Researchers at the Mount Sinai Health System evaluations of artificial intelligence (AI) in an emergency care setting. Their machine learning model, trained on more than 1 million patient records, was designed to identify which patients arriving at emergency departments (EDs) are likely to require hospital admission. The aim was to generate these predictions within hours of a patient’s arrival, so that hospitals could begin planning bed assignments and resource allocation sooner.
The study involved more than 500 ED nurses across seven hospitals in the Mount Sinai network, which spans both urban and suburban communities. These predictions were compared with nurses’ own assessments during the triage process. The research team then analyzed the accuracy of the AI model both independently and in combination with nurse assessments.
Results showed that the AI model performed consistently across different hospital sites, despite differences in patient demographics and case types. This indicated that the model itself was already a strong independent predictor of admissions. Its primary benefit would be to alert care teams much earlier in the patient’s stay, enabling faster coordination of services.
【MORE】