For more than a decade, engineers have been innovating their way through the nascent field of digital medicine: creating pills with sensors on them, disease-detecting facial recognition software, tiny robots that swim through the body to perform tasks, smartphone-based imaging and diagnostics, and sensors that track our vitals. But for all that creativity, only a small portion of these inventions get widely adopted in health care. In an essay published today in Science Translational Medicine, Eric Topol, a cardiologist, geneticist, digital medicine researcher, and director of the Scripps Research Translational Institute, discusses the state of digital medicine, ten years in. IEEE Spectrum caught up with Topol to ask a few questions for our readers.
Eric Topol on:
IEEE Spectrum: What’s lacking in digital medicine?
Eric Topol: There’s lots of promise but we’re short on proof. Over the last ten years, we’ve developed sensors for monitoring things like heart rhythm, glucose, and sleep apnea, and we’ve shown that they work well and can make a difference. But now what we’re trying to accomplish is much more intricate, because we’re talking about using deep learning algorithms that diagnose common, non-life threatening conditions like ear infections and urinary tract infections and skin rashes, and algorithms that can read radiology scans. These algorithms will help people avoid going to the doctor, but they have to be validated so that we know they’re not making mistakes. So we’ve got to do the rigorous research—ideally randomized, controlled clinical trials—and so far, we haven’t had enough of that. Generally, the trials that have been done are not large enough, long enough, randomized, or targeting the right patient population.
Spectrum: We might be short on validation, but we’re certainly not short on innovation. Engineers come up with new stuff all the time.
Topol: The engineers are out front but the medical community has lagged. That’s partly because the proof points are needed, but also because there’s resistance to change. I always get concerned when we have a remarkable innovation and it just doesn’t get the use that it deserves.
Spectrum: What’s an example of a great innovation that hasn’t been widely adopted?
Topol: The smartphone ultrasound, like the Philips Lumify. It’s a brilliant engineering advance. You can connect a probe to your smartphone and have this exquisite imaging—equivalent to a US $400,000 ultrasound machine that hospitals use—and yet it’s hardly used at all.
Spectrum: Why isn’t it used?
Topol: Challenges with reimbursement is the main reason, and other factors such as expense and training. Five of these devices have been approved by the FDA [U.S. Food and Drug Administration].
Spectrum: Is there something engineers can do to increase the likelihood that their innovations will be accepted in health care?
Topol: Ideally engineers should work in tandem—symbiotically—with medical researchers and people caring for patients. That way, they’ll better understand unmet medical needs, and can help plan the proof or validation to help promote acceptance.
Spectrum: In your essay, you said “it was naïve to expect that having a map of the genome alone would transform the future of medicine.” Tell me about that.
Topol: If you think back to the year 2000 when President Clinton and Francis Collins and Craig Venter stood on the White House lawn and said they had decoded the secret of life with the sequencing of the human genome, the expectations, at that time—of having that [information]—were extraordinary. But a genome can only get you so far. And what we’ve learned in the almost 20 years since then is that the genome is just one dimension that explains our uniqueness and individuality. You can learn a lot about a person from other layers of information, like physiology, anatomy, gut microbiome, and environment. Sensors and digital medicine tools are a vital way to track and understand some of those layers. But to date, there’s been little convergence between digital medicine and genomics, leaving us with a very narrow view of a person that is grossly insufficient.
Spectrum: You write in your essay that you expect to see voice assistant technology, like Alexa or Siri, evolve into voice medical coaches. That sounds both exciting and terrifying. How do you envision such technology being used?
Topol: The opportunity is extraordinary. No one doctor could take all of one’s data and continually process it and give you feedback. With a voice medical coach, ultimately, in the future, all of your data is being assimilated, and it could get back to you with up-to-date medical literature. I wrote a chapter about it in the book Deep Medicine. One scenario I gave was using voice medical coaches to determine how frequently you need to be screened for a particular disease, based on your specific data. Recommendations for that sort of thing now are dumbed down and are the same for everybody. But we’ve got to prove that this sort of technology actually helps people. And that takes big clinical trials.
Spectrum: You concluded your essay by saying that many serious concerns about digital medicine still loom, including algorithmic bias, black box issues, health inequities, and privacy and security. Which of those concerns you most, and why?
Topol: Privacy and security is the one we’re doing the least about, and has major ramifications and could destroy the progress of the whole field. If enough people are scared to used digital tools or AI analytics, then we have nothing.