| 摘 要: 为了有效预测重症监护室脓毒症患者的死亡风险并分析影响结局的因素,建立了脓毒症患者死亡风险预测模型,为脓毒症患者的早期预防和死亡风险控制提供科学的参考依据。本研究以重症监护医学信息市场数据库作为数据来源,从中挑选符合要求的病患,使用贝叶斯网络模型训练相关特征预测脓毒症患者的死亡风险。纳入2 352例脓毒症患者,以患者是否死亡作为最终结局建立模型,模型的风险预测准确率为78.7%,优于逻辑回归模型(72.3%)和决策树模型(71.0%)。贝叶斯网络模型相较于其他模型具有更高的信服力,能够准确预测脓毒症患者的死亡风险,模型的可解释性能够辅助医护人员进行临床决策,同时能够更加合理、科学地分配医疗资源。 |
| 关键词: 脓毒症;贝叶斯网络;重症监护室;死亡预测 |
|
中图分类号: TP391
文献标识码: A
|
|
| Mortality Risk Prediction for Patients with Sepsis in Intensive Care Unit Based on Interpretable Machine Learning |
|
LIU Kun1, LING Chen2, SHI Xiaoqiang1, ZHOU Mengyu1, XU Naiyue1
|
(1.School of Health Science and Engineering, University of Shanghai f or Science and Technology, Shanghai 200093, China;
2. Medical Instrumentation College, Shanghai University of Medicine and Health Sciences, Shanghai 201318, China)
lkun11111@163.com; lingc@sumhs.edu.cn; 19117200027@163.com; zhou_meng_yu_66@163.com; xunaiyue21@163.com
|
| Abstract: To effectively predict the mortality risk of sepsis patients in the Intensive Care Unit (ICU) and analyze the factors affecting outcomes, this paper proposes a mortality risk prediction model for sepsis patients to provide scientific reference for early prevention and mortality risk control of sepsis patients. In this study, the MIMIC-Ⅲ (Medical Information Mart for Intensive Care Ⅲ) database is used as the data source, from which patients who meet the requirements are selected. Bayesian network model is used to train relevant features to predict the mortality risk of sepsis patients. With a sample size of 2 352 sepsis patients, the model, using patient mortality as the ultimate outcome, demonstrates a risk prediction accuracy of 78.7% , surpassing the logistic regression model (72.3% ) and the Decision Tree model (71.0% ). The Bayesian network model, compared to other models, exhibits higher credibility, accurately predicting the mortality risk of sepsis patients. The interpretability of the model can assist medical staff in clinical decision-making and realize a more rational and scientific allocation of medical resources. |
| Keywords: sepsis; Bayesian network; ICU; mortality prediction |
