Dec. 20, 2022 — When going on a road trip decades ago, drivers would pull their Road Atlas from the glove box to get the lay of the land and plan their route. They might even have had a trusted passenger take charge of navigating. But with the advent of web mapping platforms like Google Maps, modern drivers can now pull up a destination on their phones and receive route recommendations in real time based on a variety of factors, such as traffic patterns, and personal preferences, such as avoiding highways or tolls.

“With common map routing software, there’s not someone on a computer saying, ‘Dean Koch’s taking a road trip, let’s design this route for her’; there’s no interpersonal concierge work done,” said Patrick Tighe, M.D., M.S., the associate dean for AI application and innovation at the UF College of Medicine and the project champion for AI-QI initiatives under the college’s strategic plan. “In health care, that’s mostly still how it’s done for quality improvement projects. That interpersonal element is important, but we can upscale this approach and automate some steps in the process while keeping the human as the master of the loop. This is where we think we want to go with our AI-QI initiatives — moving from descriptive to prescriptive analytics and taking a more holistic approach to improve clinical processes.”

With UF welcoming HiPerGator, one of the fastest AI supercomputers in American higher education, and hiring 100 new faculty with expertise in the methodology and application of AI over the last two years, the renewed energy sparked innovation around how to leverage the tool across the health care landscape. As AI continues to become a centerpiece of research endeavors at UF, a team of clinicians, analysts and patient safety experts at the College of Medicine are developing an AI-enabled infrastructure for quality improvement analytics and patient safety at UF Health.

As members of the team began to brainstorm, they determined three needs: the ability to pull experts from different disciplines together, such as software engineers, computer scientists, clinicians and other health care colleagues, to innovate and find solutions to QI challenges; data to analyze and a place to store and organize that data; and a place to do the math — a supercomputer — with the appropriate security measures built in to handle sensitive details such as protected health information.

To address these needs and create opportunities for collaboration and scholarship in the AI-QI space and catalyze publications, presentations and innovations, the Office of the Dean provided funding to establish a grants program, Rapid AI Prototyping and Development for Patient Safety, or RAPiDS. An initiative under the value pillar of the college’s strategic plan, the grants program functions as the mechanism for developing, testing and advancing innovative AI quality improvement efforts, promoting faculty career development and stimulating external funding.

“RAPiDS is not meant for turnkey clinical solutions ready for immediate implementation into clinical care,” Tighe said. “It’s geared toward those prototype ideas where you’re not sure if they’re going to work out, and that’s OK because we haven’t taken from mission-critical resources for QI activities. It’s a safe space to innovate. It brings a multidisciplinary team of folks in data modeling, coding, research computing, AI education and more together, using accessible data with an eye toward patient safety and value. We anticipate that some of these prototypes may graduate to opportunities to work with our implementation science colleagues to see how we can insert them into clinical workflows and processes that have a more direct impact on patient care.”

The first RAPiDS grant cycle, which launched in August, brought teams across UF Health and the College of Medicine together to test a secure data and computing resource remotely accessible by authorized clinical and research teams. The resource, called ALPS — AI Labs for Patient Safety — aims to create an integrated infrastructure and safe space for sensitive analytics, becoming a common place where data coexists in the same spot as a computer capable of handling the data. ALPS also develops the ideal workflows promoting teamwork across such a wide array of disciplines. Another initiative under the value pillar of the strategic plan, ALPS is a collaboration among the College of Medicine, UF Health Shands and UF Research Computing.

“We are forging collaborative connections with the hospital throughout this process,” said Liliana Bell, M.H.A., P.M.P., a senior project manager in the Office of the Dean. “We are seeing success because we are building on the experience and knowledge of teams across the health system and we’re all coming together. It’s an iterative cycle through which we are learning as we go and going as we learn.”

The goal of ALPS is to create the infrastructure for clinicians and others across the health system to access and compute data, using AI processes to facilitate quality improvement in patient safety. Harnessing this tool, clinicians can create a picture of how care is provided and what the health care experience looks like and can also leverage geospatial data that reflect some aspects of social determinants of health to improve outcomes.

Through RAPiDS, which will move into its second cycle in winter/spring, teams can use ALPS to collaborate across disciplines to bring their QI ideas to life at a system level. The ultimate goal of both initiatives is to bring those with clinical expertise and those with data analytics, statistics and technical expertise together to analyze data using AI to improve outcomes.

“Comprehensive data profiles of the patients, clinics and health systems are critical for AI to work,” said Jiang Bian, Ph.D., a professor in the department of health outcomes and biomedical informatics and chief data scientist at UF Health.

The current infrastructure of ALPS is hosted by ResVault, a robust and secure platform used by UF that can handle millions of data elements and meets all Health Insurance Portability and Accountability Act, or HIPAA, compliance requirements. Users must complete a rigorous checklist of access requirements, including Institutional Review Board and ResVault training. Only authorized users can put data into the system and run analyses, and no users are permitted to download data or reports.

“Collaboration between the hospital, the health science colleges and Information Technology is crucial for driving innovation and improving clinical processes through the use of AI,” said Gigi Lipori, M.T., M.B.A., UF Health chief information officer and senior vice president. “By bringing together experts from diverse disciplines, we can leverage the power of AI to find solutions to quality improvement challenges and enhance patient safety.”

As clinicians move through the Advanced Training Program, another initiative under the value pillar of the strategic plan, with a focus on outcomes-driven practice, RAPiDS and ALPS are outlets through which they can take their ideas for quality improvement and patient safety from a light-bulb moment to a real-world solution.

“It’s inspiring to work with such incredible people who are moving outside their comfort zones and doing it with such excitement,” Tighe said. “And it’s cool to learn from the work of amazing folks who have been doing AI work at UF for years. The chance to extend this in a different direction in the QI space is really exciting because it hasn’t been done before. It’s audacious, and with the support of the dean and college, I truly believe that UF is the place to do these kinds of things.”