Perceptions of Triage and Care for Critically Ill Patients in the ED: A national survey of emergency physicians

Physician logs of Medical ICU admission requests, which included clinical presentation and triage decisions, were prospectively collected over one year at an academic tertiary care center. Patient characteristics, University HealthSystems Consortium (UHC) expected mortality risk, and hospital courses were also captured. Utilizing machine learning techniques, we trained and evaluated an NLP-based fact extraction computer model to analyze data and recognize patterns within free-text descriptions stratified by the ICU admission decision (Accept vs. Reject). We examined triage decisions and patient outcomes of in-hospital mortality using Chi-square testing and multivariate regression.

View Publication