MIKE SNYDER WAS clearing brush behind his brother’s western Massachusetts house, erecting a fence to keep deer from the blueberries, when the tick bit him. A few days later, on a flight to Norway with his family, his palms itched and his head grew woozy. So the Stanford geneticist dumped a bunch of wearable sensors on his tray table and began doing what he does best: measuring himself.
Low blood oxygen, said the Masimo pulse oximeter housing his finger, and the globe-shaped Scanadu he held against his forehead every few minutes. Weird heart-rate, said the two Basis smart watches strapped to his wrists. Immediately he feared the worst: Lyme disease. Caught too late, Lyme hijacks the body’s immune system to seek and destroy joints, nerves, brain tissue, and—this really made him anxious—the heart. Once his temperature rose in Oslo, he sped to the doctor.
Snyder’s approach to his personal health may seem overzealous, but he’s doing it in the service of science. All of his collected metrics are just one part of a larger digital health study from his team at Stanford, published today in Plos Biology. Consumer wearables—a market expected to reach $34 billion by 2020—have the potential to move from tracking simple metrics like steps and heart rates to providing actionable health information. But before that can happen, researchers need to carefully study how biometrics change in individuals over time, and determine which wearable sensors provide data reliable and useful enough to be used in diagnosis.
The Stanford study includes data from 60 volunteers, including Patient #1, Snyder, who’s done the most time: two years, with the most sensors, seven. (As part of another Snyder study, I’ve worn three.) Combining sensor measurements with genomics and lab results, the study has generated 1.7 billion measurements: skin temperatures, sleep patterns, activity, even radiation exposure. Since Snyder is always with his gear, he knows his ‘normal’—the personal baselines unique to him. In the context of years of measurement, he knew his physiological oxygen and his heart rate on that plane to Norway were abnormal.
But this study isn’t just about helping individuals predict their own health: It’s part of a long play toward more robust mobile diagnostic tools for much larger populations. “This work with device-driven measurements is really going to help inform major cohort-based projects, like those proposed in the Precision Medicine Initiative,” writes Atul Butte, Director of the Institute for Computational Health Sciences at UCSF, in an email. “Eventually these tools have to play a role in medical care,” he says.
Different participants in the study pointed toward different biometric monitoring possibilities. Collectively, data from the study’s insulin-resistant participants suggests that a simple set of measurements, like sleep patterns and steps, could be used to predict others who are insulin-resistant—and provide warning before they develop Type 2 diabetes. The most striking data may be what Snyder’s post doc and study co-author Xiao Li accidentally discovered last August 21, just weeks after he returned from Norway: that wearables could be used to infer nascent inflammation, before users even begin to notice themselves.
“That day, I saw he had an abnormal resting heart-rate,” says Li, “so I checked his blood tests and I saw his high CRP level.” C-reactive protein, a common blood biomarker, is linked to inflammation from infections, and even immune dysfunction, like in autoimmune disorders or cancer. Xiao checked the records, and found a similar pattern from the time Snyder was first bitten by that tick. Both times, he didn’t yet know he was sick, that his immune system was a hot mess—but his sensors indicated that something was up. Once an individual establishes their baseline biometrics, says Li, resting heart rate, with or without skin temperature, can infer CRP levels indicative of inflammation.
Even Snyder was surprised by the discovery. “Not only can these inexpensive devices capture this information at a personal level, and so quickly,” he said, but they can do so with an almost negligible error rate. In Snyder and several others, data suggested inflammation on multiple occasions, which was validated with blood draws suggesting abnormal CRP. “If you see it early you can take your zinc or decongestants right away,” he says. Snyder’s group is filing for a patent on its inflammation algorithm.
Of course, inferring the presence of a blood-based molecule without drawing a drop could prove to be another Theranos-like fantasy. Wearable sensors are still imperfect in many ways: The light-based sensors used in smart watches to detect blood flow change, for example, still have very low resolution. “Some people don’t think low-res is accurate,” says data science post doc Jessilyn Dunn, who co-authored the paper. “But you don’t need such a high resolution, 100 percent accurate signal in order to extract the broader health information.”
The bigger concern for doctors could be perfectly healthy patients raising false alarms. “People tell me that everyone will be going to go to the doctor all the time,” says Snyder. He believes the fix lies in the algorithm itself, which can be tweaked towards greater robustness. Nonetheless, Robert Green, a medical geneticist at Harvard, sees some of the same problems that have come up in genomic medicine. “It’s very interesting to watch people create narratives around the information they receive: ‘I’ve got a gene for this skin disease, and I’ve always had itchy skin.’ Information could be gathered by consumers that they will take to their doctors, and demand unnecessary testing.”
And if we extend this to wearables, says Green, medical resources could end up unevenly distributed. “There’s no question that all of these advanced technologies that are not covered by reimbursement like genetic testing are going to be tried first by people of means—they justify it by saying it will filter down to the rest of society.” But Li adds that health tracking shouldn’t be cumbersome, or as expensive as Snyder’s own collection. Their group is on a special hunt, for useful sensors to lump into one device. “At the end of the day, we just need one watch with all the sensors we want, and an iPhone to dump the data,” says Li. “The Fitbit measures almost everything we want.” And a device in China called the MiBand costs just $10 to $20. “It depends on whether there is a sale or not,” Li says.
In the long term, importantly, consistent monitoring could improve care, especially in less developed areas. “In some areas in the world, there’s a lack of medical resources, and they may not be able to see the doctor immediately—either there is no facility, or they don’t have money,” says Li. “When you are sick enough, you go to the hospital. How do you decide?”
Wearables could help close that gap, when we’re ready. In Snyder’s case, although he marshalled the evidence in Norway–his personal data–the doctor, skeptical of Snyder’s tick theory and probably, his Scanadu, had him take blood tests anyways. They would reveal an infection from bacteria such as Borelia, which causes Lyme.
Article and Photo by: Rinku Patel | Wired Science