AI and Clinical Practice—the Learning Health System and AI
Nov 15, 2023
auto_awesome
Nigam Shah, Stanford University professor, discusses reshaping medicine with AI. Topics include challenges with Electronic Health Records, training AI models with unbiased data, ethical data contribution, and the impact of technology on healthcare delivery.
Train AI models using trusted medical sources to ensure accuracy.
Address biases in AI models to improve healthcare system and patient care.
Deep dives
The Importance of Physician Involvement in AI Technology in Healthcare
In this podcast episode, Dr. Kirsten Bippens-Domingo interviews Dr. Nick M. Shaw about the role of AI in healthcare. Dr. Shaw emphasizes the need for physicians to be proactive in shaping the integration of AI technology into medicine. Drawing a parallel to the development of electronic health records (EHRs), he highlights the importance of avoiding repeating the mistakes made in the past. Instead of relying solely on off-the-shelf language models, Dr. Shaw suggests training AI models using trusted medical sources, such as textbooks and clinical guidelines, to ensure the accuracy and reliability of the output. He emphasizes the importance of clearly defining goals and evaluating the impact and efficiency of AI augmentation in healthcare workflow.
Addressing Bias and Privacy in Training AI Models
Dr. Shaw acknowledges the possibility of training AI models on biased or flawed data. He suggests that biases revealed by these models should motivate us to address and rectify those biases in the healthcare system. He emphasizes the importance of intentional content selection to minimize biases during model creation. Additionally, Dr. Shaw highlights the significance of policies and guidelines to govern the output of AI models, ensuring that clinical practices aligned with evidence-based medicine are prioritized over biased recommendations. Regarding patient data privacy, he differentiates between privacy and security concerns, advocating for the responsible use of de-identified aggregate patient data to achieve learning health systems while protecting individual privacy.
Prioritizing Intentional Design and Evaluation of AI Augmentation
Dr. Shaw stresses the need to carefully design and evaluate AI augmentation in healthcare, considering both its impact on reducing clinician burden and ensuring quality patient care. He notes that augmentation should not merely shift cognitive burden onto clinicians by requiring them to double-check AI output. Instead, the anatomy of augmentation should focus on achieving productivity gains and improved outcomes for patients. Dr. Shaw emphasizes the importance of clear articulation of goals, while evaluating and verifying the benefit of AI technologies through intentional experiments spanning the entire process from model creation to deployment at scale. He calls for more studies that assess the impact of AI technology on healthcare productivity and its alignment with patient care goals.
In this Q&A, JAMA Editor in Chief Kirsten Bibbins-Domingo, PhD, MD, MAS, interviews Nigam Shah, MBBS, PhD, professor of medicine at Stanford University and chief data scientist at Stanford Health Care, to discuss how large language models are reshaping medicine and the potential pitfalls of automation. Related Content: