SHM Converge Scientific Abstract ePoster Gallery-Predicting Post - Return of Spontaneous Circulation Outcomes: A Machine Learning Approach to Sedation Analysis

Pdf Summary

This study examined how sedation affects prediction of hospital mortality after return of spontaneous circulation (ROSC) in patients who survived at least one day in the ICU. Using EMR data from 562 post-ROSC patients, the authors built a machine learning model, Bootstrap Forest, to predict mortality and estimate how much each variable contributed to model performance.<br /><br />Patients who survived and those who did not were similar in age and sex distribution. The analysis found that sedation significantly changed the predictive value of several clinical features. Most notably, Glasgow Coma Scale (GCS) was much less informative in sedated patients, explaining 11% of variance in mortality prediction compared with 28% in non-sedated patients. Sedation also altered the importance of temperature, urine output, and creatinine.<br /><br />The model remained accurate in both sedated and non-sedated groups. In sedated patients, kidney-related measures such as eGFR and creatinine, along with urine output, became stronger predictors of mortality, suggesting that systemic and renal status may matter more when neurological assessment is confounded by sedation.<br /><br />Overall, the study shows that sedation can blunt the prognostic value of GCS after ROSC, and that machine learning can help clarify how different factors contribute to mortality risk. The findings support more nuanced interpretation of neurological exam findings in sedated ICU patients and suggest that closer attention to renal function and fluid balance may improve risk assessment and patient care.

Asset Subtitle

John Z. Korin

Meta Tag

Author List Aman Khurana, Caroline N. Wojtas, Gustavo Garcia, John Z. Korin, Karen Hamad, Patricia Riano Rivero, Raja Goyal, Robert A. Smith

Category Research

Concept Sedation

Concept Glasgow Coma Scale

Concept Hospital Mortality

Concept Post-ROSC Patients

Concept Mortality Risk

Distinguished Non-Finalist

Presenter Organization Sarasota Memorial Hospital/Florida State University School of Medicine

Presenting Author John Z. Korin

Track Outcomes Research

Keywords

sedation

return of spontaneous circulation

hospital mortality

machine learning

Bootstrap Forest

Glasgow Coma Scale

ICU

renal function

creatinine

urine output

Sedation

Glasgow Coma Scale

Hospital Mortality

Post-ROSC Patients

Mortality Risk