Scholarlink

Background: Intracerebral hemorrhage (ICH) is a severe and potentially life-threatening condition characterized by bleeding within the brain parenchyma. ICU readmission among ICH patients is a critical outcome, reflecting both clinical severity and resource utilization. To accurately predict ICU readmission risk, machine-learning methods should be applied, as they offer advanced capabilities for analyzing complex clinical data and optimizing decision-making in intensive care settings. Methods: Data from the MIMIC-III and MIMIC-IV databases were analyzed to identify ICH patients. Preprocessing steps, including data imputation and resampling, were implemented to enhance data quality. Feature selection focused on clinical, laboratory, and demographic variables, guided by literature review and expert input. An Artificial Neural Network (ANN) model was developed to predict ICU readmission risk, with performance primarily assessed using AUROC as the chief metric, alongside accuracy, sensitivity, and specificity. Results: The ANN model achieved robust predictive performance, with the highest AUROC of 0.899 (95% CI: 0.860–0.911), representing a 14.3% improvement over the AUROC of 0.736 (95% CI: 0.668–0.801) reported by Miao et al. Key predictors included demographic details, clinical parameters, and laboratory measurements. Conclusion: This study introduces an ANN-based framework for predicting ICU readmission risk in ICH patients, achieving robust predictive performance with potential clinical utility. Despite its success, the model’s generalizability remains limited due to the lack of multi-institutional validation and its focus on specific clinical outcomes. Future research should expand validation efforts and explore integration with real-time patient monitoring to refine predictions and broaden its application.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Machine Learning-Based Prediction of ICU Readmissions in Intracerebral Hemorrhage Patients: Insights from the MIMIC Databases

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Shuheng Chen,
  • Junyi Fan,
  • Armin Abdollahi,
  • Negin Ashrafi,
  • Kamiar Alaei,
  • Greg Placencia,
  • Maryam Pishgar

摘要

Background: Intracerebral hemorrhage (ICH) is a severe and potentially life-threatening condition characterized by bleeding within the brain parenchyma. ICU readmission among ICH patients is a critical outcome, reflecting both clinical severity and resource utilization. To accurately predict ICU readmission risk, machine-learning methods should be applied, as they offer advanced capabilities for analyzing complex clinical data and optimizing decision-making in intensive care settings. Methods: Data from the MIMIC-III and MIMIC-IV databases were analyzed to identify ICH patients. Preprocessing steps, including data imputation and resampling, were implemented to enhance data quality. Feature selection focused on clinical, laboratory, and demographic variables, guided by literature review and expert input. An Artificial Neural Network (ANN) model was developed to predict ICU readmission risk, with performance primarily assessed using AUROC as the chief metric, alongside accuracy, sensitivity, and specificity. Results: The ANN model achieved robust predictive performance, with the highest AUROC of 0.899 (95% CI: 0.860–0.911), representing a 14.3% improvement over the AUROC of 0.736 (95% CI: 0.668–0.801) reported by Miao et al. Key predictors included demographic details, clinical parameters, and laboratory measurements. Conclusion: This study introduces an ANN-based framework for predicting ICU readmission risk in ICH patients, achieving robust predictive performance with potential clinical utility. Despite its success, the model’s generalizability remains limited due to the lack of multi-institutional validation and its focus on specific clinical outcomes. Future research should expand validation efforts and explore integration with real-time patient monitoring to refine predictions and broaden its application.