CHIME19 Fall CIO Forum: Dr. Eric Topol on the Future of AI in Healthcare

Data-driven tools that diagnose disease and promote wellness are set to transform care for a wider population.

Kevin Joy is the managing editor of HealthTech. He previously worked in content marketing for a major university health system and as a features reporter for the Columbus Dispatch.

Recently, Dr. Eric Topol captured vivid images of his heart, abdomen and left foot, among other things, using only a handheld ultrasound tool and a smartphone.

“A total-body medical selfie,” said Topol Wednesday at the CHIME19 Fall CIO Forum in Phoenix, where the photos were projected onto screens during his closing keynote address. “This all took a matter of minutes.”

But when Topol experienced abdominal pain, his doctor ordered a standard CT scan, which ultimately captured the same details. The cost: $2,800.

SUBSCRIBE: Become an Insider for access to exclusive HealthTech videos, white papers and articles.

This was one of several examples that Topol, a well-known doctor, author and founder and director of the Scripps Research Translational Institute, cited in his speech. Intuitive technologies that use artificial intelligence and deep learning — and in some cases leverage tools that many people already own, he said — boost accuracy in diagnosis, save money and ease physician burnout.

Functions can be as simple as Apple Watch’s electrocardiogram application, which can detect atrial fibrillation before an adverse event. Or they can be complex: AI-assisted colonoscopies that identify polyps and signs of cancer in less than a second have been developed in Japan.

“You don’t want to go through this procedure and have things that are missed,” Topol said.

Diagnosing Disease With AI

The futuristic notion of using algorithms and automated diagnostics in mainstream healthcare is getting closer to reality. “Human intelligence is not going to change in our lifetime, but machines are going to keep getting smarter,” Topol said. “There isn’t one field that won’t be affected.”

He cited AI’s growing role in healthcare, improving recognition of skin lesions to detect cancer, enabling detection of diabetic retinopathy with a smartphone camera and identifying metastatic breast cancer (Google claims its AI-powered tool has 99 percent accuracy, far superior to human ability).

He also shared sobering statistics: Twelve million Americans are misdiagnosed in a primary care setting each year, with one-third of those cases resulting in serious or permanent damage or death. Applying AI and analytics for care throughout a patient’s lifetime could become routine.

Looking to the future, Topol envisions “a virtual medical assistant that will assemble all your data — what you eat, your medical history, all of your multimodal data that will be used to coach you and your clinician. That’s one of the big pushes for the future in AI.”

How AI Can Help Doctors Provide Better Care

Physician burnout remains a serious issue, and Topol said organizations should embrace any technologies that can drive efficiencies without sacrificing care. He cited a conservative estimate that if every patient consultation was shortened by a minute, 1 million hours of outpatient clinic time annually would be made available.

Systemwide, these tools have the potential to handle tasks such as coding, billing and transcription. That’s a notable advantage, he noted, when hospital administrative costs account for more than one-quarter of all hospital spending in the U.S.

Another significant means of cost savings and patient satisfaction: using AI-enabled remote monitoring tools to keep people out of the hospital or to send them home sooner, with data collection and transmission taking place from a distance.

“Instead of thinking about the hospital room of the future, we ought to be thinking about the patient’s bedroom,” Topol said. “You can give them sensors that cost less than a night’s stay.”

The Limitations of AI in Medicine

Despite excitement around its potential to transform care, AI has many hurdles yet to clear.

The biggest one? “We keep using the same limited data sets” in research and development, said Topol. “We need a lot more. Any deep learning model will be compromised by a lack of comprehensive data.” A lack of diversity is also a problem.

Part of the challenge is getting people to understand the value that their data plays in the advancement of care — whether across a population or within their families. And it also will take work to convince providers that AI won’t increase their caseloads — or replace workers.

The goal is about boosting personal attention as well as medical efficiency. “We want, ‘The human will see you now,’” Topol said. “Not, ‘The robot will see you now.’”

Follow us on Twitter @HealthTechMag, or the official CHIME Twitter account, @CIOCHIME, and join the conversation using the hashtag #CHIME19.

Kevin Joy/HealthTech Magazine