Scholarlink

Study design <p>Retrospective cohort study.</p> Objective <p>The primary aim was to develop a machine learning (ML) model to predict rehospitalization during the first year of traumatic spinal cord injury (TSCI) and to identify top predictors using data obtained during initial rehabilitation. The secondary aim was to predict prolonged hospital stay among the rehospitalized group.</p> Setting <p>Eighteen SCI Model Systems centers throughout the United States.</p> Methods <p>Data were retrieved from the National Spinal Cord Injury Model Systems Database. The participants were divided into 2 groups based on rehospitalization during the first year of injury. Those who experienced rehospitalization during first year were further grouped into prolonged stay (&gt;75th quartile of the total length of stay) or non-prolonged stay. Variables considered in models included socio-demographic factors, clinical characteristics, and comorbidities.</p> Results <p>The best performing classification models were Random Forest for predicting rehospitalization and Adaptive Boosting for prolonged stay. The most important predictors in both models were the degree of functional independence, American Spinal Injury Association (ASIA) scores, age, days from injury to rehabilitation admission and body mass index. Additionally, for prolonged stays, pressure injury as a reason for rehospitalization was top predictor.</p> Conclusion <p>Functional Independence Measure (FIM) and ASIA scores emerge as key predictors of both rehospitalizations and prolonged rehospitalizations. These findings may assist clinicians in patient risk assessment. Furthermore, the identification of pressure injury as a primary predictor for prolonged stays signifies a targeted focus on preventive measures for pressure injury-related rehospitalizations, offering a specific strategy to enhance patient care and outcomes.</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

Utilization of machine learning algorithm in the prediction of rehospitalization during one-year post traumatic spinal cord injury

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年2月15日

查看全文
  • Salma Aly,
  • Yuying Chen,
  • Abdulaziz Ahmed,
  • Huacong Wen,
  • Tapan Mehta

摘要

Study design

Retrospective cohort study.

Objective

The primary aim was to develop a machine learning (ML) model to predict rehospitalization during the first year of traumatic spinal cord injury (TSCI) and to identify top predictors using data obtained during initial rehabilitation. The secondary aim was to predict prolonged hospital stay among the rehospitalized group.

Setting

Eighteen SCI Model Systems centers throughout the United States.

Methods

Data were retrieved from the National Spinal Cord Injury Model Systems Database. The participants were divided into 2 groups based on rehospitalization during the first year of injury. Those who experienced rehospitalization during first year were further grouped into prolonged stay (>75th quartile of the total length of stay) or non-prolonged stay. Variables considered in models included socio-demographic factors, clinical characteristics, and comorbidities.

Results

The best performing classification models were Random Forest for predicting rehospitalization and Adaptive Boosting for prolonged stay. The most important predictors in both models were the degree of functional independence, American Spinal Injury Association (ASIA) scores, age, days from injury to rehabilitation admission and body mass index. Additionally, for prolonged stays, pressure injury as a reason for rehospitalization was top predictor.

Conclusion

Functional Independence Measure (FIM) and ASIA scores emerge as key predictors of both rehospitalizations and prolonged rehospitalizations. These findings may assist clinicians in patient risk assessment. Furthermore, the identification of pressure injury as a primary predictor for prolonged stays signifies a targeted focus on preventive measures for pressure injury-related rehospitalizations, offering a specific strategy to enhance patient care and outcomes.