Mapping Algorithmic Bias in AI-Powered Electrocardiogram Interpretation Across the AI Life Cycle: Protocol for a Scoping Review
Bibliographic record
Abstract
BACKGROUND: Artificial intelligence (AI)-powered analysis of electrocardiograms (ECGs) is reshaping cardiac diagnostics, offering faster and often more accurate detection of conditions such as arrhythmias and heart failure. However, growing evidence suggests that algorithmic bias, defined as performance disparities across patient subgroups, may undermine diagnostic equity. These biases can emerge at any stage of the AI life cycle, including data collection, model development, evaluation, deployment, and clinical use. If unaddressed, they risk exacerbating health disparities, particularly in underrepresented populations and low-resource settings. Early identification and mitigation of such bias are essential to ensuring diagnostic equity. OBJECTIVE: This scoping review protocol outlines a structured approach to mapping the evidence on algorithmic bias in AI-enabled ECG interpretation. Following the population-concept-context framework and PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses Extension for Scoping Reviews) guidance, the planned review will systematically identify and categorize reported sources and types of bias, examine their effects on diagnostic performance across demographic and geographic subgroups, and document mitigation strategies applied throughout the AI life cycle. By synthesizing how bias and fairness considerations are handled in this field, this review aims to clarify existing evidence, highlight key gaps, and inform future efforts toward equitable and clinically trustworthy application of AI in cardiology. METHODS: We will conduct a comprehensive literature search across 5 electronic databases (PubMed, Embase, Cochrane CENTRAL, CINAHL, and IEEE Xplore) and gray literature sources. Eligible studies will include original research (2015-2025) evaluating the performance of AI-based ECG models across different subgroups or reporting on bias mitigation strategies. Two reviewers will independently screen studies, extract data using a standardized form, and resolve disagreements through consensus. This review will follow the PRISMA-ScR reporting framework. RESULTS: At the time of submission, study identification and screening has been completed. Database searches conducted in August and September 2025 yielded 430 records, with an additional 18 records identified through other sources. After duplicates removal, 398 unique records remained. Title and abstract screening led to the exclusion of 250 records, and 148 articles proceeded to full-text review. Following full-text assessment, 110 articles were evaluated for eligibility, of which 38 studies met the inclusion criteria and were included in the qualitative synthesis. The study selection process is summarized in a PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) flow diagram. Data extraction was conducted between November and December 2025. CONCLUSIONS: This review will be the first to comprehensively map the landscape of algorithmic bias in AI-powered ECG interpretation. By identifying patterns of inequity and evaluating proposed solutions, it will provide actionable insights for developers, clinicians, and policymakers aiming to promote fairness in AI-enabled cardiac care. INTERNATIONAL REGISTERED REPORT IDENTIFIER (IRRID): PRR1-10.2196/82486.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.124 | 0.195 |
| Meta-epidemiology (narrow) | 0.006 | 0.005 |
| Meta-epidemiology (broad) | 0.013 | 0.019 |
| Bibliometrics | 0.020 | 0.017 |
| Science and technology studies | 0.004 | 0.006 |
| Scholarly communication | 0.010 | 0.009 |
| Open science | 0.006 | 0.009 |
| Research integrity | 0.010 | 0.007 |
| Insufficient payload (model declined to judge) | 0.066 | 0.011 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".