作者：吴凡，张文杰

单位：

212000江苏省镇江市，江苏大学附属医院护理部 通信作者：张文杰，E-mail：zhangwj7878@163.com

单位（英文）：

Nursing Department, Affiliated Hospital of Jiangsu University, Zhenjiang 212000, China Corresponding author: ZHANG Wenjie, E-mail: zhangwj7878@163.com

关键词：

心力衰竭; 不良结局; 影响因素; 风险预测模型;

关键词（英文）：

Heart failure; Adverse outcome; Influencing factors; Risk prediction model

中图分类号：

DOI：

10.12114/j.issn.1008-5971.2023.00.007

基金项目：

江苏省医院协会医院管理创新研究课题（JSYGY-3-2020-427）；江苏大学临床医学专项基金项目（JDLCHL202007）

摘要：

目的 探讨慢性心力衰竭（CHF）患者出院后6个月内发生不良结局的影响因素，构建风险预测模型并进行内部验证和外部验证。方法 采用便利抽样法选取2020年1月至2021年3月在江苏大学附属医院住院治疗并在中国心衰中心数据填报平台接受规范管理的958例CHF患者为研究对象，将2020年收治的758例CHF患者按7∶3的比例随机分为建模组（530例）和内部验证组（228例），2021年1—3月收治的200例CHF患者纳入外部验证组。通过中国心衰中心数据填报平台收集患者的临床资料。采用多因素Logistic回归分析探讨CHF患者发生不良结局的影响因素，构建风险预测模型，采用Hosmer-Lemeshow检验和ROC曲线分析风险预测模型的拟合优度及预测效能，并进行内部验证和外部验证。结果 958例CHF患者中341例发生不良结局，发生率为35.6%。建模组530例CHF患者根据是否发生不良结局分为发生组（n=177）和未发生组（n=353）。两组年龄、合并慢性冠状动脉疾病情况、合并心房颤动情况、合并高脂血症情况、合并低钠血症情况、合并贫血情况、尿酸、肾小球滤过率、血肌酐、尿素氮、脑钠肽（BNP）、查尔森共病指数（CCI）评分、自理能力比较，差异有统计学意义（P<0.05）。多因素Logistic回归分析结果显示，年龄、合并贫血、尿酸、尿素氮、BNP、CCI评分、自理能力是CHF患者发生不良结局的影响因素（P<0.05）。构建风险预测模型，公式为：P=1/{1+exp[-（0.604×年龄+1.172×贫血+0.782×尿酸+0.856×尿素氮+0.760×BNP+0.681×CCI评分3～5分+2.095×CCI评分6～7分+2.064×自理能力中度依赖+4.180×自理能力重度依赖-6.204）]}。HosmerLemeshow检验结果显示，差异无统计学意义（P=0.471），提示该风险预测模型拟合较好；ROC曲线分析结果显示，该风险预测模型预测CHF患者发生不良结局的AUC为0.901[95%CI(0.871,0.923)]，最佳截断值为0.452，灵敏度为0.828，特异度为0.864，约登指数为0.649。内部验证组228例患者中，88例发生不良结局，该风险预测模型预测内部验证组CHF患者发生不良结局的灵敏度为0.818、特异度为0.843、正确率为83.3%。外部验证组200例患者中，76例发生不良结局，该风险预测模型预测外部验证组CHF患者发生不良结局的灵敏度为0.816、特异度为0.856、正确率为84.0%。结论 年龄、合并贫血、尿酸、尿素氮、BNP、CCI评分、自理能力是CHF患者出院后6个月内发生不良结局的影响因素，基于以上因素构建的风险预测模型对CHF患者出院后6个月内发生不良结局具有一定预测价值。

英文摘要：

【Abstract】 Objective To investigate the influencing factors of adverse outcomes within 6 months after discharge in patients with chronic heart failure (CHF) , construct a risk prediction model and conduct internal and external validation. Methods A total of 958 patients with CHF hospitalized in Affiliated Hospital of Jiangsu University from January 2020 to March 2021 and received standardized management on the China Heart Failure Center Reporting Platform were selected by convenience sampling method. Among them, 758 patients admitted from 2020 were randomly divided into the modeling group (n=530) and the internal validation group (n=228) in a 7∶3 ratio. Two hundred CHF patients admitted from January to March in 2021 entered the external validation group. Clinical data of patients were collected through the data platform of China Heart Failure Center. The multivariate Logistic regression analysis was used to analyze the influencing factors of adverse outcomes in patients with CHF. A risk prediction model was constructed, the goodness of fit and prediction efficiency of the risk prediction model were evaluated by Hosmer- Lemeshow test and ROC curve, and the internal and external validation was conducted. Results Among the 958 patients with CHF, 341 had adverse outcomes, accounting for 35.6%. Five hundred and thirty CHF patients in the modeling group were divided into the occurrence group (n=177) and the non-occurrence group (n=353) according to the adverse outcomes. There were significant differences in age, whether combined with chronic coronary artery disease, whether combined with atrial fibrillation, whether combined with hyperlipidemia, whether combined with hyponatremia, anemia, uric acid, glomerular filtration rate, serum creatinine, urea nitrogen, brain natriuretic peptide (BNP) , Charlson Comorbidity Index (CCI) score and self-care ability between the two groups (P < 0.05) . Multivariate Logistic regression analysis results showed that age, anemia, uric acid, urea nitrogen, BNP, CCI score and self-care ability were influencing factors of adverse outcomes in patients with CHF (P < 0.05) . The risk prediction model formula of adverse outcomes in patients with CHF was: P=1/ {1+exp [- (0.604×age+1.172×anemia+0.782×uric acid+0.856×urea nitrogen+0.760×BNP+0.681×CCI score of 3-5 points+2.095×CCI score of 6-7 points+2.064×self care ability of moderately dependent+4.180×self care ability of heavily dependent-6.204) ] } . The results of Hosmer-Lemeshow test showed that the difference was not statistically significant (P=0.471) , which indicated that the goodness of fit of the risk prediction model was good. ROC curve analysis results showed that the AUC of risk prediction model in predicting adverse outcomes in patients with CHF was 0.901 [95%CI (0.871, 0.923) ] , the best cutoff value was 0.452, the sensitivity was 0.828, and the specificity was 0.864, the Youden index was 0.649. Among 228 patients in the internal validation group, 88 patients had adverse outcomes. The sensitivity, specificity and accuracy of this risk prediction model in predicting adverse outcomes in patients with CHF in internal validation group were 0.818, 0.843 and 83.3%, respectively. Among 200 patients in the external validation group, 76 patients had adverse outcomes. The sensitivity, specificity and accuracy of this risk prediction model in predicting adverse outcomes in patients with CHF in external validation group were 0.816, 0.855 and 84.0%, respectively. Conclusion Age, anemia, uric acid, urea nitrogen, BNP, CCI score and self-care ability are influencing factors of adverse outcomes within 6 months after discharge in patients with CHF. The risk prediction model constructed based on the above factors has predictive value for adverse outcomes within 6 months after discharge in patients with CHF.

参考文献：