Summary of article from Association of Health Care Journalists, by Karen Blum:
Two recent studies highlight the potential of GPT-4, an AI model by OpenAI, to assist in emergency department triage and hospital admission predictions. Researchers at UCSF found that GPT-4 could identify patients with more severe conditions with 89% accuracy, slightly outperforming physicians. A second study by the Icahn School of Medicine at Mount Sinai demonstrated GPT-4’s ability to predict hospital admissions with up to 83% accuracy. These findings suggest AI could help streamline emergency care, though further validation and efforts to mitigate biases are needed. Despite the promise, doctors will still need to independently assess patients to ensure accurate treatment.