心力衰竭患者死亡风险预测模型的优化研究  

Optimized research on prediction model of death risk in patients with heart failure

作　　者：揭秉章[1] 陈立伟[1] 杨明[1] 韩凌[1] Jie Bingzhang;Chen Liwei;Yang Ming;Han Ling(Department of Cardiology,Fuxing Hospital Affiliated to Capital Medical University,Beijing100038,China)

机构地区：[1]首都医科大学附属复兴医院心血管内科,北京100038

出　　处：《中华心力衰竭和心肌病杂志（中英文）》2024年第3期186-192,共7页Chinese Journal of Heart Failure and Cardiomyopathy

基　　金：北京市西城区卫生和计划生育委员会青年科技人才培养项目(xwkx2016);首都医科大学科研培养基金项目(pyz2018104)。

摘　　要：目的旨在探讨影响心力衰竭(心衰)患者死亡风险的因素,特别关注核心温度与表层温度差值(ΔT)与B型利钠肽(BNP)作为预测因子的效能,并通过联合多个预测指标,以优化心衰死亡风险的预测准确性。方法回顾性入选了2016年1月1日至2019年12月31日期间于首都医科大学附属复兴医院住院的心衰患者,采用单因素和多因素Cox比例风险模型分析,确定心衰患者死亡的独立预测因子。通过绘制受试者工作特征曲线(ROC)并计算曲线下面积(AUC),评估不同指标及联合预测模型在预测心衰患者死亡中的价值。结果共纳入253例患者,87例患者死亡。单独应用ΔT预测心衰患者死亡的AUC值为0.758,95%可信区间(CI)为0.694~0.823,灵敏度为72.9%,特异度为80.4%;单独应用BNP预测心衰患者死亡的AUC值为0.663,95%CI为0.592~0.735,灵敏度为62.4%,特异度为66.1%;单独应用ΔT的预测价值高于单独应用BNP(Z=2.143,P=0.032);ΔT与BNP联合使用可以使AUC值提高至0.771,灵敏度提升至81.2%,但特异度下降至67.9%。与单独应用ΔT预测相比,联合预测的效能并未显著提高(Z=-0.744,P=0.457)。多因素Cox回归分析结果显示,ΔT、BNP、年龄及应用洋地黄类四个指标是心衰患者死亡的独立预测因素,联合这四个指标的预测模型的AUC值显著提升至0.798,灵敏度为82.4%,特异度为69.0%。这一联合预测模型与ΔT单独预测相比,其预测效能显著提高(Z=-1.965,P=0.049)。结论本研究优化了心衰患者死亡风险的预测模型,为临床实践中评估和管理心衰患者提供了新的预测工具,有助于改善临床决策的质量。ObjectiveTo explore the factors that affect the risk of death in patients with heart failure(HF),with a focus on the efficacy of core-surface temperature difference(ΔT)and B type natriuretic peptide(BNP)as predictors,and to optimize the accuracy of predicting HF mortality risk by combining multiple predictive indicators.MethodsPatients with HF admitted to Fuxing Hospital Affiliated to Capital Medical University from January 1,2016 to December 31,2019 were retrospectively enrolled.Univariate and multivariate Cox proportional risk models were used to determine independent predictors of death in patients with HF.The value of different measures and combined prediction models in predicting death in patients with HF by using of the receiver operating characteristic curve(ROC)and area under the curve(AUC).ResultsA total of 253 patients were included in the study,87 of whom died.The AUC value ofΔT alone as a predictor of death in HF patients was 0.758 with 95%confidence interval(CI)of 0.694~0.823 and sensitivity of 72.9%and specificity of 80.4%.The AUC value and 95%CI,sensitivity and specificity for BNP alone as predictor of death in HF was 0.663(0.592~0.735),62.4%and 66.1%,respectively.The predictive value ofΔT alone was higher than that of BNP alone(Z=2.143,P=0.032).Combination ofΔT and BNP can increase the AUC value to 0.771 and the sensitivity to 81.2%while decrease the specificity to 67.9%.Compared withΔT prediction alone,the combined prediction performance was not significantly improved(Z=-0.744,P=0.457).Multivariate Cox regression analysis showed thatΔT,BNP,age and digitalis application were independent predictors of death in patients with HF.The AUC value of the combined prediction model was significantly increased to 0.798 with the sensitivity of 82.4%and the specificity of 69.0%.Compared withΔT alone,the prediction efficiency of this combined prediction model is significantly improved(Z=-1.965,P=0.049).ConclusionsThis study optimized the predictive model for mortality risk in patients with HF,providing a

关 键 词：心力衰竭 死亡风险 预测模型 COX回归分析 ROC曲线 

分 类 号：R541.6[医药卫生—心血管疾病]