RISK OF ALGORITHMIC BIAS IN CONTEMPORARY ATRIAL FIBRILLATION RESEARCH

BackgroundIn recent years there is increasing attention to artificial intelligence (AI) and machine learning (ML) in the diagnosis treatment of atrial fibrillation (AF). However, AI approaches carry risk algorithmic bias from nonrepresentative training sets that may yield harm populations.ObjectiveEvaluate use descriptive data representing individual, social, structural variables research related AF.MethodsBibliometric analysis all papers yielded by a PubMed search using terms (atrial fibrillation) AND ((artificial intelligence) OR (machine learning)) (algorithm) (train*). Non-English were reviewed Google Translate. Full text was for papers.Results174 reviewed. Excluded 37 (21%) review articles, non-human studies, or did not describe AF diagnostics/treatment. Of remaining 137 papers, 78 (57%) described only rhythm status set; 48 (35%) additional individual level information (age/gender most common, comorbidities less common); 4 (3%) social determinants (e.g. tobacco, alcohol, home address); 2 (1.4%) (i.e. insurance status). Structural inferrable describing single-center studies countries with universal health care systems. Limitations included single reviewer, raising possibility misinterpretation some sets.ConclusionAt present, rarely incorporates patient metrics relevant determine representativeness broad application. In light historical examples demonstrating population generated samples, future should include explicit accounting determinant factors mitigate against avoid harm. populations. ObjectiveEvaluate AF. Evaluate MethodsBibliometric papers. Bibliometric Results174 sets. 174 ConclusionAt At