Health-care professionals have long felt that wireless devices and sensor networks could be used as aids in caring for elderly patients, especially those suffering from various forms of dementia. As Baby Boomers age, the impacts and costs of dementia on public health are likely to grow significantly. A 2013 research paper published in the New England Journal of Medicine concluded that dementia already inflicts a financial burden on par with that of heart disease and cancer. But to what degree can technology improve dementia patients’ care and treatment—and could technology be used to cap or even lower the overall costs of caring for these individuals?
Those are questions that a group of researchers and physicians are setting out to answer through a clinical trial known as Dementia Care Ecosystem: Using Innovative Technologies to Personalize and Deliver Coordinated Dementia Care. Funded by a $10 million grant from the Centers for Medicare & Medicaid Services, the trial is being managed through a partnership between the University of California San Francisco’s Memory and Aging Center and the University of Nebraska Medical Center. It will involve 2,100 patients who are at least 45 years old, live at home, have been diagnosed with some form of dementia and reside in California, Nebraska or Iowa. The trial will evaluate a model of care, based on the use of an online dashboard and telephone-based support for the patient’s primary caregiver, that—according to the program’s website—”emphasizes coordinated, continuous, and personalized care and aims to improve health and satisfaction for patients and their caregivers.”
The trial will also evaluate, through a subset of 300 patients, the use of wireless technology to monitor participants’ movements and habits. Via machine learning and analysis of baseline activity for each patient, the software that collects the sensor data will generate alerts in the event that apparent changes, either sudden or gradual, in a patient’s movement or routines may signal a health problem or the progression of his or her symptoms.
“At the heart of the program is what we call the Care Team Navigator,” explains Katherine Possin, one of the study’s directors. The Care Team Navigator is a person who is assigned to coordinate care and support the patient and the patient’s caregiver. He or she will use software called the Navigator Dashboard, an online workflow management tool that has three main modules, dedicated to the caregiver, decision-making and medications. The Care Team Navigator assigned to each patient will use the Navigator Dashboard to learn about that person’s disease, and will be able to ask questions regarding treatment. The decision-making module is designed to help the patient and caregiver create a roadmap for how to proceed as his or her disease progresses. The pharmacy module serves as a means for a pharmacist to monitor any drugs that the patient may be prescribed, and to look for potential conflicts and harmful side effects. The Navigator Dashboard gives each Care Team Navigator access to a team of nurses, social workers and pharmacists with whom to consult. “The Care Team Navigator can triage anything with this team,” Possin says.
For each of the 300 participants who will be involved in testing the wireless technology—which is referred to as Functional Monitoring—trial program assistants will install a number of Bluetooth beacons, made by Estimote, throughout his or her home. The team will also issue each patient a Sony SmartWatch 3, which runs the Android Wear operating system and has an integrated Bluetooth radio, an accelerometer and a gyroscope. The smart watch will be paired with a smartphone that each patient will also be issued for use during the trial, according to Katrin Schenk, an associate professor of physics at Randolph College. A few additional sensors will also be placed around the home, such as on the refrigerator door or on the handle of the toilet flusher, which will be used to deduce the patient’s actions.
Schenk worked with Stephen Bonasera, a geriatrician from the University of Nebraska (formerly of the UCSF’s Memory and Aging Center) to develop the technology. She says researchers from the University of California, Berkeley—specifically, professor Alexandre Bayen and his graduate student George Netscher—helped develop the Functional Monitoring system’s machine-learning capabilities.
The beacons will continually transmit their universally unique identifiers (UUIDs). The Bluetooth radio inside the patient’s smart watch will collect each UUID reading, along with the transmission’s signal strength and a timestamp, and upload the readings to the smartphone. The phone will then forward all of this data to the Functional Monitoring system, which will determine the wearer’s distance from the specific beacon that transmitted each UUID, based on the received signal strength. In this way, it will be able to determine in which room of the house the patient was located when each UUID was collected.
During the first month of the trial, Possin explains, the Functional Modeling system, through its cloud-based servers, will simply collect all movement and location data transmitted via each patient’s smartphone, in order to establish a baseline for every participant. This baseline establishes what a normal day would look like, in terms of activity and the locations within the home where the patient spends the most time.
Once the baseline has been established, the Functional Monitoring software will trigger alerts if it detects deviations from a patient’s baseline movements or habits. These changes could be sudden—such as a marked increase in the number of times a patient flushes her toilet, which could indicate that she has a urinary tract infection—or gradual, such as declining activity in the kitchen, which could indicate a change in eating habits. “Maybe, typically, they go to the kitchen every morning at the same time,” Possin explains, “and then one day, they don’t get out of bed.”
If the Functional Monitoring software detects an anomaly, it will send an alert to the Navigator Dashboard, which the patient’s Care Team Navigator will quickly see and investigate through whatever means is most appropriate. That could mean a quick phone call to check in with the patient’s caretaker or the patient, or it might require an unscheduled visit to the patient’s home.
“Over time,” Possin says, “we can use these events to train the system further, using machine learning.” For example, if a patient moves more often than normal between the first floor and the basement, especially on the same day each week, this likely indicates that he or she is doing laundry. Over time, this would not be deemed an anomaly. However, if the wearer were to go downstairs and not return for an extended period of time, or if these movements were detected during the middle of the night, the software would likely trigger an alert.
The Care Ecosystem team began recruiting 2,100 patients for the trial in March of this year, according to Possin, and plans to begin recruiting 300 patients for the Functional Monitoring subset this month. The trial, which has already begun with the first batch of recruited patients, will continue until August 2017, meaning that most patients will be actively involved in the trial for one year or more.
“There is a lot of work that needs to be done to help dementia patients live at home longer,” Possin reports, “both because it’s often their preference—because it allows them to be with or near family members—and because having to move to a health-care facility is so expensive.”
The 2,100-patient study is randomized so that for every three patients, two are aided, through a caregiver, by the triage services (caregiver, decision-making, pharmacy consultancy and so forth) offered via the Navigator Dashboard. One out of every three patients will not receive these services.
Every six months, the Care Ecosystem team will interview the caregivers—and, in some cases, the patients—to evaluate how effective the personalized care has been. At the end of the trial, the treatments’ effectiveness and costs will be evaluated.
“If the trial does improve health care [for the patients who receive the personalized treatments] and does not increase costs,” Possin states, “the idea is that Medicare will work with us to scale the program out nationally.”