To bring human judgment to bear on big data, Washington State University researchers received a $1.77 million grant for a project designed to help people stay in their homes as they age.
The statistics are mind-bending. According to the Population Reference Bureau, in less than 50 years, there will be more than 98 million Americans ages 65 and older, making up almost a quarter of the population. Today’s situation—with some 46 million people aged 65 and older making up 15 percent of the population—is historically unique and already challenges our health care resources.
Many technology companies and scientists are looking to technology as part of the solution. But the human factor is important too, and that’s where Roschelle “Shelly” Fritz, assistant professor of nursing at WSU Vancouver, and two colleagues from WSU Pullman come in. They are examining the potential of applying human judgment to smart-home data to detect behavioral patterns. The project seeks to determine whether technology can help people stay in their homes longer.
In a smart home, many household systems—such as appliances, lighting, TVs and other digital equipment, security, heating and cooling—are networked and can be controlled remotely via the internet. The homes Fritz is studying are designed with sensors strategically placed to record the residents’ movement. The sensors deliver moment-by-moment data showing those movements throughout the day and night. Fritz’s job is to interpret that data using her clinical judgment.
To investigate how smart-home technology can monitor the health and safety of older adults from afar, the project received a five-year, $1.77 million grant from the National Institute of Nursing Research—part of the National Institutes of Health. Its title: “A Clinician-in-the-Loop Smart Home to Support Health Monitoring and Intervention for Chronic Conditions.”
The three lead investigators, all female, have complementary perspectives. Diane Cook, professor in the School of Electrical Engineering and Computer Science, Voiland College of Engineering and Architecture, provides the technology expertise. Maureen Schmitter-Edgecombe, a professor of psychology, is exploring whether health interventions affect cognitive health. Fritz provides the clinical expertise.
Fritz is well suited to the task. She has been a nurse for 25 years, with experience in emergency rooms, public health, hospital administration and employee health, as well as teaching. And her work is drawing attention. In November, she was invited to be on a panel in Seattle for Xconomy’s event on artificial intelligence and health care. She was the only nurse in the room, among technology company executives, neurosurgeons and others—all sharing their work with each other in pursuit of their common goal—to transform health care.
What behavior change implies
For the smart home study, multiple sensors are deployed in six homes in retirement communities in Spokane, Wash., a seventh in Beaverton, Ore., and an eighth in Southwest Washington. (The research may expand to Vancouver and other locations during the grant period.) The sensors detect motion and record the residents’ data moment by moment. The clinician analyzes the data to determine the occupants’ routines, such as getting out of bed to go to the bathroom or the kitchen at night. When the clinician flags a particular pattern as important, the engineers train the machine to recognize similar patterns that could be encountered in the future. Then, automatic alerts can be developed for caregivers. If the change is significant, it may indicate the need for a health intervention, such as a visit or call.
So, for example, if a person who regularly spends just a couple of minutes on a trip to the bathroom at night is out of bed for 30 minutes, it suggests some kind of incident, such as a fall, and a caregiver would be alerted.
“We are not doing diagnostics,” Fritz said, “but training an intelligent machine to understand what a change in health state looks like, much as a home health nurse would assess that.”
The sensor technology adds something lacking in previous research, where people tell the researcher about their experience. That something is accuracy. “Sensor data has more credibility than when someone self-reports,” Fritz said. While someone might neglect to tell a caregiver about a recent slip, for example, the behavior change cannot be ignored by the smart home.
During the grant period, the research will focus on developing the machine-learning piece of the puzzle. Taking into account the person’s health condition—such as a chronic disease—Fritz will tell the engineers what she believes she is seeing in the data, and they will develop an algorithm and alerts when motion indicating a change in health state is noted.
“We are not doing diagnostics,” Fritz said, “but training an intelligent machine to understand what a change in health state looks like, much as a home health nurse would assess that.” Essentially, they are training the machine to learn about the individual’s health and behavior.
How to understand the data
The research also seeks to inform engineers’ ability to communicate effectively with clinicians, Fritz said. “The only way the data can impact clinical decision making is if we are provided information we can understand. How do you take big data like this and present it to a health care provider in a way that’s meaningful and relevant to their work—which medicines or treatments to prescribe?”
Students are working on the project along with the researchers. Cook’s and Schmitter-Edgecombe’s student engineers and psychologists are developing visual analytics based on the data and working with Fritz’s nursing informatics students to determine what visuals nurses like best. “Classes will go back and forth doing that,” Fritz said. “It’s very fun and innovative, multidisciplinary teaching.” A Ph.D. nursing student, Shandeigh Berry, is working with Fritz as a research assistant.
The project’s long-term goals address today’s major health concerns, from the explosive growth of the aging population to the shortage of caregivers to the cost of health care. One big question: Can technology make it possible to extend the reach of caregivers and nurses? Perhaps.
Cost is another matter. “It’s a very simple sensor technology,” Fritz said. “What’s expensive is the infrastructure and brains behind the algorithm.”
Smart homes are not new. For example, Cook has developed a “Smart Home in a Box,” deployed in more than 110 homes around the world. Bringing in a clinician to study the data, along with automating health monitoring, assessment and evaluation of the intervention impact, is an important new step in determining whether the system can help individuals manage chronic health conditions.
Although any product is far from market-ready, Fritz said, “I feel really good about it. The more we get into it, the more I realize that what we’re trying to do is actually feasible.” ■