Cardiology · Medical Literature
Conversational artificial intelligence HeAlth supporT in Atrial Fibrillation Self-Management (CHAT-AF-S): rationale and randomised controlled trial design.
Abstract
Atrial fibrillation (AF), a common arrhythmia, is associated with impaired quality of life and increased stroke risk and mortality. Clinical guidelines recommend leveraging digital technologies to support patient education and AF self-management. Conversational artificial intelligence (AI) technologies may support patient engagement with self-management by enabling human-like conversations. This study aims to evaluate the effectiveness of a conversational AI intervention (Conversational artificial intelligence HeAlth supporT in Atrial Fibrillation Self-Management (CHAT-AF-S)) in improving quality of life in patients with AF. CHAT-AF-S is a 3-month randomised controlled trial with 1:1 allocation and embedded process evaluation. We will randomise 480 adults (aged 18 years and older) with documented AF to the CHAT-AF-S intervention or usual care. Primary outcome is the Atrial Fibrillation Effect on QualiTy-of-life overall score. We will follow the intention-to-treat principles and data analysts will be blinded. Intervention participants will be invited to complete a user experience survey and take part in an interview to explore the feasibility, acceptability, perceived use and barriers and enablers to implementing the intervention. Qualitative data will be analysed thematically. Ethics approval was obtained from the Western Sydney Local Health District Human Ethics Research Committee (2023/ETH00765). Written and informed consent will be obtained from all study participants before commencing any study procedures. Results will be disseminated via peer-reviewed publications and presentations at international conferences. Australian New Zealand Clinical Trials Registry (registration number: ACTRN12623000850673).
Clinical relevance
In adults with documented atrial fibrillation, does a 3-month conversational AI intervention (CHAT-AF-S) — comprising voice-based automated phone calls and text messages — improve AF-related quality of life compared with usual care? __SNAPSHOT__ {"studyDesign":"Randomized controlled trial (RCT) — published protocol/design paper with embedded process evaluation (results not yet reported); multicenter, single-blind, 1:1 parallel allocation","setting":"Australia; multicenter (specific sites not reported in abstract); Western Sydney Local Health District ethics oversight","sampleSize":"Target n=480 (intervention: 240, usual care: 240)","population":"Adults aged ≥18 years with documented atrial fibrillation; additional inclusion/exclusion criteria not reported in the abstract","intervention":"CHAT-AF-S: a 3-month self-management program comprising voice-based conversational AI (automated phone calls using speech recognition and natural language processing) and text messages to support AF self-management","comparator":"Usual care","primaryOutcome":"Atrial Fibrillation Effect on QualiTy-of-life (AFEQT) overall score at 3 months","followUp":"3 months","fundingSource":"Not reported in abstract; investigators declare no conflicts of interest"}
Key findings
This is a protocol/design publication; no efficacy or outcome data are reported. The study is ongoing and results are not yet available. The sample size of 480 was calculated to provide adequate power to detect a between-group difference in AFEQT score, addressing the key limitation of the underpowered predecessor trial (CHAT-AF, n=103). Analysis will follow intention-to-treat principles with blinded data analysts. An embedded process evaluation will assess feasibility, acceptability, perceived utility, and barriers and enablers to implementation via user experience surveys and qualitative interviews with intervention participants, with thematic analysis of qualitative data. No p-values, effect sizes, confidence intervals, or NNT/NNH are reportable at this stage.
Practice implications
This publication establishes the rationale and methodological framework for CHAT-AF-S but does not establish clinical effectiveness. The predecessor proof-of-concept RCT (CHAT-AF, n=103) showed a significant within-group improvement in AFEQT score in the intervention arm (7.1 points; 95% CI 2.9–11.3) but was underpowered to detect between-group differences and did not demonstrate superiority over usual care. CHAT-AF-S is designed to remedy those limitations through a larger, multicenter, adequately powered design. Until trial results are reported, no conclusions about the effectiveness of conversational AI for AF self-management can be drawn. Patients not well represented in the predecessor trial included non-English speakers and those without mobile phone access; it is unclear whether CHAT-AF-S addresses these equity gaps. This summary is based on the published protocol and abstract only, which is an important limitation. __LIMITATIONS__ Protocol-only publication: no outcome data available; clinical effectiveness is entirely unestablished at this stage.,Single-blind design (participants cannot be blinded to intervention assignment), introducing potential performance and detection bias for patient-reported outcomes.,The AFEQT is a patient-reported outcome measure; it is a validated instrument but susceptible to expectation and response bias in an unblinded participant population.,Generalizability is uncertain: the predecessor trial recruited exclusively English-speaking patients post-discharge from a single Australian tertiary center; whether CHAT-AF-S broadens this to diverse populations and settings is not reported in the abstract.,Three-month follow-up may be insufficient to detect durable quality-of-life benefit or to evaluate clinically important hard endpoints (stroke, hospitalization, mortality).,The intervention relies on telephone and text-message access, potentially excluding elderly, digitally disadvantaged, or hearing-impaired AF patients — groups with high disease burden.,No pre-specified secondary outcomes beyond the process evaluation are described in the abstract; it is unclear whether cardiovascular events or health care utilization are captured.,Funding source is not disclosed in the abstract, limiting assessment of potential sponsorship bias.,The minimum clinically important difference (MCID) for AFEQT used in the power calculation is not stated in the abstract, making it impossible to assess whether the trial is powered for a clinically meaningful effect. __BOTTOM_LINE_CATEGORY__ not_ready_for_clinical_application __BOTTOM_LINE__ CHAT-AF-S is a well-framed, adequately powered multicenter RCT designed to address the limitations of its underpowered predecessor (CHAT-AF) and rigorously evaluate whether a 3-month conversational AI program improves quality of life in AF patients. At present, this is a protocol publication only and no practice-informing efficacy data exist. Clinicians should await the trial results before considering conversational AI tools as adjuncts to standard AF self-management support.