Title: The Future of Emergency Care: Artificial Intelligence in Triage
Emergency departments across the country are grappling with overcrowding and increased patient volumes, leading to challenges in prioritizing care. However, a new study suggests that artificial intelligence (AI) could be a game-changer in addressing this issue.
Researchers at UC San Francisco used anonymized records of 251,000 adult emergency department (ED) visits to evaluate the effectiveness of an AI model in extracting symptoms from patients’ clinical notes to determine their urgency of treatment. The AI model, ChatGPT-4 large language model (LLM), was accessed via UCSF’s secure generative AI platform, which has robust privacy protections.
The LLM was able to identify which ED patient in a sample of 10,000 matched pairs had a more serious condition 89% of the time. Moreover, in a sub-sample of 500 pairs, the AI was correct 88% of the time, compared to 86% for physicians.
The use of AI in the triage process has the potential to free up critical physician time to treat patients with the most serious conditions while providing backup decision-making tools for clinicians juggling multiple urgent requests.
“AI has the potential to significantly improve emergency care by accurately identifying patients who require immediate attention,” said lead author Christopher Williams, MB, BChir, a UCSF postdoctoral scholar at the Bakar Computational Health Sciences Institute.
However, Williams emphasized that AI is not yet ready for prime time. Despite the success of this study, further validation and clinical trials are necessary before AI can be used responsibly in the ED.
One of the critical issues to address is eliminating bias from the model. Previous research has shown that these models may perpetuate racial and gender biases in healthcare due to the biases within the data used to train them.
“Upcoming work will address how best to deploy this technology in a clinical setting while being cautious and deliberate in how it is applied,” said Williams.
This study is one of the few to evaluate an LLM using real-world clinical data and the first to use more than 1,000 clinical cases for this purpose. The findings highlight the potential for AI to revolutionize emergency care, provided that the issues of bias and responsible use are addressed.
The future of emergency care is bright, with the potential for AI to enhance patient care and improve the efficiency of triage processes. However, it is crucial to prioritize safety, accuracy, and fairness as we move towards the integration of AI in emergency medicine.
