生再入院的独立风险因素（均 P ＜0.05）。模型的 AUC 为0.785，内部验证后模型的 AUC 为0.720，两组的模型拟合曲线显示校准曲线和标准曲线基本贴合，两组的临床决策曲线均显示在一定范围内。结论 构建的CHF患者再入院风险预测模型具有良好的预测效能和临床适应性，可在临床推广。

Objective To analyze the risk factors for readmission in patients with chronic heart failure (CHF),construct and validate a prediction model,and provide a reference for proposing strategies to prevent and control CHF readmissions.Methods From September 2020 to April 2021,402 CHF patients admitted to the Department of Cardiology of a tertiary A hospital in Shandong Province were selected using convenient sampling and followed up for six months.Patients were randomly divided into a modeling group and a validation group at a ratio of 7∶3.Variables were screened using the Random Survival Forest algorithm,and a risk prediction model for CHF readmission was constructed using multivariate Cox proportional hazards regression.A nomogram was plotted.The predictive efficacy of the model was evaluated using the area under the time-dependent receiver operating characteristic curve (AUC),calibration curves,and decision curve analysis.Results The 6-month readmission rate among the 402 CHF patients was 37.31%.Chronic disease outpatient insurance status,discharge method,number of hospitalizations within one year,nutritional control status score,cystatin C,red cell distribution width,and serum high-sensitivity troponin were independent risk factors for readmission in CHF patients (all P ＜0.05).The AUC of the model was 0.785,and the AUC after internal validation was 0.720.The model fitting curves of both groups showed that the calibration curves were basically consistent with the standard curves.The clinical decision curves of both groups showed net benefits within a certain threshold probability range. Conclusions The constructed risk prediction model for CHF readmission demonstrates good predictive efficacy and clinical applicability and is worthy of clinical promotion.

R473.54

山东省自然科学基金面上项目（ZR2020MG071）；国家自然科学基金面上项目（72474121）