Artificial intelligence (AI) is making strides in the healthcare field, as a recent study led by UCSF found that AI is as effective as a physician in prioritizing which patients need treatment urgently in an emergency department setting. With emergency departments often overcrowded and overwhelmed, AI could potentially help streamline the triage process and allocate care and resources more efficiently.
In this study, researchers used anonymized records of 251,000 adult emergency department visits to evaluate how well an AI model could extract symptoms from patients’ clinical notes to determine their need for immediate treatment. The AI analysis was compared with the patients’ scores on the Emergency Severity Index, and the AI was found to be highly accurate in identifying which patients needed urgent care. In a sample of 10,000 matched pairs of patients, the AI correctly identified the patient with a serious condition 89% of the time.
The researchers also tested the AI model against the judgment of physicians in a sub-sample of 500 pairs, and found that the AI was correct 88% of the time, compared to 86% for the physician. This suggests that AI could potentially assist healthcare providers in making critical decisions in a fast-paced environment, allowing physicians to focus on treating patients with the most serious conditions.
Lead author Christopher Williams emphasized the potential benefits of AI in helping prioritize patients in emergency situations, such as determining which patient should be transported to the hospital first or which urgent request should be addressed by a physician. However, Williams noted that AI is not yet ready for routine use in the emergency department without further validation and clinical trials.
One important consideration in using AI in healthcare is the potential for bias within the model. Previous research has shown that AI models can perpetuate racial and gender biases in healthcare, which could impact patient care. Williams highlighted the need to address and eliminate bias in AI models before deploying them in clinical settings to ensure equitable and unbiased patient care.
While the study demonstrated the effectiveness of AI in prioritizing emergency department patients, Williams stressed the need for further research and validation before widespread implementation. Future work will focus on how best to deploy AI technology in clinical settings and how to ensure that it benefits all patients, regardless of their background or medical condition. Ultimately, AI has the potential to revolutionize healthcare by assisting clinicians in making critical decisions and improving patient outcomes in emergency situations.
