PRE-DETECT-HF

Early detection of decompensated heart failure using voice analysis

The challenge

Heart Failure (HF) is one of the major life-threatening manifestations of cardiovascular diseases and one of the leading cause of deaths on a global level. It is associated with high hospital readmission rates, morbidity, and mortality.[1]

Currently, the problem of HF is being addressed through a combination of medical interventions, lifestyle modifications, and patient management strategies. However, despite these current approaches, several challenges exist in managing HF:

Disease progression: HF is a chronic and progressive condition, and managing its progression can be challenging. The underlying causes may continue to worsen over time, impacting the outcome for the patient.

Hospital readmissions: HF patients experience frequent hospital readmissions due to decompensation (recurrence of heart failure symptoms) or other complications. This can affect patient quality of life and worsen their outlook. It also puts a strain on healthcare resources and increases healthcare costs. Many of the hospitalisations could be prevented.

Limited predictive capability: The current approach lacks highly accurate and real-time predictive models to identify patients sufficiently early who are at high risk of decompensation. The ability to proactively detect worsening symptoms or impending HF exacerbations could significantly improve outcomes.[2]

Patient adherence: Adherence to medication regimens, lifestyle modifications, and self-care management can be challenging for HF patients. Non-adherence may lead to poor symptom control, disease progression, and increased hospitalisations.[3]

Resource constraints: Access to specialised healthcare professionals, diagnostic tests, and advanced interventions may be limited, increasingly even in high-income countries. This can impede optimal care for HF patients, especially in rural areas with more pronounced resource constraints.[4]

The standard of care in HF monitoring is to keep track of patient’s symptoms and weight as an impending decompensation may be evident in a substantial weight increase within a few days. However, these indicators tend to signal inappropriately and very late.[5]

The solution

Lead partner and start-up Noah Labs proposes an innovative digital medical device – Noah Labs Vox – consisting of a voice-based machine learning (ML) model to predict HF-related decompensation based on voice analysis early and accurately.

Noah Labs Vox offers significant advantages over existing devices and solutions. By utilising advanced algorithms and data analytics, the device can accurately detect signs of decompensation several days before traditional indicators become apparent. This early prediction capability is crucial for timely intervention and proactive management of HF patients.

One key advantage of the device is its non-invasive nature. Unlike existing devices that may require implants or frequent blood drawing, the speech-based ML model enables remote monitoring and analysis of voice data. This non-intrusive approach enhances patient comfort and compliance while minimising the need for invasive procedures.

By leveraging the power of ML, the device can identify unique vocal biomarkers that may go unnoticed by conventional diagnostic methods. This provides healthcare professionals with new information to monitor and assess the progression of HF, enabling early intervention and preventive measures.

In addition to its other advantages, the device has the distinct advantage of not requiring any other wearable devices for monitoring. This eliminates the potential instability and variability associated with relying on multiple devices and sensors, thereby reducing the margin of inaccuracies in the measurements and analysis.

Expected impact

As part of this EIT Health project, the consortium is conducting a multi-centre observational study to clinically validate Noah Labs’ technology, aiming to improve algorithm accuracy and accelerate the solution’s market readiness.

The project promises several potential benefits for both healthcare systems and patients. It leverages voice analysis and advanced algorithms to remotely monitor heart failure (HF) patients, reducing healthcare burden and enabling earlier detection of complications.

The voice-based ML model developed within the project provides predictive insights into HF decompensation, assisting in resource allocation by identifying high-risk patients. This remote monitoring capability is expected to enhance cost-effectiveness by decreasing hospitalisations and emergency department visits.

Overall, the approach is anticipated to lead to a significant reduction in HF-related hospitalisations, lower mortality rates, and substantial cost savings for healthcare systems.

External Partners

• Noah Labs
• IPE Institut für Politikevaluation GmbH
• German Foundation for the Chronically Ill
• DKV Servicios
• ProductLife Group

References

[1] Savarese, G. et al. (2023). Global burden of heart failure: a comprehensive and updated review of epidemiology. Cardiovascular Research, 118(17), 3272-3287.

[2] Allen. L. A. et al. (2012). Decision Making in Advanced Heart Failure: A Scientific Statement From the American Heart Association. Circulation, 125, 1928–1952.

[3] Ruppar, T. M. et al. (2016). Medication Adherence Interventions Improve Heart Failure Mortality and Readmission Rates: Systematic Review and Meta-Analysis of Controlled Trials. Journal of the American Heart Association, 5 (6).

[4] Manemenn, S. M. et al. (2021). Rurality, Death, and Healthcare Utilization in Heart Failure in the Community. Journal of the American Heart Association, 10 (4).

[5] Stevenson, L. W. et al. (2023). Remote Monitoring for Heart Failure Management at Home: JACC Scientific Statement. Journal of the American College of Cardiology, 81(23), 2272-2291.