A new study, sponsored by the U.S. Department of Veterans Affairs, applied wearable biosensors to post-acute heart failure patients and deployed FDA-cleared analytics from vendor physIQ to detect vital sign anomalies.

It demonstrated promising predictive power for artificial intelligence-based analytics in terms of sensitivity, specificity and early warning lead time – suggesting the potential to transform to a proactive, personalized care model for at-risk patients, investigators from the Utah School of Medicine and the VA Salt Lake City Health Care System announced.

[Also: Microsoft says AI and cloud computing are giving precision medicine a boost]

The report, "Continuous Wearable Monitoring Analytics Predict Heart Failure Decompensation: The LINK-HF Multi-Center Study," evaluated how AI-based personalized physiology analytics could be applied to wearable biosensor data to predict when a patient might be at risk of hospitalization or an acute care event.

The multicenter observational study enrolled 100 heart failure patients across four VA hospitals prior to hospital discharge and provided them a 90-day supply of wearable biosensors from VitalConnect and a smartphone to continuously transmit patient physiologic data to the physIQ cloud.

[Also: Department of Veterans Affairs looks to private developers for innovation]

AI-based, FDA-cleared personalized physiology analytics from physIQ then applied machine learning algorithms to the multivariate set of continuous vital sign data to learn the individual's dynamic vital sign patterns and establish a personalized baseline.

From this baseline, the analytics detected subtle changes that may be an indicator of deteriorating health or a precursor to an acute event. Retrospective analysis of the patient records in this study evaluated the analytics' ability to detect vital assign anomalies that corresponded with documented clinical events and rehospitalizations.

"The results of this study suggest a highly favorable relationship between sensitivity and specificity of event detection, as well as a sufficient warning lead time for clinicians tasked with managing patients at risk for admission for heart failure exacerbation," said Josef Stehlik, MD, principal investigator of the study.

"Furthermore, the study demonstrated that patients were highly compliant with wearing the biosensors throughout the monitoring period, making this a promising solution for a patient population with significant unmet clinical needs," he said. 

There currently are more than 6 million people in the United States diagnosed with heart failure and one million annual hospitalizations, according to the American Heart Association. Among them, approximately 20 percent of patients are readmitted within 30 days of hospital discharge, underscoring the challenges both clinicians and patients face with trying to manage this disease, according to a report in the Journal of the American College of Cardiology.

"Heart failure continues to be a major challenge in healthcare today," said Stephen Ondra, MD, chief medical officer at physIQ. "Sadly, there has been far too little improvement over the past 15 years with respect to our ability to proactively manage acute deterioration. As a result, healthcare has struggled to adequately improve patient quality of life and reduce the exorbitant costs associated with this chronic disease."

The results of this study are very encouraging as they show tremendous promise in an approach that supports a proactive care delivery model – one where clinicians can see personalized changes early enough in the deterioration process to take action that can potentially head off an acute or even catastrophic clinical event, Ondra added.

"This could be an enormous benefit to patients, providers, health systems and payers as they embrace value-based care," he said.

Twitter: @SiwickiHealthIT
Email the writer: bill.siwicki@himssmedia.com