Smartphone speech analysis detects worsening heart failure prior to an event
Click Here to Manage Email Alerts
NATIONAL HARBOR, Md. — A novel speech analysis system was able to detect impending heart failure events in ambulatory patients with congestive HF more than 3 weeks ahead of time, a speaker reported.
The system (HearO, Cordio Medical) was not only able to predict HF events with greater sensitivity compared with daily weight and oxygen saturation monitoring, but also demonstrated fewer average false positives per patient per year, according to the results of the HearO community study presented at the Heart Failure Society of America Annual Scientific Meeting.
“The energy generator of speech are the lungs. We look at this analysis technology as being primarily reflective of changes in lung fluid content or pulmonary congestion, but it’s worth noting that hydration may affect speech parameters at various levels, including the soft tissues in the vocal tract as well as the vocal folds,” William T. Abraham, MD, professor of internal medicine and board-certified advanced heart failure and transplant cardiology physician at The Ohio State University Wexner Medical Center, said during a presentation. “The Cordio HearO community study is a multicenter, noninterventional, single-arm, open clinical study being performed outside of the U.S., primarily in Israel, and the goal here was simply to explore the diagnostic performance in ambulatory heart failure patients.”
Abraham stated that the front end of the system is a smartphone app, able to function on an Android or iOS device, which asks the patient each day to speak five preset sentences into the phone. From there, it is uploaded into a HIPAA-secure cloud-based system where the speech analysis and comparison with the stable baseline data occurs.
In a prior feasibility study of the system published in JACC: Heart Failure, researchers assessed whether speech measures identified using the speaker verification, speech processing and smartphone app algorithm would predict clinical states of pulmonary congestion indicative of HF events in patients with acute decompensated HF.
At hospital admission and discharge, patients were asked to record five sentences, repeated three to four times each in their native language.
As Healio previously reported, 94% of cases were identified as distinctly different from baseline and in 87.5% of cases, distinct differences from baseline were detected in all five speech measures.
For the present analysis, 253 ambulatory patients with congestive HF were enrolled for the HearO community study (mean age, 69 years; 19% women; 37% with left ventricular ejection fraction < 40%).
Each patient was asked to complete five voice recordings per day for 2 years. Among them, the primary language was 72% Hebrew, 19% Russian, 7.5% Arabic and 1.6% English.
Abraham said the speech processing system creates a patient-specific speech model while they are in a clinically stable state and compares that model going forward against daily speech samples to detect changes thought to be related to changes in lung fluid content.
The primary outcome was HF events, defined as HF hospitalization and treatment with IV diuretics.
Abraham reported that the HearO system could predict approximately 80% of HF events on average 22.5 days before the events occurred.
Researchers observed a false-negative rate of approximately 18%.
Over a total number of 186,152 analysis days, the system’s false-positive rate was on average 2.8 alerts per year, translating to one false positive every 4.3 months per patient per year.
The HearO system demonstrated greater sensitivity for the detection of impending HF events when compared with current standards for detection such as daily weight (80% vs. 35%) and oxygen saturation measurements (80% vs. 25%) and had fewer false positives per patients per year.
Moreover, researchers observed a high rate ( 70%) of compliance to daily voice recording among trial participants.
“Speech analysis technology may be a useful tool in remote monitoring of heart failure patients, providing early warning of worsening heart failure, including impending decompensation resulting in hospitalization,” Abraham said during the presentation. “Ongoing and future studies will continue to define the role of speech measures in the management of heart failure, and I think this approach has the potential to reduce [acute decompensated] HF hospitalizations and improve patient quality of life and economic outcomes. But of course, we’ll need to show that in a subsequent prospective and randomized controlled trial.”
Abraham added that HearO is currently being evaluated in a new U.S. trial that will test the system’s sensitivity across various dialects, accents and languages representative of the general U.S. population.
Perspective
Back to Top
Debra Moser, PhD, RN, FAHA, FAAN
That was probably one of the most fascinating trials because it gets at the main problem of trying to figure out, early on, when people are congested because they’re either not good at perceiving their own symptoms or they’re not adherent to daily weighing, which is not that sensitive anyway. The other options are implantable hemodynamic monitoring devices that are invasive. It would be fabulous to have this app with patients talking to it a few times a day to be able to detect congestion. It looks like it can detect it 3 weeks ahead of an acute exacerbation. It could really keep people out of the hospital.
It’s an observational study and needs a randomized control trial, but I think it’s a promising device that could change the way we detect congestion early on.
Perspective
Back to Top
The voice analysis approach may be a novel way to do remote monitoring. The current methods we have are so crude in terms of daily weights, etc, that perhaps technology will lead the way. Decompensation is a huge problem that we have in this country, but if you could find those patients early, intervene and abort that admission, it would be a huge advance. This is an early study of a technology, but it’s very exciting and I’m looking forward to seeing additional studies from that group.
Collapse
Abraham W, et al. Late Breaking Clinical Trials Session I. Presented at: The Heart Failure Society of America Annual Scientific Meeting; Sept. 30-Oct. 3, 2022; National Harbor, Maryland (hybrid meeting).
Click Here to Manage Email Alerts