SALT LAKE CITY—More veterans leave the hospital with a diagnosis of heart failure than any other condition, making improving care for this rapidly growing segment of the population a high priority. A critical step is keeping veterans with heart failure out of the hospital in the first place.

“As a chronic illness, heart failure is characterized by acute worsening that often requires hospitalization. There are multiple reasons to reduce the rate of hospitalization,” said Josef Stehlik, MD, MPH, medical director of the heart failure and heart transplant program at the George E. Wahlen Medical Center in Salt Lake City and co-chief of the advanced heart failure program at the University of Utah.

“As heart failure worsens, patients become short of breath and bedbound. If they’re admitted to the hospital, deconditioning begins very fast, even with efforts to have patients moving and in rehabilitation. It’s just not the same as being at home,” he told U.S. Medicine.

Patients admitted to the hospital also have a high risk of being readmitted, exacerbating the deterioration in their physical condition. Twenty percent of hospitalized heart failure patients return to the hospital within 30 days of discharge, and 30% are readmitted within 90 days.

Stehlik and colleagues at the University of Utah and elsewhere have been evaluating a system that has the potential to significantly reduce those readmissions.

7-10 Days’ Warning

In a recent study published in Circulation: Heart Failure, they describe a system that links a chest sensor to artificial intelligence technology that can identify changes in the heart well before heart failure patients experience a crisis, giving valuable time for intervention to avert hospitalization.1

The system uses a disposable silicon sensor that adheres to the chest. The sensor sends information on the patient’s movement, electrocardiogram and respiration rate to an AI platform developed by physIQ via the patient’s smartphone.

In the study, 100 patients discharged from VAMCS in Salt Lake City; Houston; Palo Alto, CA; and Gainesville, FL, were given the sensors to wear 24 hours per day for as long as three months. The AI established the patient’s baseline characteristics—heart rate, heart rhythm, respiration, walking, sleep, body posture and more—over the first two to three days. Unexpected deviations from the norm indicated potential worsening of the patient’s heart failure.

The system correctly predicted the impending need for hospitalization or emergency department visit with sensitivity of up to 88% and specificity of 86%. While the study did not include interventions, the system predicted hospitalization on average 10.4 days before it occurred with a median of 6.5 days of advanced notice.

“Continuous sensor data has enormous potential to transform how we understand trajectories of human health; but these data streams present multiple challenges,” explained study co-author Matthew Pipke, JD, chief technology officer of physIQ. “Not only is the volume of data massive, but the behavior of physiology in the unconstrained ambulatory setting is rich and complex. Properly validated machine learning and deep learning analytics, like those tested in this VA study, are the key to effectively processing wearable biosensor data to provide clinical users with information they can act on.”

“This study is one of the first to actually successfully apply machine learning approach to use heart rate, respiration rate and body motion to elevate a signal from the noise,” in heart failure, Pipke told U.S. Medicine.

“Now that we have shown that the system can accurately make predictions, we can work on implementing it into clinical practice,” Stehlik said. “Otherwise it won’t help the patients.”

Veteran Approved

The AI platform can work with a number of different sensors and over the study period, several different ones were used. Since the first feasibility testing in 2013, the chest sensors have gotten smaller, and their batteries have lasted longer.

At least one veteran found even the larger size used five years ago to be no problem and the information provided quite helpful.

Veteran Frank Anderson was in the Salt Lake City VAMC for consultation prior to a heart transplant when Stehlik asked whether he’d like to participate in the study.

“I said, ‘Sure.’ It was a way for me to give back to the VA for what they’re doing for me,” Anderson told U.S. Medicine.

Anderson had the sensor patch for the four months before his transplant took place. “When walking, I could tell if my heart rate was going up or not. If I noticed that my heart was running faster, I knew to slow down,” he noted. “It wasn’t at all inconvenient. I had to change the battery on a daily basis. That was all.”

Today, the single-use sensors can be worn for five days and then replaced, so there’s no need to change the battery.

A few weeks after Anderson enrolled in the study, a problem with his heart took him to another hospital. “Everybody asked what I was wearing and said, ‘We want that, too,’” Anderson said, laughing. “Nurses, doctors, everyone was very interested in the study. I’m so glad I could participate in it.”

Next Steps

In the next phase, the researchers plan to have an alert transmitted from the system to the heart failure team when a significant deviation indicates an emerging problem. The team hopes that a follow on study will be funded that will enable them to conduct a clinical trial that would randomize several hundred veterans to intervention based on data from the system or nonintervention. The VA’s Innovation Ecosystem funded the current study.

“We’re designing a statistical algorithm to respond to changes in the patient that clinicians can use to guide care,” Stehlik noted. “One step further in the future, we hope to learn what kinds of responses work under different circumstances based on the data coming in and response to treatment in the past. That would allow us to apply the algorithm to provide a more precise response for a larger number of patients without increasing the burden on physicians.”

  1. Stehlik J, Schmalfuss C, Bozkurt B, Nativi-Nicolau J, Wohlfahrt P, Wegerich S, Rose K, Ray Ranjan, Schofield r, Deswal A, Sekaric J, Anand S, Richards D, Hanson H, Pipke M, Pham M. Continuous Wearable Monitoring Analytics Predict Heart Failure Hospitalization: The LINK-HF Multicenter Study. Circ Heart Failure. 25 Feb 2020;13(3):e006513. DOI: 10.1161/CIRCHEARTFAILURE.119.006513