心力衰竭患者多次非计划性入院风险预测模型构建及验证

Construction and validation of a risk prediction model for multiple unplanned hospital admissions in patients with heart failure

关键词:  心力衰竭  多次非计划性入院  危险因素  预测模型

Key words:  heart failure  multiple unplanned admissions  risk factors  prediction model

摘要:

目的探讨影响心力衰竭患者多次非计划性入院的危险因素，构建并验证多次非计划性入院的风险预测模型。 方法采用便利抽样法，选取2020年1月至2022年10月就诊于南通市第六人民医院心内科的498例心力衰竭患者，将其中2020年1月—2021年6月的332例心力衰竭住院患者的临床资料进行统计并分析多次非计划入院的危险因素，利用Logistic回归构建风险预测模型；选择2021年7月—2022年10月的166例患者数据，利用ROC曲线对模型进行验证。 结果332例心力衰竭住院患者中，发生多次非计划性入院的75例，发生率为22.59％。回归分析显示： 年龄＞70岁，房颤史、慢性肾脏疾病、COPD、NYHA分级≥Ⅲ级、Hb＜110g/L、服药≥7种是心力衰竭多次非计划性入院的影响因素。最终构建的预测模型的ROC曲线下面积为0.864（95%Cl： 0.802~0.925），灵敏度为90.8%，特异度为76.8%，最大Youden指数为0.676。 结论本研究构建的心力衰竭多次非计划性入院风险预测模型预测效果较好，可为临床医务人员识别高危患者并及时干预提供指导。

Abstract:

ObjectiveTo explore the risk factors of multiple unplanned admissions in patients with heart failure, and to construct and verify a risk prediction model. MethodsA total of 498 patients with heart failure who were admitted to the Department of Cardiology of Nantong Sixth People’s Hospital from January 2020 to October 2022 were selected by the convenient sampling method. The risk factors of multiple unplanned admissions were analyzed with logistic regression among 332 patients admitted from January 2020 to June 2021(training set), a risk prediction model was constructed. The prediction model was verified among 166 patients admitted from July 2021 to October 2022(validation set) with ROC curve. ResultsIn training set, 75 cases had multiple unplanned admission(22.59%). Logistic regression analysis showed that age >70 years, history of atrial fibrillation, chronic kidney disease, COPD, NYHA class ≥Ⅲ, Hb <110g/L, and polypharmacy were independent risk factors of multiple unplanned hospital admission for heart failure. The area under the ROC curve of the final prediction model was 0.864(95%CI： 0.802-0.925), the sensitivity was 90.8%, the specificity was 76.8%, and the maximum Youden index was 0.676. ConclusionThe risk prediction model of multiple unplanned admission for heart failure constructed in this study has a good prediction effect, which may be used to identify high-risk patients and to intervene in time.