Propelling quality improvement with AI
AI at UW
Artificial intelligence is being incorporated into healthcare technologies and practices more and more in order to promote efficiencies, simplify tasks, reduce cost and ultimately improve patient outcomes.
At the UW, several groups are actively investigating new ways to deploy AI for use in improving medical research and healthcare delivery.
The UW Medicine Predictive Analytics Committee is led by Drs. Stephan Fihn, professor, and Andrew White, professor (General Internal Medicine) from the Department of Medicine and Drs. Peter Tarczy-Hornoc and Trevor Cohen from the Department of Biomedical Informatics and Medical Education.
Fihn, a general internist at UW Medicine since 1977 and head of the Division of General Internal Medicine from 1995-2021, and White, a hospital medicine provider and director for the Certificate Program in Patient Safety and Quality, share a passion for quality improvement, evaluation and informatics.
Propelling new innovations
As co-chairs of the Predictive Analytics Committee, they design and deploy advanced predictive models to facilitate evaluations of new innovations. With this approach, researchers are able to utilize machine learning to look at existing data and variables, decipher patterns and then predict statistically likely future outcomes.
The committee will soon be deploying predictive models for five new projects, including one focused on sepsis, which is already in use.
Three of these models are spearheaded by Department of Medicine faculty.
Large language modeling
Large language modeling is a computational model notable for its ability to achieve general-purpose language generation and other natural language processing tasks such as classification.
They are also being charged with implementing large language models (LLM) into Epic, UW Medicine’s electronic health record system. AI systems such as LLMs have the ability to generate information in natural language based on combinations of data inputs.
For use in areas of Epic such as the inbox manager and ambient listening note documentation, deploying LLMs could streamline clinicians' workflows and reduce time spent on administrative tasks. Ultimately, the committee hopes to evaluate the initiative to illustrate these outcomes and others like impact on clinician burnout.
Looking forward
These are just a few examples of applications of the data-processing and predictive capabilities of AI for streamlining processes and improving various aspects of research and quality improvement in healthcare.
The Predictive Analytics Committee and other groups at the UW and UW Medicine are leading the way for improving quality with machine learning.