At the UF College of Medicine, physician-scientists are leveraging artificial intelligence to model this expansive clinician-clinician network, creating a robust patient safety graph capturing the almost infinite interactions between caregivers to change how we think about hospital operations and pioneer new ways of delivering quality care.  

The patient safety graph is one of the many ways the College of Medicine is turning to AI to improve patient safety under the AI-QI initiative, a series of programs designed to make a tangible impact in the realm of patient care by extending AI research into quality improvement activities at UF Health.

Patrick Tighe, MD ’05, MS, is the associate dean for AI application and innovation at the College of Medicine and the project champion for the AI-QI initiative under the college’s strategic plan. We sat down with Tighe to discuss these groundbreaking AI-QI programs and how they can help physicians enhance quality in health care. 

Q: What is the driving force behind the college’s AI-QI initiative?

Tighe: For over a decade, UF and the College of Medicine have led the way in developing AI advances in health care in collaboration with so many in engineering, pharmacy and psychology. As that technology and science matures, we have been asking, “What do these advances look like when we apply them in a clinical setting? How does it advance care for our patients?” The AI-QI initiative allows us to pivot from the classic objectives of expanding knowledge to translating these findings to clinical stakeholders and their individual workflows.

Q: One of the ways you are using AI is to model hospital operations and hidden collaboration points between caregivers. Could you tell us more about this?

Tighe: One of the first opportunities we saw to better quantify our clinical processes and how we take care of patients was by modeling health care as a complex system. To do this, we developed the clinician-clinician patient safety graph, which uses AI to look at how teams collaborate to deliver care for patient populations. We found that in a typical year, thousands of caregivers are connected by millions of patient care interactions, and we can now measure those interactions at scale and use that information to propose improvements
to patient care.

Looking at these interaction points between caregivers helps us better understand health care teams and how they cooperate daily. Most recently, we’ve extended the graph to be multimodal, linking clinician networks to similar patient-centered networks of medications, labs, consults and procedures. We can now see who clinicians work with and how these collaborations translate into pathways of clinical care.

Q: How else is the AI-QI Initiative pioneering new approaches to quality and patient safety?

Tighe: We recently launched a new grants program called Rapid AI Prototyping and Development for Patient Safety, or RAPiDS, to encourage the creation of unconventional AI solutions that AI health researchers and clinicians might not typically pursue because they aren’t sure they will work. The goal of this program is to create a safe place for developing, testing and advancing innovative AI quality improvement efforts.

We also brought together teams from across UF and the College of Medicine to test a secure data and computing resource called ALPS — AI Labs for Patient Safety — that will create an integrated infrastructure for sensitive analytics and develop an ideal workflow for promoting teamwork across disciplines.

Q: What do you see as the potential impact of these AI-QI efforts on patient care and hospital operations?

Tighe: We often talk about the “Swiss cheese” model for safety, when “holes” in layers of stacked safety systems allow a medical error through. While it’s a nice model, it’s of limited use for mapping out a single safety issue, let alone a complex system like a hospital. I think the AI-QI efforts, through projects like the patient safety graph, will help us not only create this new kind of map, but also use the map to simulate how we can change clinical workflows to improve patient safety.

I’m also so excited to link AI-QI to clinical AI research across the college. What if we could predict a patient’s medical error or surgical complication and then use the patient safety graph to automatically predict potential improvements to the care plan, all before the patient ever arrived at our hospital? One of the long-term goals of AI-QI is to not just help translate and implement the phenomenal AI research underway here into clinical practice, but to do so in an impactful and sustainable way.