{"id":"W4200191097","doi":"10.1038/s41591-021-01595-0","title":"Underdiagnosis bias of artificial intelligence algorithms applied to chest radiographs in under-served patient populations","year":2021,"lang":"en","type":"article","venue":"Nature Medicine","topic":"Artificial Intelligence in Healthcare and Education","field":"Medicine","cited_by":784,"is_retracted":false,"has_abstract":true,"ca_institutions":"Vector Institute; University of Toronto","funders":"Natural Sciences and Engineering Research Council of Canada; Canadian Institute for Advanced Research; Microsoft Research","keywords":"Context (archaeology); Socioeconomic status; Medicine; Artificial intelligence; Exacerbation; Computer science; Machine learning; Population; Internal medicine","routes":{"ca_aff":true,"ca_fund":true,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":false},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.03126141,0.000561872,0.0007450652,0.00298229,0.0009102057,0.00325634,0.001178638,0.001255213,0.001022876],"category_scores_gemma":[0.1751597,0.0003045488,0.0009750898,0.001953661,0.002210829,0.002025601,0.002313419,0.001671168,0.0005659277],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001333141,"about_ca_system_score_gemma":0.001323894,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.004299473,"about_ca_topic_score_gemma":0.00682885,"domain_scores_codex":[0.9687195,0.02012535,0.002897919,0.003699358,0.00374425,0.0008137295],"domain_scores_gemma":[0.8397682,0.1133355,0.01993912,0.013523,0.01239716,0.001036919],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.0004787857,0.0001123982,0.9044192,0.000208348,0.0005185335,0.0002396462,0.001735427,0.006236675,0.001282611,0.001845103,0.003443937,0.07947928],"study_design_scores_gemma":[0.0001639024,0.000741653,0.8021462,0.001068258,0.001020473,0.004259077,0.006912307,0.1141449,0.01747597,0.0281916,0.02373822,0.0001375816],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9613097,0.00377085,0.02462287,0.004690946,0.000257133,0.0001407698,0.0009574344,0.0002269206,0.004023236],"genre_scores_gemma":[0.9894064,0.0004827827,0.007472285,0.001271223,0.0001276174,0.00003639246,0.000883568,0.00004748373,0.0002723727],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.03126141,"threshold_uncertainty_score":0.1653281,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.2655884020617279,"score_gpt":0.4424643733409022,"score_spread":0.1768759712791743,"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."}}