基于机器学习的模型的开发和验证,以预测危及生命的心室失律症的住院死亡率:回顾性队列研究
Le Li1, Ligang Ding1, Zhuxin Zhang1
1National Center for Cardiovascular Diseases, Fu Wai Hospital, Peking Union Medical College, Chinese Academy of Medical Sciences, Beijing, China.
Journal of medical Internet research
|November 15, 2023
概括
机器学习模型显著改善了对危及生命的心室失常症 (LTVAs) 患者的住院死亡率预测. 这些先进的模型优于传统的评分系统,可以更好地识别高风险个体,从而改善心脏护理.
科学领域:
- 心脏病学 心脏病学
- 医疗信息学 医疗信息学
- 医疗保健中的机器学习
背景情况:
- 危及生命的心室节律失常 (LTVAs) 是突然心脏骤停的主要原因.
- 需要有效的预测模型来识别患有LTVA的高风险个体.
研究的目的:
- 开发和评估基于机器学习 (ML) 的模型,用于预测被诊断为LTVA的患者的住院死亡率.
- 将ML模型的性能与传统评分系统进行比较.
主要方法:
- 使用了3140名LTVA患者的数据集,随机分为培训和内部验证集.
- 对2851名患者的单独队列进行了外部验证.
- 将5ML算法与SAPS-II和LODS评分系统进行比较,使用接收器操作特征曲线 (AUC) 下的面积.
主要成果:
- 与传统的评分系统相比,ML模型的预测性能明显优越.
- CatBoost获得了最高的AUC (90.5%),紧随其后的是LightGBM (90.1%).
- 传统模型 (SAPS-II和LODS) 的预测值不令人满意 (AUC分别为78.0%和74.9%).
结论:
- 基于ML的模型为LTVA患者的住院死亡率提供了更高的预测准确度.
- 这些发现表明,在临界心脏状况下,转向ML以实现更有效的风险分层.
相关概念视频
01:30Dysrhythmias V: Evaluating Dysrhythmias
40
Dysrhythmias, also known as arrhythmias, are disturbances in the heart's rhythm that range from benign to life-threatening. A thorough evaluation is crucial for appropriate management and involves a comprehensive medical history, physical examination, and various diagnostic tests.Medical HistorySymptoms: Collect detailed information on palpitations, dizziness, syncope, chest pain, and fatigue. Note their onset, frequency, and triggers.Previous Cardiac Issues: Document any history of heart...
40
01:29Cardiomyopathy V: Interprofessional Care
14
Managing cardiomyopathy involves addressing underlying or precipitating causes, treating heart failure with medications, and implementing dietary changes and a balanced exercise and rest regimen.Lifestyle ModificationsCardiomyopathy patients should adopt a low-sodium diet to reduce fluid retention and manage heart failure. A personalized exercise and rest plan helps maintain physical fitness without overstraining the heart. Avoiding alcohol and tobacco is essential to prevent further damage to...
14