时间序列深度学习模型实时动态预测心脏术后死亡与肾脏替代治疗
该研究利用MIMIC数据库中2008至2019年接受冠状动脉旁路移植术(单独或联合瓣膜手术)的ICU患者数据,开发基于循环神经网络的时间序列深度学习模型(DLM),动态预测院内死亡和连续性肾脏替代治疗(CRRT)需求。7402例患者中,院内死亡1.3%,需CRRT者1.8%;模型在死亡和CRRT预测上的精确率-召回率曲线下面积分别为0.877和0.906,F1分数分别为0.808和0.757。模型可每小时更新预测,自注意力层突出关键时间点,提升临床可解释性。
要点
- 基于每小时ICU数据的DLM可动态预测心脏术后死亡和CRRT,性能较高。
- 随着生理数据积累,模型预测性能逐步提升。
- 自注意力层可识别驱动预测的关键时间点,增强可解释性。
核心结论
- 时间序列深度学习模型可每小时连续更新心脏术后ICU患者的死亡和CRRT预测,性能高且具可解释性。
- 该模型有望在病情明显恶化前几天识别高危患者,促进早期干预。
研究设计与样本
- 回顾性研究,使用MIMIC数据库中2008至2019年接受冠状动脉旁路移植术(单独或联合瓣膜手术)的ICU患者数据。
- 数据按70:20:10分为训练集、验证集和测试集,开发循环神经网络模型,使用固定变量和动态变量预测院内死亡和CRRT需求。
主要发现
- 共纳入7402例患者,院内死亡1.3%,需CRRT者1.8%。
- 使用完整ICU序列时,死亡预测的精确率-召回率曲线下面积为0.877,F1分数0.808;CRRT预测分别为0.906和0.757。
- 模拟实时部署每小时更新预测,随着生理数据积累模型性能提升。
- 自注意力层突出驱动预测的关键时间点,提供临床可解释性。
临床怎么用
- 可利用该模型对心脏术后ICU患者进行动态风险评估,早期识别高危患者并提前干预。
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
原文
OBJECTIVE: Predicting postoperative deterioration following cardiac surgery remains challenging. Conventional risk scores rely on static variables and fail to capture evolving physiologic trajectories. We developed a time-series deep learning model (DLM) using serial intensive care unit (ICU) measurements to dynamically predict mortality and continuous renal replacement therapy (CRRT) following cardiac surgery. METHODS: Using the Medical Information Mart for Intensive Care database, we analyzed ICU admissions of patients undergoing coronary artery bypass grafting, isolated or combined with valve surgery, between 2008 and 2019. Data were split into 70:20:10 for training, validation, and testing. Recurrent Neural Network models were developed to predict in-hospital mortality and CRRT requirement using fixed and dynamic variables. Fixed variables included demographics, comorbidities, and procedure type. Dynamic variables comprised hourly ICU data (vitals, ventilation, vasopressors, and labs). Risk was updated at each timestep based on the most recent measurements, and a self-attention layer highlighted influential timepoints. RESULTS: Of the 7402 patients included, 1.3% died in-hospital and 1.8% required CRRT. Utilizing full ICU sequences, the DLMs achieved an area under the precision-recall curve of 0.877 and 0.906, and F1 scores of 0.808 and 0.757 in mortality and CRRT prediction, respectively. Simulating real-world deployment with hourly, updated predictions, model performance improved as physiologic data were accumulated. Furthermore, the self-attention layer highlighted critical timepoints that drive predictions, offering valuable clinical interpretability. CONCLUSIONS: This is the first time-series DLM to provide continuously updated, hourly predictions of both mortality and CRRT for cardiac surgery ICU patients, demonstrating high performance and interpretability. By reliably identifying high-risk patients several days before overt deterioration, this approach may facilitate earlier, proactive clinical intervention.
引用
- El Moheb M, Bitar ER, Putman K, et al. Real-time dynamic prediction of mortality and renal replacement therapy after cardiac surgery using a time-series deep learning model. J Thorac Cardiovasc Surg. 2026. PMID: 42668112. doi:10.1016/j.jtcvs.2026.08.015
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。