New York City-based Mount Sinai Health System found that artificial intelligence can help emergency department teams identify which patients are likely to need hospital admission hours earlier than current methods.
For the two-month, seven-hospital study, researchers compared predictions from a machine learning model — trained on more than 1 million prior visits — with triage assessments from more than 500 ED nurses, according to an Aug. 11 news release. Nearly 50,000 patient visits were included.
The AI model alone accurately forecasted admissions across urban and suburban settings, with no significant accuracy boost when combined with human predictions. Researchers said earlier forecasts could reduce ED overcrowding and boarding, improve patient flow and allow better resource allocation.
Mount Sinai plans to test the model in real-time workflows, measuring effects on boarding times, throughput and operational efficiency.
The study, titled “Comparing Machine Learning and Nurse Predictions for Hospital Admissions in a Multisite Emergency Care System,” was published July 9 in Mayo Clinic Proceedings: Digital Health.
