{"id":"W4220718380","doi":"10.2196/36388","title":"Evaluation and Mitigation of Racial Bias in Clinical Machine Learning Models: Scoping Review","year":2022,"lang":"en","type":"article","venue":"JMIR Medical Informatics","topic":"Artificial Intelligence in Healthcare and Education","field":"Medicine","cited_by":184,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"","keywords":"Scopus; Machine learning; Artificial intelligence; MEDLINE; Computer science; Systematic review; Publication bias; Meta-analysis; Selection bias; Medicine; Data science; Pathology","routes":{"ca_aff":false,"ca_fund":false,"ca_venue":true,"about_ca":false,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.00789671,0.00006719551,0.0003046022,0.0001141808,0.00008459113,0.000004859947,0.00005777366,0.00008260001,0.0004548557],"category_scores_gemma":[0.003249428,0.00006126336,0.00004666489,0.0002860903,0.0000796749,0.0001588347,0.00006177017,0.0006064842,0.000004797317],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00009425062,"about_ca_system_score_gemma":0.0009465854,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00006516064,"about_ca_topic_score_gemma":0.00005043349,"domain_scores_codex":[0.996699,0.0003661211,0.00155018,0.00007731829,0.001178087,0.0001292835],"domain_scores_gemma":[0.998852,0.0003361634,0.0003476046,0.000111668,0.0002112148,0.0001414096],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.00007017574,0.0002239622,0.01824749,0.007771036,0.0000148457,0.000002430774,0.01335526,0.0003963187,0.000002110621,0.0003254129,0.0002712364,0.9593197],"study_design_scores_gemma":[0.0003417532,0.0005349993,0.001182067,0.01755688,0.00008226824,0.00003918243,0.00441375,0.9740191,0.00004446064,0.001264007,0.0004163847,0.0001051369],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.983185,0.009730686,0.0006235517,0.002671502,0.0002974792,0.001972481,0.000002309035,0.00002591997,0.001491083],"genre_scores_gemma":[0.9885899,0.008074421,0.0005176233,0.002360797,0.0001076008,0.0002137345,0.000114026,0.00000700374,0.00001483589],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.9736228,"threshold_uncertainty_score":0.4980351,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.4425374768641928,"score_gpt":0.5513769212540904,"score_spread":0.1088394443898976,"validation_status":"score_only:v0-immature-baseline","note":"Baseline scores from an immature model (maturity gate not passed). Scores rank; they never assert a category."}}