How Agencies Are Tackling AI for Health Care

Federal health agencies see promise in artificial intelligence (AI) in health care settings to analyze bias in data models, ensure health equity and relieve employee burnout.

Government’s newest agency, the Advanced Research Projects Agency for Health (ARPA-H), is building partnerships nationwide with new programs that aim to be at the forefront of health innovation. It’s looking at moonshot projects that will take 10 to 15 years before there’s any measured progress, but the impacts are huge.

“Take the full eye transplantation [in the THEA program],” said Director Renee Wegrzyn at NVIDIA’s AI Summit in Washington, D.C., earlier this month. “There will be patients in the last phase of that program that hopefully will receive a whole eye transplantation 20 years from now.”

APECx and PRINT are two ARPA-H programs that are looking at using AI to process data associated with current untreatable viruses and generate 3D bio-printed models that can be used to reduce waiting times for organ transplants, noted CTO Gerardo Castaneda.

“We’re a great agency to pilot and test things out,” said Director Renee Wegrzyn at NVIDIA’s AI Summit in Washington, D.C., earlier this month. “We’re spurring that economic innovation with the federal government and with [small firm investors] who haven’t previously been eligible.”

AI to Curb Burnout in Health Care

AI could provide a solution to prevent employee burnout at Defense Department hospitals.

Defense Healthcare Management Systems has petabytes of standardized data dating back to the early 2000s. Its Chief Data Scientist Jesus Caban said the agency wants to train AI and machine-learning predictive models with the data. This would reduce the time care providers spend re-reading clinical notes for medical screenings.

Clinicians can then spend more time focusing on the patient and make better informed decisions based on previous visits. This information could let military employees know if they are medically qualified for a job sooner.

“A significant number of providers spend all day reviewing clinical notes one by one,” said Caban at the event. “Being able to filter 1,000 clinical notes and say, hey, these are the five notes that you should look at; they might have some potential disqualifying conditions for that job. That’s No. 1 for us to finish by the end of this year.”

Assistant Secretary for Technology Policy (ASTP) and National Coordinator for Health IT Micky Tripathi, who also serves as acting chief AI officer at the Department of Health and Human Services, sees a wide range of use cases for AI.

“It’s just a way of compressing that administrative time,” said Tripathi at the event. “In medical settings, for example, and I saw a survey that something like 60 to 70% of doctors are using a lot of foundation models to check for drug interactions. You realize [AI is] not used in just HR.”

Better Data for Health Equity

Tripathi noted that HHS after its reorganization aims to increase AI security and create policies around its ethical use.

Some of the challenges with AI involve issues like scalability and data management. Plus, those without insurance or little coverage are often left out of clinical data sets.

“Data is inherently biased by the health care system that we live in today and has health equity dimensions,” said Tripathi. “That’s all reflected in the data, and AI is unfortunately going to pick that up and make other inferences that increase health equity issues.”

Belinda Seto, deputy director at National Institutes of the Health’s Office of Data Science Strategy, wants to use AI to break down health equity barriers. Seto said focusing on holistic and diverse health data could decrease disparities.

“We have to understand the patient as a whole being… the social determinants of health are very important, as are the environmental determinants of health,” said Seto at the event. “When you take into account this vast disparate array of data, AI is tremendous in being able to consume the data and detect patterns.”

Another challenge is getting diverse, quality data into clinical trials.

Wegrzyn highlighted the NITRO program focused on reversing osteoarthritis. Since osteoarthritis is extremely common in Indigenous communities, the NITRO program partnered with the Cherokee Nation to ensure Indigenous patient data was included.

“You need to know a lot about different demographics of that disease,” said Wegrzyn. “There’s so much data that would go into it to make sure that it really is accessible for everybody.”