A focus on digital health: Conference highlights present, future applications

You’re feeling under the weather, but your temperature is 98.6 degrees Fahrenheit — and that’s normal, right?

Well, not necessarily, says Michael Snyder, PhD, chair of genetics at Stanford Medicine. Speaking Feb. 28 at a conference on the evolution and future of digital health, Snyder cited a 2002 paper that reported normal body temperatures ranging between 92.0 F and 100.8 F among 2,749 healthy adults, with 25% measuring 94.6 F or below.

“If your normal, healthy baseline’s here — 94.6 — and they measure you at 98.6, they’ll tell you you’re normal, you’re healthy, everything’s fine,” said Snyder, the Stanford W. Ascherman, MD, FACS Professor in Genetics. “But I guarantee if you’re 4 degrees over your baseline, you’re not healthy. Something’s off.”

The ability of digital, genomic and imaging technologies to provide personalized data about health matters because what’s normal for one person may not be normal for the next, he said. The challenge, according to Snyder and other speakers at the Digital Health 2024 conference, is how to take better advantage of these innovations.

The two-day conference at the Li Ka Shing Center for Learning and Knowledge drew nearly 200 attendees from industry, nonprofit organizations, government, medicine and academia to hear dozens of speakers address a range of topics at the intersection of health and digital technology. The event was sponsored by Stanford Center for Digital Health, the Stanford Healthcare Innovation Lab and Times Higher Education.

Eleni Linos, MD, DrPH, professor of medicine and of dermatology, gave opening remarks. “One of the goals of today is not just to inspire, to teach — we have incredible speakers speaking to us about digital innovation, about AI, about ethics and trust — but also to connect with each other,” said Linos, who directs the Center for Digital Health and holds the Ben Davenport and Lucy Zhang Professorship in Medicine.

Encouraging a spirit of collaboration and innovation, she concluded her opening remarks with quotes from Walt Disney: “We keep moving forward — opening up new doors and doing new things — because we’re curious.”

“It’s kind of fun to do the impossible.”

Resistance to new tech in medicine

Snyder said one obstacle to a wider adoption of new health technology is medicine’s “conservative nature.” He said that a decade ago, mapping the genomes of healthy people was “not well-received by physicians,” who thought the practice would turn up genetic mutations that would make patients hypochondriacs, leading to huge medical expenses.

“That actually has not been the case,” Snyder said.

Digital technology can also provide longitudinal health profiles, which can make detecting medical problems easier, Snyder said. He said that whereas most physicians would discourage a healthy person from getting a whole-body MRI, he would enthusiastically endorse the move.

The naysayers, he said, are concerned that MRI will turn up nodules, which even healthy people have, sounding an unnecessary alarm. But whether you have nodules is not the point, he said. “The issue is, ‘Do you have any growing nodules?’” — that is, potentially cancerous tumors. “And you know that through the longitudinal profiles.”

AI in medical imaging

Akshay Chaudhari, PhD, an assistant professor of radiology, called MRI “a fantastic imaging modality” but one that can generate even more useful information with the help of machine learning. A member of the Integrative Biomedical Imaging Informatics program at Stanford Medicine, Chaudari spoke on the keynote panel.

His research focuses in part on using artificial intelligence to develop techniques for improving the quality of MRI images and producing them more quickly, as well as on finding ways to extract more health information from existing medical images. “There’s a lot of embedded and latent information in these images that we’re just not used to acting on in the hospital,” he said.

Chaudhari said that medical AI is undergoing “a fundamental paradigm shift”: Whereas in the past, researchers would train algorithms to perform one task, that approach is quickly being replaced with foundation models — machine-learning models trained on massive datasets, along the lines of ChatGPT, that can perform a wide variety of tasks.

“At Stanford Radiology, we have around 1 petabyte” — the equivalent of 1,024 terabytes — “of medical imaging data. How do we aggregate all that information? How do we combine that with new data that we get from the medical record? How do we really use information that exists in unstructured clinical notes?”

Machine-learning algorithms haven’t necessarily been developed to contend with the “nuances and intricacies of medical datasets,” he said. “Being able to adapt some of those algorithms to work with our specific datasets is definitely one challenge.”

Patients receptive to digital health

Linos asked Maame Yaa A. B. Yiadom, MD, associate professor of emergency medicine, how clinicians can enhance patients’ understanding and acceptance of AI and digital health. Yiadom, the principal investigator for the Stanford Emergency Care Health Services Research Data Coordinating Center, said she has found that patients are receptive to such technologies and may think they’re being used more in medical care than they actually are. “They’re excited about it,” she said.

She and her colleagues ran a pilot program for virtual visits in the Stanford Medicine Emergency Department. Patients in the ED who were not high-acuity — not sick enough to require being seen immediately — were given the choice to meet with a physician virtually.

With hardware and software to enable virtual visits and an assistant to help with equipment in the exam room, the physician could assess the patient’s condition and even request consults from specialists, such as orthopaedists and surgeons, as well as medical imaging, Yiadom said. (A case study describing the program was published Oct. 19, 2022, in The New England Journal of Medicine Catalyst Innovations in Care Delivery.) The remote physicians could see as many as 30 patients in a shift, whereas as an ED physicians on a normal shift might see 24 patients, Yiadom said.

“We thought patients who were older might not like this because they want to have a personal touch — and they thought it was fantastic,” she said. “Showcasing how we can present the art and science of medicine in a convenient, accessible way can be really appealing for patients.”