Mobile's role in population health management

-reproduced from Fierce Healthcare

University of Illinois’s Schatz says that while smart clothes, such as undershirts or socks with sensors built in, would be ideal for measuring population health from a functionality standpoint, such tools have not hit the mass market yet. For now, he says, phones and fitness devices are the most realistic options for tools that already are widely deployed.

Still, Schatz says, devices like Fitbit and Jawbone lack the accuracy to be able to measure relevant medical information.

“Fitness devices are limited because of the packaging,” he says. “The battery is supposed to last for a month without recharging; it’s something that you put on and leave on for a long period of time, and because of that, the battery has to be saved somehow.”

That, he says, is done through limited sampling.

“A typical Fitbit will measure motion once a second,” Schatz says. “A typical medical device, if you’re wearing it on your wrist, will sample 100 times per second. In order to measure walking accurately, you need on the order of 20 samples per second.”

Because of that disparity, Schatz says, while commercial wearables are good for helping younger, healthier people keep track of estimated walking trends, they’re virtually useless for those who would truly need to monitor their activity.

“If you try [Fitbits or Jawbones] on an older person that doesn’t swing their arms very much, like a typical COPD patient, the numbers are totally wrong,” he says. “If you’re trying to use them for medical applications where you’re ... trying to measure something with high accuracy, they’re simply no good at all.”

On the other hand, Schatz says, because phones are basically general purpose computing devices, you can get whatever sampling you want. Schatz led development of an app called MoveSense that uses internal sensors--both accelerometers and motion sensors to determine if the phone is moving--to measure movement of users, plus a backend server that uses analytics to predict health status based on those measurements. If the phone isn’t moving, the app is not transmitting, he says.

The tool, so far, has been tested on less than 200 users, according to Schatz, and has been able to predict clinical values for cardiopulmonary heart and lung patients, such as lung function, with high accuracy, but only in small controlled settings. The real proof of the tool’s effectiveness, he says, will come if its accuracy continues when it is deployed to 1,000 people in everyday settings, which he says is in the works.

“Phones are the sweet spot in that ... you need something that people can carry all the time without thinking about it,” Schatz says. “You need to get around an adherence problem of getting people to use it properly, and you have to be able to actually measure something which a health system would care about.”

***

For a copy of the paper "Classification Models for Pulmonary Function

using Motion Analysis from Phone Sensor," Qian Cheng, Joshua Juen, Shashi Bellam, Nicholas Fulara, Deanna Close, Jonathan C. Silverstein, Bruce Schatz, or "Mining Discriminative Patterns to Predict Health Status for Cardiopulmonary Patients," Qian Cheng, Jingbo Shang, Joshua Juen, Jiawei Han, Bruce Schatz, please contact the communications office.