564. Temporal External Validation of a Machine Learning Model for Sepsis Risk Prediction in Burn Patients

564. 机器学习模型在烧伤患者脓毒症风险预测中的时间外部验证

阅读:1

Abstract

INTRODUCTION: Sepsis remains a leading cause of mortality in burn patients, where the persistent post-burn hypermetabolic state may obscure early clinical signs and limit opportunities for timely intervention. To address this challenge, we previously developed a streamlined machine-learning model for sepsis prediction based on six admission-level variables (age, burned body surface area, deep partial-thickness burns, full-thickness burns, inhalation injury, and hypertension). The objective of this study was to perform a temporal external validation of this model in an independent patient cohort to confirm its robustness before clinical implementation. METHODS: A retrospective analysis was performed using data from a national multicenter burn registry. The previously established admission-level model was applied to a new cohort of burn patients admitted after the original development period. Model performance was assessed for discrimination (AUC), sensitivity, specificity, and predictive values. A web-based calculator was subsequently developed to support bedside implementation. RESULTS: In the validation cohort, the score preserved strong predictive performance with metrics closely aligned to the derivation study: AUC 0.91, sensitivity 0.81, specificity 0.85, and negative predictive value 0.98. Validation confirmed not only preserved discrimination but also reproducible patterns of feature interdependence, including non-linear effects of burn size and age on sepsis risk. These findings confirm classification accuracy across temporally distinct patient samples. CONCLUSIONS: Temporal external validation in a multicenter national registry confirmed the accuracy and stability of the previously developed sepsis prediction model in burn patients. As the first burn-specific validated predictive tool in this setting, it enables reliable early-warning risk stratification immediately upon admission. APPLICABILITY OF RESEARCH TO PRACTICE: This validated score supports a development from reactive diagnosis to proactive risk management in burn care. The companion web-based calculator provides an accessible, objective tool for immediate bedside risk assessment, enabling intensified surveillance for high-risk patients (e.g., more frequent biomarker assessment, earlier advanced hemodynamic monitoring) while potentially supporting stewardship in those at low risk. FUNDING FOR THE STUDY: N/A.

特别声明

1、本页面内容包含部分的内容是基于公开信息的合理引用;引用内容仅为补充信息,不代表本站立场。

2、若认为本页面引用内容涉及侵权,请及时与本站联系,我们将第一时间处理。

3、其他媒体/个人如需使用本页面原创内容,需注明“来源:[生知库]”并获得授权;使用引用内容的,需自行联系原作者获得许可。

4、投稿及合作请联系:info@biocloudy.com。