When patients flared, researchers found, they did indeed walk less.
Predicting Flares
In another talk, Laure Gossec, MD, PhD, professor of rheumatology at Sorbonne University, Paris, discussed the way clinicians and researchers have turned to activity trackers to try to get a better handle on RA patient flares.
“They’re important for patients, but also they should be important for physicians,” Dr. Gossec said. “But we haven’t seen yet how we can collect them and make sure we’re not overdefining flares, or missing something. Questionnaires and changes in disease activity status are the usual ways to detect flares, but they don’t always agree.”
Patients’ disease activity can, of course, fluctuate wildly between visits, which provide only a snapshot in time. In a study called ActConnect, researchers turned to activity trackers on the theory that when patients are suffering a flare, they’ll walk less, because patients say activity interruption is an important aspect of their flares. In the study, 170 RA and axial spondyloarthritis patients were asked questions each week to assess flare, and their physical activity was monitored with trackers. Dr. Gossec said patients (who received the devices for free) were compliant in wearing them and in uploading their data regularly with a Bluetooth connection.
When patients flared, researchers found, they did indeed walk less. There was a significant link between physical activity and flares (P=0.03). Patients in flare tended to walk roughly 1,000 steps less per day—or about 10–15% below the normal amount of daily activity of those not in flare.
Researchers turned to a commercial source—a phone company interested in venturing into healthcare with a machine-learning angle—to analyze the data with machine learning to try to use the steps walked to predict flare. They came away with a positive predictive value of 91% and a negative predictive value of 99%.2
Dr. Gossec acknowledged the machine-learning process is a bit of a “black box.” It stands to reason, though, that the process likely involves not just the total number of steps, but the pattern of those steps—when they’re taken, skipped exercise days and pace, she said.
She said the results show it is possible to use big data analysis on rheumatology datasets. But it is not a simple proposition.
“If we want to use big data machinery to analyze things like steps per minute, we need expertise,” Dr. Gossec said. “It’s a very specific expertise. It also takes time and money.”
The ultimate use of the data is also a consideration. “Will that information go to the insurance company before it goes to the doctor?” she asked. “These are all issues we need to think about.”
