基于机器学习的高血压危急病患者的死亡率预测:比较分析,公平性和解释性
Shenghan Zhang1, Sirui Ding2, Zidu Xu3
1Department of Biomedical Informatics, Harvard University, Boston, MA, United States.
Frontiers in artificial intelligence
|December 29, 2025
概括
机器学习模型准确地预测了重症高血压患者的死亡率. 特征选择提高了模型的公平性和可解释性,以便更好地进行临床决策.
科学领域:
- 关键护理医学 关键护理医学
- 生物医学信息学是生物医学信息学.
- 医疗保健中的人工智能
背景情况:
- 高血压是心血管,脑血管和脏疾病的主要危险因素,严重病患者的死亡率增加.
- 准确的死亡率预测对于在这种高风险人群中及时进行干预至关重要.
- 机器学习 (ML) 和深度学习 (DL) 为分析电子健康记录 (EHR) 数据提供了先进的工具.
研究的目的:
- 开发和评估ML和DL模型,用于预测高血压患者的住院死亡率.
- 评估这些预测模型的公平性和可解释性.
- 为了利用MIMIC-IV重症监护数据集进行稳健的模型开发.
主要方法:
- 开发了梯度增强机 (GBM),物流回归,支持向量机 (SVM),随机森林,多层感知器 (MLP) 和长短期记忆 (LSTM) 模型.
- 利用了EHR的综合特征,包括人口统计数据,实验室值,生命体征和并发症.
- 评估模型使用5倍交叉验证,SHapley添加式扩展 (SHAP) 进行特征重要性,以及人口平价差异 (DPD) 和平等赔率差异 (EOD) 进行公平性.
主要成果:
- GBM模型实现了最高的性能 (AUC-ROC 96.3%,精度为89.4%).
- 死亡率的关键预测因素包括格拉斯哥昏迷量表 (GCS) 分数,布拉登量表分数,血液尿素,年龄,红细胞分布宽度 (RDW),二碳酸盐和乳酸盐.
- 使用顶级特征的模型显示偏差降低 (较低的DPD和EOD);对所有特征的模型来说,debiasing技术改善了公平性.
结论:
- ML模型显示了预测临床重症高血压患者的死亡率的巨大潜力.
- 特征选择提高了模型的解释性,减少了复杂性,并可能提高公平性.
- 整合可解释和公平的AI工具可以支持临床决策在重症监护.
相关概念视频
01:25Errors occurring during blood pressure monitoring
1.3K
Blood pressure monitoring is a crucial clinical procedure in diagnosing and managing various cardiovascular conditions. Despite its significance, the accuracy of blood pressure measurements can be compromised by multiple factors, potentially leading to either falsely high or low readings. These inaccuracies are critical as they can significantly impact patient care. So, it is vital to understand these challenges deeply and adopt strategic approaches to minimize errors.
Several factors...
Several factors...
1.3K
01:10Pre-Procedural Guidelines for Assessing Blood Pressure
788
Accurate blood pressure assessment is crucial for diagnosing and managing various health conditions. To ensure the reliability of these measurements, healthcare professionals must adhere to standardized pre-procedural guidelines. These guidelines enhance patient safety and improve the overall quality of healthcare. The following steps are essential for obtaining accurate and consistent blood pressure readings, from using the appropriate tools to ensuring effective communication with the...
788
01:20Comparing the Survival Analysis of Two or More Groups
531
Survival analysis is a cornerstone of medical research, used to evaluate the time until an event of interest occurs, such as death, disease recurrence, or recovery. Unlike standard statistical methods, survival analysis is particularly adept at handling censored data—instances where the event has not occurred for some participants by the end of the study or remains unobserved. To address these unique challenges, specialized techniques like the Kaplan-Meier estimator, log-rank test, and...
531
01:30Hypertension III: Clinical Manifestations and Diagnostic Studies
437
Hypertension is asymptomatic and also referred to as the "silent killer" until it progresses to a severe stage or causes target organ disease. Patients may experience symptoms stemming from the strain on blood vessels and tissues in various organs or the heart's increased workload.Physical exams might show no abnormalities other than high blood pressure. Signs of vascular damage, when present, correspond to the organs supplied by the affected vessels, leading to target organ damage. For...
437