SHM Converge Scientific Abstract ePoster Gallery-Predicting Return of Spontaneous Circulation and Hospital Discharge After in-Hospital Cardiac Arrest Using Machine Learning

Pdf Summary

This retrospective pilot study examined whether machine learning models could predict return of spontaneous circulation (ROSC) and live hospital discharge (LHD) after in-hospital cardiac arrest (IHCA). The cohort included 322 IHCA cases from Staten Island University Hospital between 2018 and 2024, with patients required to have blood work within 24 hours before arrest. More than 50 clinical, laboratory, and echocardiographic variables were analyzed.<br /><br />Four supervised machine learning models were trained: logistic regression, decision tree, random forest, and XGBoost. Model interpretability was evaluated using SHAP values to identify the most important predictors. The study found that several consistent predictors emerged across models for both ROSC and LHD, although exact performance metrics were not highlighted in the text provided.<br /><br />The patient population had substantial comorbidity burden, including hypertension, dyslipidemia, diabetes, COPD, coronary artery disease, chronic kidney disease/end-stage renal disease, malignancy, heart failure, and valvular disease. Laboratory abnormalities were common, with elevated white blood cell count, creatinine, BUN, and AST among the recorded baseline features.<br /><br />The authors conclude that this pilot work demonstrates the feasibility of using machine learning to predict ROSC and LHD after IHCA. They emphasize that, despite the modest sample size, the models identified meaningful and consistent predictors. The study provides a foundation for future research using larger multicenter datasets to improve predictive accuracy and generalizability, with the goal of supporting clinical decision-making during cardiac arrest care.

Asset Subtitle

Elie Bou Sanayeh

Meta Tag

Author List Aysan Sattarzadeh, Elie Bou Sanayeh, Elie Moussa, Georges Khattar, Hadi Itani, Kaiyu Jia, Martin Miguel Amor, Omar Sadaqah, Quacy Burke, Suzanne El Sayegh, Yiela Saperstein

Category Research

Concept In-hospital cardiac arrest

Concept Post-ROSC care

Concept Live Hospital Discharge

Concept Supervised Machine Learning

Concept Predictive Model

Distinguished Non-Finalist

Presenter Organization Staten Island University Hospital, Northwell

Presenting Author Elie Bou Sanayeh

Track Outcomes Research

Keywords

in-hospital cardiac arrest

return of spontaneous circulation

live hospital discharge

machine learning

predictive modeling

supervised learning

SHAP values

logistic regression

random forest

XGBoost

In-hospital cardiac arrest

Post-ROSC care

Live Hospital Discharge

Supervised Machine Learning

Predictive Model