The goal of this clinical trial is to learn if using a WhatsApp-based chatbot called PROMBot-FA can help people recover after atrial fibrillation ablation. The main questions it aims to answer are:

• Is the chatbot easy and practical for people to use after their procedure?
• Do people find the chatbot helpful for sharing their symptoms and health updates?

Researchers will compare follow-up care supported by the chatbot to standard hospital follow-up visits. Participants will:

• Be randomly assigned to either use the chatbot or receive usual follow-up care.
• Report their symptoms, well-being, and any problems after ablation for 3 months.
• Complete short questionnaires about their experience using the chatbot.

This study will help researchers understand whether chatbot follow-up is feasible and acceptable for people after atrial fibrillation ablation and guide future larger studies.

The primary objective of this pilot study is to evaluate the feasibility and acceptability of integrating PROMBot-FA, a WhatsApp-based chatbot platform for collecting PROMs and clinical parameters, into routine post-ablation follow-up care at a tertiary hospital. Specific aims are to (1) assess recruitment, retention, and engagement rates during the first 3 months following AF ablation; (2) evaluate the acceptability and usability of the chatbot among patients and the clinical dashboard among healthcare professionals; and (3) explore preliminary effects on clinical outcomes, health-related quality of life, and integration into existing clinical workflows. This pilot trial will generate essential data for refining the intervention and designing a definitive effectiveness trial.

Methods Study Design This study is a single-center, parallel-group, randomized, open-label pilot trial. Patients undergoing catheter ablation for AF will be randomly assigned in a 1:1 ratio to receive either chatbot-supported follow-up through the PROMBot-FA platform (intervention group) or standard care (control group). Both groups will complete standardized baseline and 3-month follow-up assessments.

Ethical Considerations The study protocol was submitted to the Ethics Committee of Hospital de Santa Marta, Unidade Local de Saúde de São José (approval number pending). All participants will provide written informed consent before enrollment. The study will be conducted in accordance with the Declaration of Helsinki and complies with the European Union General Data Protection Regulation. Participants may withdraw from the study at any time without impact on their clinical care. Data will be pseudonymized and stored securely with access restricted to authorized study personnel.

This study is conducted within the scope of the CardioCausalAI project, funded by the Foundation for Science and Technology, Portugal (reference 2024.07369.IACDC), which develops causal artificial intelligence models for cardiovascular care.

Study Setting The study is conducted at the Department of Cardiology, Hospital de Santa Marta, Lisbon, Portugal. This tertiary referral center specializes in cardiothoracic care and performs a high volume of cardiac ablation procedures annually. Since 2011, the hospital has implemented a structured 1-year follow-up program involving periodic telephone consultations, creating a strong organizational foundation for digital health integration.

Participant recruitment, informed consent procedures, and clinical assessments occur at the hospital. The intervention is delivered remotely, with patient-chatbot interactions occurring in participants' home environments. Clinical coordination and data monitoring are centralized at the hospital to ensure integration with standard care pathways.

Participants Inclusion criteria

• age 18 years or older;
• documented diagnosis of paroxysmal or persistent AF based on clinical assessment and coded according to the International Classification of Diseases, 11th Revision;
• undergone catheter ablation for AF at Hospital de Santa Marta during the study recruitment period;
• able to use the PROMBot-FA chatbot independently or with caregiver assistance, defined as Portuguese literacy sufficient to understand instructions and ability to operate a smartphone;
• clinically stable at discharge with no acute post-procedural complications precluding participation; and
• provision of written informed consent.

Exclusion Criteria

• severe cognitive impairment preventing comprehension of study procedures, such as moderate-to-severe dementia or other neurological conditions limiting capacity to consent;
• severe uncorrected visual impairment or motor disability preventing smartphone operation even with caregiver support;
• explicit refusal to use digital technology or unwillingness to share health data required for the study;
• clinical or contextual conditions rendering 3-month follow-up unfeasible, including anticipated relocation outside the hospital catchment area or concurrent participation in another interventional study with overlapping endpoints; or
• life expectancy less than 6 months.

Assessment of digital literacy will be conducted individually. Older adults with limited prior experience using smartphone applications may be enrolled if they demonstrate willingness to learn and have an available caregiver willing to provide assistance. All participants and caregivers will receive structured training and ongoing technical support.

Sample size The sample size was calculated to achieve 80% statistical power at a significance level of 0.10, using a test for proportions for two independent groups. Assuming a 20% incidence of AF recurrence in the control group and an expected reduction to 5% in the intervention group, 30 participants per group were initially estimated to ensure adequate power. This estimate was derived post hoc using Fisher's exact test. To account for an anticipated 20% dropout rate, the sample size was adjusted to 38 participants per group, for a total of 76 participants.

Randomization and Allocation Concealment Participants will be randomly assigned to the intervention or control group using computer-generated random numbers in permuted blocks of variable size (4 to 6) to maintain allocation concealment. Randomization will be stratified by AF type (paroxysmal vs persistent) to ensure balance between groups. Allocation sequences will be generated by a statistician not involved in participant recruitment using R statistical software version 4.5.0 (R Foundation for Statistical Computing).

Allocation concealment will be maintained through sequentially numbered, opaque, sealed envelopes prepared by personnel independent of the recruitment process. Envelopes will be opened only after eligibility confirmation and completion of baseline assessments.

Blinding Due to the nature of the intervention, it is not feasible to blind participants or healthcare professionals providing care. However, outcome assessors conducting medical record reviews for clinical endpoints (eg, hospitalizations, cardiovascular events) will be blinded to group allocation where possible. Statistical analyses will be conducted by a researcher blinded to group assignment until completion of primary analyses. Participants will be instructed not to disclose their group allocation during follow-up assessments conducted by blinded assessors.

Procedures Preparation and Training The research team will configure the PROMBot-FA chatbot (operating via WhatsApp) and related remote monitoring platform, including definition of finite-state dialogue flows with AF-specific questions, message content, and input-validation rules. Pilot testing with 2-3 test users will also be conducted to identify usability issues and to verify delivery, latency, and data integrity. Health professionals involved in recruitment and follow-up will receive training on study procedures, including informed consent, platform registration, and use of the clinician dashboard to monitor incoming data.

Recruitment and Baseline Assessment Eligible patients undergoing AF ablation will be approached during their hospital stay. After receiving written study information, discussing questions, and providing informed consent, participants will complete a baseline assessment that will include demographics, medical history, recent clinical parameters (e.g., blood pressure and heart rate), and health-related quality of life (Atrial Fibrillation Effect on Quality-of-Life [AFEQT] - European Portuguese validated version). For participants randomized to the intervention group, PROMBot-FA will be set up before discharge by confirming WhatsApp is installed on the participant's smartphone, registering the participant in the remote monitoring platform, linking their mobile number to PROMBot-FA, and sending a confirmatory WhatsApp message to verify correct enrollment. Patients (and caregivers, if applicable) will receive a brief demonstration and training to ensure they can interact with the chatbot autonomously, and will also receive a leaflet with a brief explanation of the chatbot

Remote Follow-up Participants in the intervention group will engage with PROMBot-FA via WhatsApp for 12 weeks. The chatbot will request self-reported health metrics (e.g., body mass, blood pressure, heart rate, mass, step count, AF-related symptoms) and general well-being, supplemented with brief educational prompts. Supplementary home-monitoring devices (e.g., a digital sphygmomanometer) will be provided if needed. In the first month, chatbot interactions will occur daily, focusing on post-ablation symptom surveillance and medication adherence. After that, the cadence will taper to reduce burden in later weeks while maintaining essential monitoring. Chatbot prompts will remain open for 24 hours, after which nonresponses will be recorded as missing data. A study nurse or physician will review patient entries weekly and may proactively contact participants reporting concerning symptoms or lapses in engagement. PROMBot-FA will not replace routine follow-up; all patients will continue standard clinical care, with chatbot data available to the study clinical team to complement or inform consultations at the treating cardiologist's discretion.

Final Evaluation At the end of the 3-month follow-up period (three months), all participants will undergo post-intervention assessments during a routine hospital visit. If necessary, these assessments will be conducted via secure videoconference. The patient-reported outcomes will focus on the effect of AF on health-related quality of life. Participants in the intervention group will also complete validated measures of usability and perceived usefulness. They will also take part in a brief semi-structured interview to share their experience of using the chatbot.

Semi-structured interviews will also be conducted with healthcare professionals directly involved in patient care (eg, cardiologists, nurses, and technicians). These interviews are intended to elicit qualitative insights regarding their experience with the chatbot-based intervention, as well as their perceptions of its usability, clinical relevance, and impact on workflow integration.

Secondary Data Extraction and Study Closure Clinical records (e.g., emergency visits, hospital readmissions, additional consultations, diagnostic tests) will be extracted from hospital systems to complement primary outcomes, covering both the 3-month follow-up and the 6 months preceding ablation. After all data are collected, the study will be closed and prepared for analysis.

Technical Description of the Chatbot The PROMBot-FA chatbot is a rule-based, finite-state-machine chatbot delivered via WhatsApp, supporting either numeric, multi-choice, or free-text inputs, with branching and input-validation rules. It will serve as the primary data-collection instrument for patient-reported outcomes and physiological measures during post-ablation follow-up, including body mass (kg); blood pressure (systolic/diastolic, mmHg), heart rate (bpm), daily step count from the patient's smartphone or smartwatch, adherence to therapy, and a structured symptom checklist (palpitations, dyspnea, chest pain, dizziness/syncope, fatigue, changes to the punction area, peripheral edema). An additional questionnaire is generated if the participant reports certain symptoms, so that the clinician can have access to more information regarding that symptom. All items are time-stamped and centrally stored in the study database in their native formats (continuous numerics for body mass, blood pressure, heart rate, steps; categorical for symptoms).

Primary Outcomes (Feasibility) The primary outcomes of this trial are feasibility parameters, including recruitment rate (eligible patients providing consent), retention rate (participants completing the 3-month follow-up), and engagement rate (chatbot prompts responded to within 24 hours).

Clinical outcomes include AF recurrence within three months post-ablation, defined as documented symptomatic relapse or the requirement for antiarrhythmic therapy, cardioversion, or repeat ablation.

Post-ablation complications comprise unplanned hospital readmissions, emergency department visits, thromboembolic events, major bleeding, or cardiac tamponade. Additional outcomes are all-cause mortality within twelve months, the number of cardiology or internal medicine consultations within three and twelve months (routine or symptom-driven), and additional diagnostic tests (e.g., ECG, Holter monitoring) performed during follow-up. Clinical data will be obtained from hospital records and complemented by baseline information on AF history, comorbidities, and procedure-related characteristics.

Patient-Reported Outcomes and Biosignals Health-related quality of life will be assessed using the Atrial Fibrillation Effect on Quality-of-Life (AFEQT) questionnaire, a validated, disease-specific instrument for AF. The AFEQT comprises 20 items across four domains (Symptoms, Daily Activities, Treatment Concerns, and Treatment Satisfaction) with an overall score (0-100) derived from the first three domains; higher scores indicate better quality of life. A change of ≥5 points is generally considered clinically meaningful. The AFEQT will be administered at baseline and follow-up to assess changes in AF-related symptoms and the broader impact of management strategies on daily life.

To complement the AFEQT, participants will complete brief, rotating weekly questions aligned with the ESC quality of treatment indicators (Arbelo et al., 2021), assessing symptom status (06.3SQI1), physical functioning (06.3SQI2), emotional well-being (06.3SQI3), cognitive function (06.3SQI4), and knowledge of AF-related risk factors. Each question will be administered once per week to minimize participant burden.

All patient-reported outcomes and biosignals will be collected via the chatbot. Body mass, blood pressure, and heart rate will be self-reported daily for the first four weeks and weekly thereafter, with blood pressure and heart rate measured using a digital sphygmomanometer. Physical activity will be tracked as daily step counts via the smartphone pedometer. Symptoms, including palpitations, dyspnea, chest pain, dizziness/syncope, fatigue, and edema, will also be reported daily during the first four weeks and weekly thereafter.

Chatbot System Usability and Perceived Usefulness System usability will be assessed using the System Usability Scale (SUS), a widely adopted 10-item, self-administered questionnaire that generates a global usability score ranging from 0 to 100. Scores above 68 are generally interpreted as reflecting good usability. The European Portuguese version has been translated, culturally adapted, and validated, demonstrating good construct validity through significant correlations with the PSSUQ (r=0.70) and with a general usability item (r=0.48). User satisfaction and perceived usefulness will be evaluated using the Post-Study System Usability Questionnaire (PSSUQ), a 19-item scale validated in European Portuguese, which has shown excellent internal consistency (α=0.80), satisfactory inter-rater reliability (ICC=0.67), and strong construct validity (r=0.84, P<.05). In this instrument, lower scores indicate greater satisfaction. Both SUS and PSSUQ will be administered at follow-up to provide a comprehensive evaluation of participants' (patients and health professionals) experiences with the PROMBot-FA intervention.

To complement quantitative assessments, semi-structured interviews will be conducted with patients to qualitatively explore their experiences with PROMBot-FA. Healthcare professionals directly involved in patient care (eg, cardiologists, nurses, and technicians) will also participate in semi-structured interviews to provide insights into usability, workflow integration, perceived utility, time demands, and potential improvements.

Inclusion Criteria:

• age 18 years or older;
• documented diagnosis of paroxysmal or persistent AF based on clinical assessment and coded according to the International Classification of Diseases, 11th Revision;
• undergone catheter ablation for AF at Hospital de Santa Marta during the study recruitment period;
• able to use the PROMBot-FA chatbot independently or with caregiver assistance, defined as Portuguese literacy sufficient to understand instructions and ability to operate a smartphone;
• clinically stable at discharge with no acute post-procedural complications precluding participation; and
• provision of written informed consent.

Exclusion Criteria:

• severe cognitive impairment preventing comprehension of study procedures, such as moderate-to-severe dementia or other neurological conditions limiting capacity to consent;
• severe uncorrected visual impairment or motor disability preventing smartphone operation even with caregiver support;
• explicit refusal to use digital technology or unwillingness to share health data required for the study;
• clinical or contextual conditions rendering 3-month follow-up unfeasible, including anticipated relocation outside the hospital catchment area or concurrent participation in another interventional study with overlapping endpoints; or
• life expectancy less than 6 months.

Data sharing is not planned at this stage. De-identified data may be shared in the future upon reasonable request after publication, pending ethics committee approval and compliance with data protection regulations (GDPR).

Due to the nature of the intervention, it is not feasible to blind participants or healthcare professionals providing care. However, outcome assessors conducting medical record reviews for clinical endpoints (eg, hospitalizations, cardiovascular events) will be blinded to group allocation where possible. Statistical analyses will be conducted by a researcher blinded to group assignment until completion of primary analyses. Participants will be instructed not to disclose their group allocation during follow-up assessments conducted by blinded assessors.