文献速递重症综合管理与镇静镇痛

多模型融合预测ICU患者28天死亡率:一项包含亚组验证与临床风险分层的综合回顾性队列研究

Multi-Model Fusion for 28-Day Mortality Prediction in ICU Patients: A Comprehensive Retrospective Cohort Study with Subgroup Validation and Clinical Risk Stratification.
PubMed 文献 · Shock · 2026-01-27
一句话结论多模型融合框架预测ICU患者28天死亡率AUC-ROC为0.862,优于逻辑回归、随机森林和梯度提升模型,并可将患者分为三个临床风险组。

该回顾性队列研究纳入某三级学术医疗中心2022年8月至2024年8月连续收治的654例成人ICU患者,开发并验证了逻辑回归、随机森林、梯度提升和多模型融合四种预测方法。多模型融合框架的AUC-ROC为0.862(95%CI 0.821-0.903),显著优于所有单一模型,并在各亚组中表现一致(AUC-ROC范围0.815-0.875)。风险分层界定三个死亡率类别:低风险(概率<0.3,观察死亡率5.6%)、中风险(0.3-0.7,31.7%)和高风险(>0.7,83.3%)。作者主张将该框架整合到ICU临床决策支持系统中。

要点

  • 多模型融合预测ICU患者28天死亡的AUC-ROC为0.862(95%CI 0.821-0.903),优于逻辑回归(ΔAUC=0.121,P<0.001)、随机森林(ΔAUC=0.034,P=0.018)和梯度提升(ΔAUC=0.027,P=0.032)。
  • 多模型融合的敏感度为84.6%、特异度为81.0%,Brier评分为0.140,Hosmer-Lemeshow检验卡方=6.28、P=0.616。
  • 风险分层:低风险(概率<0.3)观察死亡率5.6%,中风险(0.3-0.7)31.7%,高风险(>0.7)83.3%。
  • SOFA评分、APACHE II评分和GCS在所有特征重要性方法中均为稳健预测因子。
  • 多模型融合在各亚组中表现一致,AUC-ROC范围0.815-0.875,涵盖不同年龄层、疾病严重程度和干预需求。

适用人群与场景

  • 某三级学术医疗中心2022年8月至2024年8月连续收治的成人ICU患者,共654例。

诊断标准与关键阈值

  • 风险分层界值:低风险为预测概率<0.3,中风险为0.3-0.7,高风险为>0.7。

研究设计与关键结果

  • 研究类型为回顾性队列研究,样本量654例,单中心,人群为成人ICU患者。
  • 多模型融合框架采用 constituent 模型概率平均,AUC-ROC为0.862(95%CI 0.821-0.903),显著优于逻辑回归(ΔAUC=0.121,P<0.001)、随机森林(ΔAUC=0.034,P=0.018)和梯度提升(ΔAUC=0.027,P=0.032)。
  • 多模型融合敏感度84.6%、特异度81.0%,Brier评分0.140,Hosmer-Lemeshow检验卡方=6.28、P=0.616。
  • 各亚组AUC-ROC范围0.815-0.875,涵盖不同年龄层、疾病严重程度和干预需求。
  • SOFA评分、APACHE II评分和GCS在所有特征重要性方法中均为稳健预测因子。
  • 风险分层:低风险(<0.3)观察死亡率5.6%,中风险(0.3-0.7)31.7%,高风险(>0.7)83.3%。

证据类型、等级与局限

  • 证据类型为回顾性队列研究。
  • 局限:原文未报告局限性。

对床旁实践的改变

  • 作者主张将多模型融合框架整合到ICU临床决策支持系统中,以提高预后精确性并指导个体化治疗策略。
  • 多模型融合框架可用于ICU成人患者28天死亡风险预测,并依据预测概率<0.3、0.3-0.7、>0.7分为低、中、高风险组。

提炼说明

  • 本条依据论文摘要整理;摘要未涉及「处置流程与药物剂量」「红旗征与不能漏的情况」「指南或共识出处」,故未列出这些栏目;全文请点「阅读原文」。
英文原文(摘要)· 点击展开

BACKGROUND: Accurate prediction of 28-day mortality in intensive care unit (ICU) patients represents a critical challenge in modern critical care medicine, with profound implications for clinical decision-making, resource optimization, and patient-centered care. Traditional severity scoring systems and conventional machine learning approaches have consistently demonstrated limitations in discriminative performance, calibration accuracy, and generalizability across heterogeneous patient populations.

METHODS: This retrospective cohort study employed comprehensive data from 654 consecutively admitted adult ICU patients at a tertiary academic medical center between August 2022 and August 2024. We developed and rigorously validated four distinct prediction methodologies: Logistic regression, random forest, gradient boosting, and an innovative multi-model fusion (MMF) framework incorporating probability averaging from constituent models. Our evaluation framework encompassed exhaustive discrimination metrics [area under the receiver operating characteristic curve (AUC-ROC), sensitivity, specificity, F1-score], sophisticated calibration assessment (Brier score, calibration curves, Hosmer-Lemeshow test), detailed subgroup analyses across clinically relevant patient strata, and multidimensional feature importance evaluation.

RESULTS: The MMF paradigm demonstrated statistically superior performance with an AUC-ROC of 0.862 (95% CI 0.821-0.903), significantly outperforming all individual models (logistic regression: ΔAUC = 0.121, P < 0.001; random forest: ΔAUC = 0.034, P = 0.018; gradient boosting: ΔAUC = 0.027, P = 0.032). The fusion model achieved an optimal equilibrium between sensitivity (84.6%) and specificity (81.0%) while maintaining exceptional calibration characteristics (Brier score: 0.140; Hosmer-Lemeshow test: chi-square = 6.28, P = 0.616). Remarkable performance consistency was observed across all patient subgroups (AUC-ROC range: 0.815-0.875), encompassing diverse age strata, disease severity spectra, and intervention requirements. Sequential Organ Failure Assessment score, Acute Physiology and Chronic Health Evaluation II score, and Glasgow Coma Scale emerged as consistently robust predictors across all feature importance methodologies. Clinically meaningful risk stratification delineated three distinct mortality categories: low-risk (<0.3 probability, 5.6% observed mortality), moderate-risk (0.3-0.7, 31.7%), and high-risk (>0.7, 83.3%).

CONCLUSION: The MMF framework establishes a clinically actionable and methodologically sophisticated paradigm for 28-day mortality prediction in ICU patients, effectively addressing fundamental limitations of conventional approaches through enhanced discriminative performance, exemplary calibration accuracy, and consistent generalizability across diverse patient populations. These compelling findings strongly advocate for its integration into ICU clinical decision support ecosystems to advance prognostic precision and guide personalized therapeutic strategies.

来源:Shock,查看原文。版权归原作者所有。

参考来源

  1. Long Z, Tong S. Multi-Model Fusion for 28-Day Mortality Prediction in ICU Patients: A Comprehensive Retrospective Cohort Study with Subgroup Validation and Clinical Risk Stratification. Shock. 2026. PMID: 41670964. doi:10.1097/SHK.0000000000002800

供执业医师抢救与诊疗参考,不替代临床判断;剂量请结合体重、肝肾功能和本院药品规格复核,以指南原文为准。