Key Points
- Novel speech processing technology has the potential to detect heart failure exacerbations before they happen, and may be better than the conventional weight monitoring approach.
- Additional studies are needed to explore the generalizability, practicality, efficacy, and effectiveness of the HearO model.
Speech processing technologies are ubiquitous, and speech parameters can be affected by hydration status and lung fluid levels. HearO technology purports to be able to detect changes in speech parameters over time that are related to lung fluid, independent of the language spoken. This technology creates a patient specific speech model while clinically stable and compares the model against daily speech samples; in this manner, patients act as their own controls. While there may be the potential for this approach to predict worsening pulmonary congestion from heart failure (HF), it has yet to be developed and tested in a clinical setting. On November 12, 2023 the results of the “SPEECH: Validation of a Speech Analysis Application to Detect Worsening Heart Failure Events in Ambulatory Heart Failure Patients” were presented at AHA Scientific Sessions 2023 on behalf of the Cordio HearO Community Study Group.
The purpose of this study was to use the HearO system to analyze ambulatory HF patients’ speech data and to refine and test its ability to detect impending HF worsening. They included outpatient NYHA Class II-III adult heart failure patients regardless of ejection fraction. All patients had baseline creation of their personalized voice model at enrollment. Participants then were instructed to take voice recordings daily at home every morning. HF worsening events were notes, and the voice recordings were analyzed retrospectively. One group was used for development, and the other group was used for testing. There were 362 HF patients in the development group, with 158,024 total recording days and 58 HF events (i.e. worsening requiring hospitalization or outpatient intravenous diuretic therapy). There were 153 HF patients in the test group, with 94,202 recording days and 14 HF worsening events. Patients were in their late 60s, and the majority in both groups spoke Hebrew, with very few speaking English. The mean ejection fraction was ~ 37%. Overall compliance with daily recording was 82.5%. In the test group, the HearO system was able to detect 71% of impending HF worsening events, on average 26 day prior to the event. In the development group, the sensitivity of HearO to detect worsening HF events was superior to daily weight monitoring.
Dr. William T. Abraham of Ohio State University concluded: “The sensitivity detection rate of HearO speech analysis-based telemonitoring was superior to that of daily weight monitoring. Novel speech analysis technology may be useful in remote monitoring of HF patients, providing early warning of worsening HF events including impending decompensation resulting in hospitalization…Ongoing and future studies, including an ongoing US pivotal trial, will continue to define the role of speech measures in the management of heart failure patients.”
