There’s a new institute brewing at the University of Washington at the intersection of artificial intelligence, machine learning and healthcare.
The Institute for Medical Data Science launched publicly this week at the 2023 UW Medical Data Science Symposium.
“The UW has very significant strengths that put us in a position to be a leader in the medical data science space,” said the Institute’s interim director Sean Mooney in an interview after the symposium, held Monday and Tuesday.
The UW is a leader in data science, biostatistics, computer science and machine learning and artificial intelligence and the region has a rich milieu of tech companies, noted Mooney. In addition, UW Medicine is a top healthcare institution.
“What we are building here is strong and unique,” Mooney said. It also dovetails with similar efforts at other institutions, such as Stanford’s department of data science within its medical school, he noted. All academic medical centers have teams that perform analytics on their data, but few have institutes devoted to the effort.
The institute is currently supported through a five-year $750,000 funding pool from UW Provost Mark Richards, as well as grants to individual institute members.
The funding pool will be used to support five competitively selected pilot projects at the UW. The institute will seek a permanent director and its website is slated to launch later this year.
Connecting interventions with outcomes
Speakers at the UW event showcased studies incorporating AI into mental health support, evaluating a chatbot that identifies patients with social needs, and more. They also evaluated various methods to extract data from the electronic healthcare record.
Keynote speaker Atul Butte spoke about an effort he leads as director of the Bakar Computational Health Sciences Institute at the University of California, San Francisco and as chief data scientist for the University of California health system.
Under his leadership, the health system has combined its electronic healthcare data across multiple hospitals and clinics into a unified pool.
With anonymized data from millions of patients, researchers can ask questions about the efficacy of various interventions or correlate outcomes with the system’s growing genomic datasets. Already, the program has helped save money by showing the overuse of certain interventions, said Butte.
For instance, the UC data showed that many patients were unnecessarily receiving certain pain medications intravenously instead of in a cheaper pill form. As a result, the entire medical system changed its practice to provide the medication more often as pills, said Butte.
The data also sets up the health system to do comparative effectiveness studies. “We should be testing every use of every drug and device and procedure we do against the standard,” said Butte. And if patients are not benefitting as much as expected, “we should figure out why.”
The UW has also implemented a similar type of system, said Mooney. The new institute is designed in part to leverage its data, he said.
“Health systems worldwide have been establishing electronic healthcare systems that manage healthcare, but they have been underutilized for the purpose of improving healthcare,” said Mooney.
Forging new startups
Mooney is also interested in helping forge connections between industry and the new UW institute.
“There’s tons of opportunities for entrepreneurs in this space,” said Mooney. “We have a good community to do that.”
Mooney is bullish on AI and natural language tools like ChatGPT that can combine with electronic health record data. He also sees opportunities for startups working with wearable devices that can be integrated into the UW Medicine system for purposes like managing chronic health conditions.
Multiple AI startups have emerged with tools to evaluate mammogram data, noted UW Medicine radiology professor Christoph Lee at the symposium. Lee leads a UW effort with other healthcare partners to assess such tools, connecting their readout with patient outcomes. That’s the type of academic-industry collaboration he envisions more of as the new institute grows, said Mooney, who is also a UW biomedical informatics and medical education professor and UW Medicine’s chief research information officer.
The Institute for Medical Data Science is overseen by the UW School of Medicine, the College of Engineering and the School of Public Health.
Other UW faculty members on the institute’s executive team are computer scientist and electrical engineer Shwetak Patel, a serial entrepreneur and co-founder of several startups; mechanical engineering and bioengineering professor Jonathan Liu, co-founder of Seattle microscopy startup Alpenglow Biosciences; biostatistics professor Lurdes Inoue, and UW Medicine radiology chair Dushyant Sahani and chief data officer Peter Tarczy-Hornoch.
