Skip to main content

HIMSS: Using Predictive Analytics to Prevent Falls

After a 2014 CNN investigation found months-long waits for care in the Phoenix Veterans Affairs healthcare system—and that at least 40 vets had died while enduring them—local VA leaders sought ways to help their patients more efficiently. Combining technological (a predictive-analytics program) with human (community paramedics) resources is now doing that for elderly veterans at risk of falls.

There are several reasons falls represent a fertile area for intervention, Andrew Muth, MD, clinical informatics fellow at the University of Arizona, told HIMSS attendees Wednesday: They’re a leading cause of injury among the elderly, with high morbidity and mortality. They’re expensive. But the risk factors are well known, and we know what interventions can be effective against them. 

An existing program already worked to identify hospitalized veterans at risk of 30-day mortality or readmission, so the fall analysis was layered onto that. And while outcomes have been difficult to quantify, “I really feel like this model has a role to play,” Muth said, “in moving from fee-for-service to fee-for-value.”

Basically the VA uses a number of criteria to stratify patients by their fall risk. It provides this information to participating local fire departments (currently Chandler and Tempe) by way of a dashboard. The FD then contacts at-risk vets and offers assistance. If it’s accepted, a home visit follows, with telehealth connection back to the VA, and department community paramedics provide interim care as needed to ameliorate risks. 

Determining who needs this outreach begins with a Care Assessment Need (CAN) score, a tool the VA uses to gauge a patient’s probability of hospitalization or death within a certain time frame. Then other criteria are weighted and added on: things like joint replacements, ED visits, a history of falls, etc. Vets are prioritized for contact based on these totals. 

Telehealth provides additional insight into their home environments, and on their visits community paramedics can help mitigate hazards like throw rugs, loose handrails, and broken assistive devices. 

It’s hard to count falls that haven’t happened because someone intervened beforehand, so the VA currently looks at surrogate measures like orders placed during home visits and hospital care for associated injuries like broken hips. “We believe they’re getting excellent care,” Muth said. “It’s just hard to document that in a meaningful way.” 

Besides preventing painful, expensive, and resource-consuming falls, the real benefit of the program, Muth said, is putting predictive analytics into action—making sophisticated real-life use of all that data we’re now amassing. Collecting data you do nothing with doesn’t do any good, he noted. But interestingly, over 2017, the numbers of patients seen under the program dropped. A total of 69 were visited in Chandler last year, 36 in Tempe.

Muth thinks that’s just because there aren’t that many veterans at fall risk in those two suburbs. Future plans may thus include expanding the effort to more cities. 

Back to Top