It started with headaches. Melissa Carleton originally wrote them off as an annoyance, a normal irritation of her pregnancy, then in its third month. But after the headaches grew in intensity, the doctors confirmed the worst: Lande had a fist-size brain tumor, and just five days after her diagnosis, she had a seizure so severe that she fell into a coma. Mercifully, her baby boy was healthy.
Over the next few months, Carleton’s progress was slow. Her doctors used amantadine to try to wake up parts of the brain that had shut down. Physical therapists began the long, slow, preliminary work that could eventually help her regain conscious control of her physical faculties. Carleton’s hand shifted; her head moved to one side—fleeting moments that slowly revealed a long, but real, road to recovery.
But there were major obstacles in the way, and Carleton’s husband, Brian Lande, was shocked at the rudimentary method with which medical staff assessed progress. A doctor would sit in the room with Carleton for about five minutes. Their observations during those sessions became the basis for her treatment plan. That represents 0.3 percent of her day, Lande noted. What if they missed a movement? What if that splinter of time gave them a false impression of what was happening?
I met Lande when he was a program manager at Darpa (the Defense Advanced Research Projects Agency), where he had been responsible for a project using sociometric badges to study behavior. Sociometric badges are wearable sensor ID cards that use microphones to measure how people talk—employing Bluetooth and infrared proximity data—and, crucially, use an accelerometer to measure how they move around. In my research, I’ve used them to study interaction and collaboration at work. Brian wondered if we could use them to capture and understand Carleton’s movements. For the past month or so she’s been wearing two accelerometers, one on each wrist, that take hundreds of movement samples every second.
While analyzing this data, I was able to pull out intentional movements by looking at significant differences between her left and right arm movements. Basically, if someone sat her up or moved her bed, both accelerometers would register similar energy levels. If she intentionally moved one arm, it registered as an event. What we saw was quite encouraging:
The improvement seen on May 13 wasn’t a fluke. Lande had noticed his wife moving her right arm more frequently but didn’t grasp that it was at such a high level. Looking at the data we can see improvement in both arms, although more marked in her right:
The data strongly suggest that Lande is improving. If the gains seen on May 13 hold steady, we can be positive about her effort to recover. (Doctors may want to focus on her left arm a bit more, as it appears to be slightly behind in overall strength.)
But there’s another conclusion: Without the accelerometers, doctors may well have missed the burst of activity on May 13. Her treatment might have changed; worse, the doctors might have decided to stop some of their interventions entirely. In this case, we were lucky. What about the hundreds of thousands of other patients in similar situations?
Coma patients also tend to “wax and wane” in wakefulness and awareness. It’s critical for doctors and physical therapists to know when their patients are ready to participate in therapy or a neurological assessment. The traditional way to assess this would require sticking a patient in an fMRI machine every few hours, at a cost of thousands of dollars per session. Accelerometers, on the other hand, cost about $1 each.
Over the next few weeks, I’m going to turn this into an automatic tool for Carleton, Lande, and their medical team to help them understand Carleton’s progress and assess the effectiveness of different treatments. Hopefully this will also help prove to their insurance company that the treatment she’s receiving is effective. It’s hard to argue with objective data.
Insurance companies are overall quite bullish on the idea of using wearables to assess treatments, and of using this data in their business. “Wearable device data will be current and shown over an extended period of time for underwriters to make better decisions,” says Jeff Root, founder of Rootfin Life Insurance Services. “Maybe insurance providers can even market lower rates directly to clients based on … milestones achieved through the wearable device.”
This would be similar to Progressive’s (PGR) Snapshot program, which uses an accelerometer to monitor driving behaviors. If the data show you’re a safer driver, you get a discount. While we’ve seen similar programs in health—some health insurers, for example, give discounts based on pedometer data—applying such tools to therapy treatments is still far outside mainstream practice.
One company that’s working to change this is Ginger.io, a Massachusetts Institute of Technology spinoff that uses sensor data from cell phones to assess disease states and work with medical providers to create preventive interventions. Using calling patterns, movement data, and location, they’re able to predict with incredible accuracy how likely you are to become depressed, get sick, or even be hospitalized. Today this technology primarily helps hospitals manage costs and patients stay well. It’s not hard to imagine this program feeding into insurance providers to ensure that therapies that actually change behavior and well-being are supported, while those that don’t have an impact are phased out.
As we’ve seen with Carleton, this data can be used to assess treatments and direct future care. What’s more, by having a tangible record of improvement, we’re able to imagine a future where our loved ones are healed, where we’re able to become whole once again.Carleton and Lande’s son, West, was born on May 22. His mother, still comatose, is slowly gaining strength.
Go here to donate to help defray Melissa Carleton’s medical costs.