Listen instead
Every morning, a patient with chronic heart failure reads six prespecified sentences aloud into a smartphone. This brief vocal exercise, taking under a minute, is designed to capture subtle changes in speech patterns that indicate fluid accumulation in the lungs. It represents a shift toward software-only remote patient monitoring, avoiding the need for specialized home medical hardware.
On November 8, researchers led by William T. Abraham, MD, presented the results of the DETECT-HF clinical trial at the American Heart Association Scientific Sessions. The study, published simultaneously in JACC: Heart Failure, evaluated Cordio Medical's HearO system, a voice-analysis platform designed to identify early signs of worsening heart failure. While the technology successfully flagged oncoming clinical events weeks in advance, it missed a critical pre-specified statistical target, highlighting the challenges of translating speech analysis into a consistent diagnostic tool.
For healthcare systems, the stakes of remote monitoring are high. Heart failure remains a primary driver of hospital readmissions, costing the healthcare system billions of dollars annually. While existing invasive monitoring tools, such as pulmonary artery pressure sensors, are highly accurate, they require surgical implantation. Non-invasive alternatives, like daily weight scales and blood pressure cuffs, often suffer from poor long-term patient compliance. Voice-based monitoring offers a low-friction alternative, but only if the underlying algorithms can deliver reliable, actionable alerts without overwhelming clinical staff.
Clinical Evidence From DETECT-HF
The DETECT-HF trial was a prospective, international, multicenter, single-arm, blinded study that enrolled 563 patients diagnosed with symptomatic chronic heart failure. Participants recorded the six sentences daily on their personal smartphones. To evaluate the system's baseline diagnostic accuracy, the HearO system's analysis remained blinded to both patients and their physicians, meaning the results did not influence ongoing care.
During the monitoring period, 76 first heart failure events occurred. The HearO system identified 52 of these events, resulting in a clinical sensitivity of 68.4 percent. On average, these voice-based alerts occurred 24.1 days before the patients formally presented with worsening symptoms. This three-week window theoretically provides clinicians with sufficient time to adjust medications, such as oral diuretics, to prevent clinical deterioration.
However, the trial did not meet its pre-established statistical target for sensitivity. The 95 percent confidence interval for the system's sensitivity was 58.0 percent to 78.9 percent. Because the lower bound of this interval fell below the prespecified target of 60.0 percent, the study did not achieve its primary statistical goal. This margin suggests that the software may require further optimization to ensure reliable performance across diverse patient populations and recording environments.
Usability and Operational Metrics
For any digital health tool, patient adherence is as critical to clinical utility as diagnostic accuracy. In the DETECT-HF trial, patients demonstrated consistent engagement with the smartphone application, maintaining an average daily adherence rate of 85 percent. On a usability scale of 1 to 5, patients rated the platform between 4.7 and 4.9, indicating that the daily task was easily integrated into their routines.
Importantly, the trial reported zero device-related adverse events, confirming the safety of the non-invasive software. However, the system generated an average of 2.65 unexplained alerts per patient-year. In a clinical setting, unexplained alerts can lead to alarm fatigue among nursing staff and physicians. If a clinic manages 500 heart failure patients using this technology, staff would need to triage more than 1,300 unexplained alerts annually, highlighting the need for structured clinical response protocols.
The Path to Commercialization
Cordio Medical's HearO system currently holds an FDA Breakthrough Device Designation but is not yet cleared for commercial marketing in the United States. To secure regulatory clearance, the company must demonstrate that the software's clinical benefit outweighs the operational burden of its false-positive rate.
Beyond regulatory clearance, the primary hurdle for voice-based diagnostics is securing reimbursement from insurance providers and Medicare. To achieve widespread clinical adoption, developers must prove that acting on early voice alerts directly reduces hospitalization rates and overall care costs. An alert delivered 24 days in advance is only valuable if it triggers an effective, structured intervention that keeps the patient out of the hospital.
Source: The HealthTech Signal
