A Washington State University multidisciplinary team along with Gonzaga University computer science assistant professor Gina Sprint created a new algorithm in 2020 to analyze data from noninvasive technology such as home motion sensors. Their study monitored 14 volunteers doing various everyday tasks at home. Among them, seven members had mild cognitive impairment and in some cases dementia, while the others were healthy.
The researchers found that the healthy group moved about twice as fast in their homes as the cognitively impaired members, who also showed higher variability in time spent on activities of daily living. The participants live in a Seattle retirement community, and WSU researchers have followed them since 2011. The youngest is 73.
"We also found that participants with cognitive impairment tended to sleep more than the healthy adults, particularly daytime sleep as well as nighttime sleep," said Diane Cook, with WSU and principal investigator of the smart home research. "Although the total sleep time is larger, they also have a greater number of sleep interruptions."
"This particular algorithm allows us to quantify and characterize behavior differences between groups of older adults who are cognitively healthy and those who have some cognitive impairment, and that's a first step toward being able to detect early symptoms of cognitive decline in a person's own home and allows us to respond and treat the condition quicker, and, hopefully, keep people functionally independent in their own homes longer."
Cook is the WSU Huie-Rogers chair professor in the School of Electrical Engineering and Computer Science. Variations of this work are ongoing at the WSU Center of Advanced Studies in Adaptive Systems. The recent cognitive study involved Sprint, Cook and Roschelle Fritz, an assistant professor of nursing at WSU Vancouver.
Cook said for typical sensors in a residence, the hardware hasn't changed much in recent years, so current work has focused on developing software for analysis. The new algorithm is called Behavior Change Detection for Groups, and the study's findings appeared in the IEEE Journal of Biomedical and Health Informatics. Such software will be helpful to gain long-term insights, Cook said.
"We have been collecting sensor data from the devices placed in people's homes over multiple months and years," Cook said. "The data we collect allows us to extract digital behavior markers, so just indicators of what a person's routine looks like. Then our psychology partners visit the participants on a regular basis — typically twice a year — and administer traditional neuropsychology clinical assessments."
"Our software, then, is designed to try to predictively link these behavior markers with the clinical scores, so our ultimate goal would be to automate assessment of a person's health in their everyday settings. It is just motion sensors, so there is almost no identifying information."
Another way it can be beneficial is in detecting nuances such as poorer sleep patterns. "Because if we're really close to a family member, we may not always be aware of these changes. And I think as clinicians or doctors, it might be difficult for them to pick up on changes if the person only visits them for a 15-minute doctor's appointment."
Researchers examine the information extracted from the smart home sensors that includes participants' walking speed, the time they spent on activities of daily living and the regularity such as the time of day and the amount of time spent on each activity, as well as how much that might vary from day to day.
"We can look also at their overall movement, not just walking speed, but how much are they moving around the house on a given day and again the regularity from day to day," Cook said.
"So, these activities are actually automatically recognized again by machine-learning algorithms, so the data is automatically labeled with things like cooking, dressing, eating, entering or leaving the home, hygiene, relaxing, sleeping, socializing, washing dishes, working, bathing and bed-toilet transition, which is getting up in the middle of the night to use the restroom — very important to monitor for older adults — and those are all automatically recognized from sensor data."
Using home motion sensors can be ideal for older adults who don't want to change their routines or aren't cognitively able to because there isn't record-keeping or the wearing of smartwatches to collect data, Cook said. However, a different study is ongoing to include smart devices worn by participants along with the home motion sensors to try to assess a person's functional health.
This study expands beyond just cognitive health, especially if someone is compensating well despite memory issues by using notes, calendars or technical devices. The research will assess a person's functional ability to perform critical daily activities independently. "That study is about a year out, or given the pandemic, two years out; we haven't been able to interact face-to-face with older adults and recruit," she said.
The group also has an ongoing research project with participants at Touchmark on South Hill in Spokane to examine multiple chronic health conditions such as Parkinson's. As examples, the technology could help automatically detect falls within the home or frequent trips to the bathroom due to urinary tract infection. WSU students help in reviewing the health-related events.
In the future, researchers hope that advancing the technologies and software can support older residents in the individuals' homes and in retirement communities.
"There are already products where you can install sensors similar to what we use, and the companies will write rules to look for specific types of sensor readings and give caregivers reports," Cook said.
"But what we're looking at is pushing the envelope in terms of the insights and interventions that they can provide, so it's not a matter of being there or not, it's just a matter of how mature and useful they are."
©2021 The Spokesman-Review (Spokane, Wash.). Distributed by Tribune Content Agency, LLC.
