0
0
Stanford Medicine researchers develop AI to analyze doctors’ notes in electronic medical records, improving care quality and patient outcomes.
Stanford Medicine researchers have developed a cutting-edge artificial intelligence (AI) tool that can efficiently read and analyze thousands of doctors’ notes in electronic medical records, providing valuable insights to improve patient care. The study, published online on December 19 in Pediatrics, highlights how AI could revolutionize the way healthcare professionals manage and evaluate patient information.
Traditionally, physicians and researchers have to painstakingly review hundreds or even thousands of medical charts to gather important data. However, this study demonstrates that large language models—AI tools capable of detecting patterns in complex written language—can automate this process, saving significant time while offering crucial insights for improving medical practices. In particular, the AI tool could be used to monitor medical charts for hazardous drug interactions or help doctors predict which patients will respond positively to certain treatments.
The focus of this study was to investigate whether AI could help doctors ensure children with Attention Deficit Hyperactivity Disorder (ADHD) receive proper follow-up care after being prescribed medication. ADHD medications can have disruptive side effects, such as appetite suppression, making follow-up care essential. The AI model trained on the medical records of over 1,200 children aged 6-11 years successfully detected whether patients or their parents had been asked about side effects within the first three months of medication use.
According to Dr. Yair Bannett, the study’s lead author and assistant professor of pediatrics, the AI model revealed “gaps in ADHD management” by identifying areas where follow-up care might have been overlooked. “This model enables us to identify some gaps in ADHD management,” Bannett said, suggesting that AI could be applied to improve many aspects of medical care.
The AI tool was trained on a dataset of 501 manually reviewed notes to recognize whether physicians inquired about side effects during follow-up visits. The tool then analyzed a total of 15,628 notes, a task that would have taken more than seven months of full-time work without AI. The results showed that some pediatric practices were more diligent than others in following up on medication side effects.
The tool also uncovered that pediatricians were less likely to ask follow-up questions regarding non-stimulant medications, such as anti-anxiety drugs, compared to stimulants commonly prescribed for ADHD. Bannett noted, however, that while AI could identify patterns in the records, it could not explain the underlying reasons for these trends.
Although the AI tool was highly accurate, some challenges remained. It missed follow-up inquiries that were not recorded in the patients’ electronic medical records, particularly for those receiving specialty care. Additionally, the tool occasionally misclassified notes regarding side effects of other medications.
Despite these limitations, the research underscores the potential of AI to aid healthcare professionals in making more informed decisions and improving patient outcomes. AI can process large datasets far more efficiently than humans, offering doctors insights from vast populations of patients that they could not otherwise access.
However, Bannett emphasized the importance of considering ethical concerns when implementing AI in healthcare. He pointed out that AI models, which are trained on existing healthcare data, may reflect the biases and disparities inherent in current medical systems. To mitigate these issues, researchers must ensure that AI tools are developed and used with caution.
The study was supported by the Stanford Maternal and Child Health Research Institute and the National Institute of Mental Health.
For further reading, the study is available in the December 2024 issue of Pediatrics, titled “Applying Large Language Models to Assess Quality of Care: Monitoring ADHD Medication Side Effects” (DOI: 10.1542/peds.2024-067223).
