This study will enroll heart failure (HF) patients who are under active management with an implanted pulmonary artery pressure sensor (CardioMEMS). Subjects will be provided an electronic stethoscope (the Eko DUO) to take at-home heart sound, lung sound, and ECG recordings in conjunction with regimented CardioMEMS measurements. These two datasets will be used to confirm whether an AI/ML model to track HF status can be developed.

Heart failure (HF) affects an estimated 6.2 million Americans over the age of 20 and carries a very high healthcare system burden worldwide. Annual costs for HF management in the United States were estimated at $30.7 billion in 2012 and are projected to increase to $68.7 billion by 2030. The primary cost driver for HF management is a high rate of acute decompensation and subsequent hospitalization. The mean per-patient cost of an HF-related hospitalization is estimated to be $14,631.

Among the conditions that the Centers of Medicare and Medicaid Services (CMS) monitors for their Hospital Readmission Reduction Program, HF has the highest median readmission rate at days 1-29 (23%) and days 1-60 (11.4%) postdischarge. The cost burden of HF readmission is $2.7 billion in 2013. A meta-analysis from 2012 estimated that 23.1% of HF readmissions are avoidable, although individual studies ranged from 5% to 79%. Many health plans, including CMS, have focused on interventions that monitor patients for early detection of HF decompensation. Earlier interventions can help care teams prevent avoidable hospitalizations.

Invasive hemodynamic sensor devices have enabled HF care teams to better predict and prevent HF decompensation events, and thus prevent rehospitalizations. One such device is the CardioMEMS pulmonary artery (PA) sensor (Abbott Inc., Atlanta, GA, USA). The CardioMEMS is implanted in a branch of the left PA, allowing for daily measurements of PA pressures. PA pressures are used as a surrogate marker of filling pressure, and rising filling pressures, in turn, are a marker that precedes the exacerbation of HF. The CHAMPIONS trial demonstrated that remote diuretic management using CardioMEMS reduced HF all-cause hospitalizations by 43% and mortality by 57%. Unfortunately, CardioMEMS as an HF solution is invasive, costly (average sales price of $17,750), indicated for a restricted patient population (NYHA class III HF who have been hospitalized within the last year), and has limited reimbursement coverage due to equivocal cost-effectiveness projections.

This has stimulated a search for less expensive, non-invasive sensors that may correlate with fluid status in HF patients. A study in Taiwan demonstrated that outpatient therapy guided by an inpatient device with ECG and sound sensors reduced post-discharge HF utilization by 31% when compared to a control group using symptoms to guide therapy. The LINK-HF study demonstrated that a wearable patch with ECG and sound sensors could predict HF readmissions with a sensitivity of 76% to 88%, a specificity of 85%, and a median lead time of 6.5 days.

Despite these initially promising results, however, these devices have significant disadvantages. The inpatient device used in the Taiwanese study could not be adapted into a portable form factor for outpatient use. Wearable devices can be rigid, uncomfortable, and highly visible, all of which can interfere with patient function and decrease monitoring compliance.

Therefore, there remains an unmet clinical need for a widely available, non-invasive, affordable medical device that can estimate an HF patient's hemodynamic fluid status and inform the HF care management team. The ultimate goal remains to decrease an HF patient's risk for hospital readmission, all from the comfort of the patient's home.

To meet this need, Eko has developed the DUO, an FDA-cleared, portable, hand-held, wirelessly connected medical device with ECG and sound sensors. Data from the DUO can be wirelessly streamed to a mobile phone or tablet, which can then be transmitted to a HIPAA-compliant internet cloud infrastructure for storage and analysis. In 2020, Eko introduced into the US market a package of AI/ML algorithms that follows this workflow to identify heart murmurs, atrial fibrillation, and other cardiac conditions, and intends after this proof of concept study to build upon this platform to estimate and trend PA pressures.

But beyond measuring and trending PA pressures, the DUO can be used to capture additional important HF features that will further improve any HF algorithm's performance. For example, because patients with decompensated HF often have an audible third heart sound, characteristic ECG findings, and altered time interval durations between their heart sounds and ECG signals, the Eko DUO device may be uniquely positioned to detect these types of changing signals.

In addition, because heart failure and fluid overload are reflected in the lungs as crackles (and occasionally effusions), the lung examination is and has always been a cornerstone of the overall physical examination of HF patients. By using the DUO to capture lung sounds in patients with HF, and comparing not only the presence or absence of crackles, but also how these adventitious sounds change over time, we will be able to explore the utility of the Eko DUO in helping to predict exacerbated HF.

This proof-of-concept study evaluates the feasibility of the Eko DUO in capturing and measuring signals relevant to HF exacerbation (e.g., time intervals, adventitious lungs sounds, pathologic heart sounds), as well as the feasibility of developing an AI/ML algorithm to model PA pressures in HF patients with the implantable CardioMEMS device.

Inclusion Criteria:

• Aged 18 years and older
• Patient or healthcare proxy willing to give written informed consent to participate
• Presence of an implanted CardioMEMS device or imminent implantation of a CardioMEMS device
• Expressed willingness to take DUO recordings immediately before or after taking their CardioMEMS measurements, on the same schedule prescribed by their physician
• Functioning iOS or Android smartphone or tablet that can download and run the companion Eko application
• Access to WiFi or cellular data connection at home

Exclusion Criteria:

• Patient or healthcare proxy is unwilling or unable to give written informed consent
• Patient is enrolled in another study that may interfere with the observations from this study
• Acute pericarditis
• Healing chest wall wounds (e.g., sternotomy or thoracotomy)

Patients will be identified via the medical center's registry of patients who are actively managed with CardioMEMS. Alternatively, the center's schedule of CardioMEMS implantations will be monitored for potential subjects with newly implanted CardioMEMS.