Advertisement

Electronic Tracking May Predict Cognitive Decline

      Pervasive computing technology can radically change the way clinical research is conducted, “leading to major advances in detecting prodromal change, in managing manifest dementia disease, and in transforming the effectiveness of clinical trials,” according to Jeffrey Kaye, MD, Layton Professor of neurology and biomedical engineering at Oregon Health and Science University, Portland, and director of the Oregon Center for Aging and Technology. Data obtained using pervasive computer monitoring and passive sensing may allow practitioners to accurately predict when cognitively-impaired elderly individuals will require transition to a more intense level of care.
      Dr. Kaye spoke at the Alzheimer’s Disease Research Summit 2015: Path to Treatment and Prevention, sponsored by the National Institute on Aging, in Bethesda, MD.
      A fundamental limitation of current dementia research is detecting meaningful change, he said. “The cardinal features of dementia make it challenging to detect change. The changes are slow and imperceptible. Those features are slow decline punctuated with acute, unpredictable events, all challenging to assess with current tools and methods.”
      Acknowledging the difficulty in dementia assessment and detecting change, Dr. Kaye described a project in which a pervasive computing platform is being used to assess elderly individuals in their homes in an ongoing community-wide “life lab” in Portland, OR.
      The program uses passive iris sensors that provide data about activity, sleep, gait speed, room transitions, mobility time, and location. Additional sensors track door openings and closings. Telemedicine provides physiologic measures and information on balance, body composition, heart rate, room environment, temperature, and air quality. A medication tracker (an instrumented pill box) registers the day and time medication is accessed. Phone use is tracked, and a computer acts as a sensor to track psychomotor activity (keyboarding, mousing) and also delivers a questionnaire that asks about self-reported behavior that would be harder to discover by passive sensing (mood, pain, etc.).
      In developing a community tracking program, “the approach needs to be technology agnostic, so it’s not about a fitness band or a wearable, or competition about who has the best technology,” Dr. Kaye said. “It’s about what works the best. So we tend to use passive sensing whenever possible, but we’re happy to use wearables or carryables as well.
      “This platform works best when it’s scaled out into the community,” he said. This particular project is in place in several hundred homes in the Portland area. Residents are instructed to go about their normal daily activities, and up to 100 volunteers monitor the data, allowing the system to function 24/7.
      Scatterplot data gleaned from the home sensors reveal patterns of rest and activity during typical busy periods, for instance, at night when residents get up to use the bathroom, and during regular daylight hours. As time progresses, however, in residents who develop mild cognitive impairment, the patterns shown in the scatterplots change, Dr. Kaye said.
      He cited “passively obtained” results from previous studies on individuals with mild cognitive impairment (MCI) that support measurement data within the “smart home” community program. For example, in one study, residents with MCI showed higher variance in total activity and walking when compared with age-matched controls. In another, trajectories of walking speed over time (fast, medium, slow) showed elderly individuals with MCI were nine times more likely to be in the slow group. In another study in which sleep activities were monitored and compared with controls, individuals with MCI showed clear differences after 26 weeks in wake-after-sleep onset, and times up at night.
      Taking medication is a cognitive task, Dr. Kaye said, and passive sensors can measure it. “Taking medication is a functional task that has cognition embedded in it,” he said. He described another study with elderly individuals, mostly women, in which medication adherence was used as a measure of cognitive function. Adherence was assessed continuously for 5 weeks with a medication tracker device, as participants took medications twice daily. The group was split into high and low cognition groups.
      “The lower cognition group was significantly worse in their adherence to their medication regimen,” Dr. Kaye said. “These individuals were out of the range of people who might be considered passing for a clinical trial, the 80% adherence rate.”
      He said combining all the avenues of data (weekly self-reporting on mood, falls, emergency department visits, visitors); behavioral activity data (computer use, time out of home, etc.); contextual data (weather, consumer cost index, living in a retirement community); annual clinical assessment; demographics; and controlled data can predict the likelihood of care transitions.
      In Dr. Kaye’s study of 108 individuals, involving more than 63 million observations using pervasive computing technology, “you can predict with high accuracy those will transition [to a higher level of care or nursing home from a normal state] within the next 6 months,” he said.
      He concluded that data attained from pervasive computing also can be used in clinical trials to speed the trajectory of new products in the pipeline.
      Dr. Kaye reported no disclosures.