Title

AWS re:Invent 2023 - Improving patient outcomes using generative AI in healthcare (HLC204)

Summary

Rod Tarago, a former pediatric critical care physician, now leads Clinical Informatics for Academic Medicine at AWS.
Tarago shared a story highlighting the issue of preventable medication errors in healthcare.
He referenced the "To Err Is Human" report and recent statistics showing that patient safety errors affect 25% of inpatient hospitalizations in the U.S.
The U.S. healthcare system is expensive with relatively poor outcomes compared to other countries.
Generative AI offers new opportunities to address these issues, but it must be done carefully.
Chad Vandenberg from UC San Diego Health discussed their journey with AWS, focusing on operational integration and impact on the quadruple aim.
Vandenberg shared examples of AI applications in detecting pneumonia, managing sepsis, and reducing inbox fatigue for healthcare providers.
He emphasized the potential of large language models to improve quality and patient safety by synthesizing unstructured data in EHRs.
John Apiz, a Senior Solutions Architect at UCSD Health, explained the technical architecture of their AI solutions, including data extraction, storage, and processing with AWS services.
The goal is to create intuitive interfaces for healthcare professionals to identify and mitigate patient risks efficiently.

The healthcare industry is still grappling with significant patient safety issues, with a high prevalence of preventable errors.
Generative AI is being explored as a means to improve patient outcomes, reduce costs, and enhance the experiences of both patients and clinicians.
UC San Diego Health's partnership with AWS is an example of how healthcare institutions are leveraging AI to address specific challenges such as sepsis detection and provider inbox management.
The integration of AI into clinical workflows is crucial for its adoption and effectiveness, as seen in UC San Diego Health's approach.
Large language models have the potential to revolutionize quality and patient safety by enabling the analysis of unstructured data in EHRs, which currently makes up the majority of patient records.
The technical architecture presented by John Apiz showcases the complexity and potential of AI applications in healthcare, utilizing a range of AWS services for data management and processing.
The ultimate goal of these AI initiatives is to proactively identify and prevent patient safety incidents, moving from a reactive to a preventative approach in healthcare.