Heart failure is a serious condition and a major reason why older adults are admitted to the hospital. Even after going home, many patients struggle to get the follow-up care they need, leading to high rates of return to the hospital.

While remote monitoring technology exists, it is often difficult to use, expensive, or hard for doctors to interpret.

Recent research suggests that changes in a person's voice and speech can be early warning signs of worsening heart failure, but we need to test if this can work reliably in a home setting.This study tests a new, user-friendly platform called AVATAR-SC.

This system uses a a computer-generated character to interact with patients. By analyzing short voice clips and simple health questionnaires provided by the patient, the system aims to make home monitoring easier and more effective.

The study will involve 60 participants divided into two groups:half of the patients who have recently been discharged from the hospital following a heart failure episode; the other 30 patients who visit the hospital monthly for a specific heart failure treatment (Levosimendan).

Participants will interact with the digital avatar twice a week. During these sessions, they will provide brief voice recordings and answer a few questions about how they feel. It is important to note that the AI in this study is not making medical decisions or changing the patient's current treatment; it is strictly being tested to see how well the technology functions.

As a feasibility study, the primary goal is to evaluate the practical implementation of the technology rather than clinical outcomes.

The research focuses on: Determining if patients can navigate the system easily and independently, Assessing how consistently participants engage with the scheduled sessions, Verifying the technical reliability of the platform in a home environment,identifying the strengths and challenges of the system to prepare for future, larger-scale trials.

Benefits: Participants may feel more supported and more aware of their symptoms through regular check-ins.

Risks: The risks are very low. They mostly involve getting used to new technology and ensuring that personal data is kept private and secure.

Impact If successful, AVATAR-SC could lead to a new way for doctors to keep an eye on heart failure patients at home, catching problems early through the sound of their voice and preventing unnecessary hospital visits.

Heart Failure (HF) is a leading cause of hospitalization and mortality among the elderly, characterized by high readmission rates and a significant impact on the National Health Service. Despite guideline recommendations, only a minority of patients receive timely post-discharge follow-up, contributing to clinical deterioration and rising healthcare costs. Existing remote monitoring systems have shown mixed results, primarily due to poor patient adherence, data interpretation challenges, and high costs. While vocal and linguistic analysis supported by Artificial Intelligence (AI) algorithms has shown promise for the early identification of heart failure, its practical application in home monitoring has yet to be systematically evaluated.

AVATAR-SC proposes an accessible, non-invasive platform based on advanced AI technologies. It integrates vocal biomarkers, clinical data, and Patient-Reported Outcome Measures (PROMs) to enhance home monitoring and support a more proactive management of heart failure patients.

Study Population: 60 patients with Heart Failure (HF), including 30 patients discharged following hospitalization for acute heart failure and 30 patients receiving monthly Levosimendan therapy in a Day Hospital setting.

Study Objectives: Although AI-supported vocal and linguistic analysis shows significant potential for the early detection of clinical deterioration in HF patients, there is a lack of data regarding feasibility, acceptability, adherence, integration into care pathways, and data quality within a real-world home monitoring context. A feasibility study is therefore necessary to:i) evaluate patients' ability to use AVATAR-SC independently;ii) estimate adherence to daily vocal data collection;iii) verify the technical performance of the platform;iv) define operational metrics (time, resources, completion rates);v) identify barriers and facilitators in preparation for a large-scale clinical trial.The study does not involve the use of AI systems as clinical tools, nor their application for the automated assessment of an individual patient's health status.Nature and extent of benefits and risks associated with study participation: Potential benefits include more regular symptom monitoring and increased support in disease management. Risks are minimal and primarily concern the use of technology and the processing of personal data.

Intervention: Interactions via digital avatar ≥ 2 times/week, collection of short voice clips, and essential PROMs (Patient-Reported Outcome Measures). No alerts will be generated, and there will be no influence on the patient's clinical pathway.

Inclusion Criteria

• Age ≥ 18 years
• Diagnosis of heart failure (NYHA Class II-IV), according to the Diagnosis Related Group (DRG) classification.
• Recent discharge following acute heart failure (AHF) or ongoing treatment with Levosimendan in a Day Hospital setting.
• Provision of signed informed consent in Italian.
• Currently receiving Guideline-Directed Medical Therapy (GDMT) for heart failure

Exclusion Criteria:

• Inability or incapacity to provide informed consent.
• Cognitive impairment severe enough to prevent interaction with the avatar and compromise the reliability of the collected data.
• Significant phonation disorders (e.g., results of previous otorhinolaryngological surgery or post-traumatic outcomes).
• Inability to manage a compatible device or the device provided by the study for interaction with the AVATAR-SC platform.
• Life expectancy of less than 6 months due to concurrent comorbidities.
• Technological or language barriers.

Individual Participant Data (IPD) will not be made publicly available to ensure strict compliance with EU Regulation 2016/679 (GDPR) and to protect patient privacy regarding sensitive vocal recordings. All collected data, including audio tracks and clinical features, are processed in a pseudo-anonymous form using non-reversible alphanumeric codes. The link log connecting these codes to patients' identities is stored exclusively at the coordinating center (ASST GOM Niguarda) and is not accessible to other partners or third parties.