UT Austin tries for ‘AI-native’ health system

Across the country, hospitals are implementing AI, whether it’s scribes to take notes or agents to take calls.

But what would it be like to build a hospital with all this technology available from the start? University of Texas at Austin (UT Austin) is trying to find out.

On April 21, UT announced a $1 billion donation to build what it’s calling an “AI-native” facility and research center. Projected to open in 2030, the UT Dell Campus for Advanced Research and the UT Dell Medical Center are being designed to keep AI and technology at the center.

“Our goal is really to shift from a system that reacts to illness to one that is continuously supporting health,” Claudia Lucchinetti, VP of medical affairs at UT Austin and dean of the university’s medical school, told us.

A different kind of hospital. Healthcare can be difficult for patients to navigate.

In a 2023 KFF study of 3,605 insured US adults, 58% reported having an issue with their health plan in the past year.

“Technology is often something that’s tacked on to a broken chassis,” Lucchinetti said. “We’re trying to build a system that navigates the patient.”

UT sees the building as an “active part of the care team,” where the physical space and digital systems are meant to work together to monitor the patient and make better decisions about their care, according to a UT spokesperson.

To do this, the hospital said it will have an Intelligence Performance Center that works in the background, connecting patient monitoring devices and electronic medical records with lab work—and even giving patients control over room lighting. For example, if a patient has trouble resting before a procedure, the technology could notice and allow the patient to sleep more, adjusting their meals and medications accordingly, Lucchinetti said.

Additionally, every patient at the hospital will have a digital twin, which she said could help the UT system to collect patient information for clinical trial matching and predicting additional health challenges.

Technology is also planned to transcend into the operating room with AI image-guided robotic surgery tools as well as back office systems that can initiate pharmacy delivery, documentation, prior authorization, bed flow, and supply chain challenges.

“For patients, it should feel invisible,” Lucchinetti said. “It ideally would mean they experience faster answers, earlier detection, fewer delays.”

Follow the patient. Say a 50-year-old woman is exhausted and doesn’t quite know why. Normally, that patient would call her doctor, do some tests, and get referrals. It’s a process that could take days or even weeks, Lucchinetti said.

“Each step is disconnected in many systems,” she added. “In the system we’re building, it looks different. Her data, her labs, her history, even her wearable information is continuously being analyzed.”

In this case, the patient would be contacted proactively, Lucchinetti said, and when she arrived for her appointment, there would already be a care plan in place, like imaging and specialist appointments.

“The system is taking care of her,” Lucchinetti said. “When we do that, that changes everything before, during, and after a visit.”

Potential issues? The risk of this model, Kate Eisenberg, senior medical director of health tech Dyna AI told us via email, is cybersecurity.

“As we’re able to aggregate imaging, ambient capture, claims, and behavioral information, you create both enormous, previously unattainable clinical value, and a substantially larger cohesive body of data that could be subject to bad actors, along with re-identification and secondary-use risks that current consent frameworks weren’t designed for,” she said.

Hospitals can take steps to ensure security in the AI age, she said. For one, systems should determine the risk of each tool it plans to implement. A scheduling tool, for example, wouldn’t need the same oversight as an autonomous medication selection agent.

It’s also important to continuously monitor tools after they are deployed and provide patients transparency around how AI is used in their care.

“It will take a certain amount of humility and willingness to update care delivery and governance models as new information and tools emerge. They will definitely need to be willing to learn as an enterprise,” Eisenberg said.