Article Archive
May/June 2013

Sensor System Transmits Health Data

By Jessica Girdwain
Today’s Geriatric Medicine
Vol. 6 No. 3 P. 8

Maintaining the ability to live independently and remain in their own homes is a nearly universal concern among older adults. Chronic conditions and health problems can sometimes push elders into assisted-living facilities, but a system being developed at The Center for Eldercare and Rehabilitation Technology aims to head off that possibility. ElderTech, a sensor network that alerts physicians and nurses about patients’ potential health complications as they occur, may offer a solution that allows elders to remain in their homes.

Developed by a team of experts including Marjorie Skubic, PhD, the center’s director, and Marilyn Rantz, PhD, RN, the sensor network is being used at TigerPlace in Columbia, Missouri, and another facility in Cedar Falls, Iowa.

The system is designed to be noninvasive and nonthreatening to older adults. It includes several motion sensors placed throughout the home to detect older adults’ movements and the timing of their activities. For example, the bed sensor, installed under a mattress, can sense an elder’s movements and heart and breathing rates although it has no actual contact with the patient.

But that’s only part of the story. The data the sensors provide are just that—data. It’s up to the team of physicians and nurses to analyze data patterns and develop a formula that predicts patients’ problems. The work to develop such profiles has been the domain of Richelle Koopman, MD, MS, an associate professor of family and community medicine at the University of Missouri at Columbia. She has worked with the team for four years, creating clinical alerts for the professional staff based on the information provided by the sensors.

Data Analysis
To do this the team initially looked at the information from patients who had gone to the hospital and emergency department and analyzed the corresponding data patterns fed from the sensor. They discovered that changes in a patient’s patterns of activity could predict an adverse health event up to two weeks ahead of time.

For example, when Koopman looked at the data from one woman who was hospitalized for an undisclosed condition, the sensors indicated that her respiratory and heart rates had increased and she was making more frequent trips to the bathroom. Also, she wasn’t sleeping in her bed, instead opting for a chair in the living room (the chair sensor registered her vitals). “I asked, ‘Did this woman go to the hospital for congestive heart failure?’ And yes, I was right. When you start to put the patterns together, it all makes sense,” she says.

From there, the team developed an alert program. “We look at a two-week window. When things look like they’re changing from the previous two weeks or even day to day, we’ll send an e-mail alert to the doctors and nurses that are treating the patient,” Koopman explains. “The whole point of the sensor network is early detection of illness.”

The system focuses on the variations in patients’ routines and activities and the interpretation of those deviations. “We don’t diagnose problems with the technology. Instead, it’s about letting the clinician know that there’s been a change, and it’s up to them to decide what needs to be done for further diagnosis. With the automatic alerts, we’ve made it pretty easy for clinicians to use it,” says Skubic.

The system is adept at identifying urinary tract infections (UTIs)—often long before a patient even knows something is wrong. “We can detect UTIs with increased bathroom use. The alert will be sent out and then a nurse will check the patient’s urine, and it will be abnormal,” says Koopman. At the time, the patient’s only symptom may have been increased urinary frequency. “Now we’re catching it early instead of having the patients go to the hospital with—if they have a smoldering infection that gets out of hand—full-blown sepsis five days later,” she explains.

At least 50% of the relevant clinical alerts come from the bed sensor data, Skubic notes. “That’s not surprising. If elders are not sleeping well, it could be because of a variety of conditions, but they’ll also have restlessness problems if they’re in pain,” she says.

The sensor system has likewise proven beneficial in the case of monitoring postoperative hip replacement patients. Koopman notes that if it appears an elder is not sleeping well, the sensor network can alert the nurse. The nurse would then talk to the patient about pain control and determine whether the patient was taking the proper medications at the times prescribed.

The system also can help detect mental health issues. In that case, the team looks over a month’s worth of data for comparison with a patient’s activity patterns from the previous six months. “We can see when an elder is out of their apartment during lunch or dinner, but if that drops off, then that can clue nurses in to potential depression. Or if a patient looks like they have their nights and days mixed up—more activity during the night, more napping than sleeping—that may point to cognitive decline,” explains Koopman.

However, while the day-to-day alerts are automatic, detecting long-term changes, such as depression, is not. A nurse would need to detect a change in a patient and then consult data patterns to determine whether notable changes could be indicative of depression.

Skubic shares the story of a husband and wife who were helped by a gait analysis system in their apartment. “After six months, we could see that the husband’s gait deteriorated. He was walking slower because he started taking shorter footsteps. It was an indication of cognitive decline, and he was diagnosed with early-onset dementia,” she says.

The impressive aspect of the story lies in the fact that the woman had Parkinson’s disease, and the system could distinguish between the two people. Physicians could see that her gait remained stable, which indicated that her Parkinson’s was well controlled. (The gait analysis system is not yet available in every apartment using the sensor network, though it should be soon, according to Skubic.)

Tool of the Future
Older adults’ privacy concerns create a potential drawback in utilizing the sensor network. For most patients, says Koopman, it’s not a problem. “As long as there are no videos or images of the patients, overall they said it’s not really a privacy issue. They just didn’t want people seeing them without their hair done or in their housecoat,” she says.

Another concern, of course, is false alarms. The team continues to work on refining the system so that it gives an alert when there really is something going on.

Although the sensor network isn’t yet commercially available, Skubic hopes the sensor system will move in that direction. “We would like to work with a commercial entity to see it widely deployed,” she says. The cost to install the sensor network is about $1,200 per apartment. However, Skubic notes that as sensor systems are installed in larger volume, she expects the price will go down. “That’s not a big expense to detect health conditions early,” she says.

One thing health care practitioners should know about this type of sensor network is that there are many different professionals involved. “There are social workers, computer science experts, physical therapists, nurses and doctors, and other experts. It’s an interesting challenge to learn each others’ languages,” Koopman notes. “It takes quite a number of months to integrate with the team to do this. That’s why the longevity of a team is so important.”

Skubic predicts that the sensor network may soon be widely available. “We want this to get out to the people who really want to use it. Our goal is to make the clinicians’ jobs easier and the elderly healthier. We want to help keep them at home where they want to be.”

— Jessica Girdwain is a Chicago-based freelance writer who has contributed health-related articles to several national magazines.