Scholarlink

<p>To enhance ICU Length of Stay (LoS) prediction for Heart Failure (HF) patients and unearth valuable clinical insights, this study introduces an underlying system – LoS Prediction and Analysis System (LoS-PAS). It utilizes a Temporal Comorbidity Network (TCN) to analyze patient heterogeneity and comorbidity progression over time. A proposed deep learning model integrates TCN insights, including the novel normalized LoS propensity and comorbidity propensity vector features, to predict ICU LoS in current visit. LoS-PAS provides insights into the driving factors and analyzes comorbidity patterns across age- and gender-stratified subgroups, identifying latent focal diseases and their LoS impact. Validated with real multi-center datasets, LoS-PAS enhances prediction accuracy, offers insights into disease progression, and incorporates visualization tools to assist in clinical decision-making. The results demonstrate that LoS-PAS provides evidence-based, significantly bolstering the clinical decision support system capabilities in predictive analytics and patient care optimization.</p>

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

How Long do the Heart Failure Inpatients Stay in ICU? Temporal Comorbidity Networks to Assist Clinical Decision Support Systems

  • 期刊论文

  • 发表日期：

    2025年10月29日

查看全文
  • Ye Liang,
  • Yalong Cao,
  • Chonghui Guo,
  • Hailin Li

摘要

To enhance ICU Length of Stay (LoS) prediction for Heart Failure (HF) patients and unearth valuable clinical insights, this study introduces an underlying system – LoS Prediction and Analysis System (LoS-PAS). It utilizes a Temporal Comorbidity Network (TCN) to analyze patient heterogeneity and comorbidity progression over time. A proposed deep learning model integrates TCN insights, including the novel normalized LoS propensity and comorbidity propensity vector features, to predict ICU LoS in current visit. LoS-PAS provides insights into the driving factors and analyzes comorbidity patterns across age- and gender-stratified subgroups, identifying latent focal diseases and their LoS impact. Validated with real multi-center datasets, LoS-PAS enhances prediction accuracy, offers insights into disease progression, and incorporates visualization tools to assist in clinical decision-making. The results demonstrate that LoS-PAS provides evidence-based, significantly bolstering the clinical decision support system capabilities in predictive analytics and patient care optimization.