{"id":"W4416419371","doi":"10.1186/s12245-025-01069-x","title":"Safety and accuracy of AI in triaging patients in the emergency department","year":2025,"lang":"en","type":"article","venue":"International Journal of Emergency Medicine","topic":"Artificial Intelligence in Healthcare and Education","field":"Medicine","cited_by":5,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"","keywords":"Triage; Emergency department; Reliability (semiconductor); MEDLINE; Sample (material); Patient safety","routes":{"ca_aff":false,"ca_fund":false,"ca_venue":false,"about_ca":true,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.001106287,0.00008417565,0.0002527685,0.0005094099,0.00002140838,0.000001572845,0.000196671,0.00004184138,0.0004125346],"category_scores_gemma":[0.002400018,0.00005443415,0.00005734666,0.0003897281,0.00003617372,0.0001206282,0.00002169823,0.0002781094,0.000001010464],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001082481,"about_ca_system_score_gemma":0.0001590981,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0006807106,"about_ca_topic_score_gemma":0.0003176631,"domain_scores_codex":[0.997631,0.00009152369,0.0015971,0.00009708813,0.0004687833,0.000114518],"domain_scores_gemma":[0.9985004,0.0002111174,0.0003874503,0.0001002533,0.0007559361,0.00004483676],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.0007664505,0.0003956165,0.9439881,0.0000599594,0.00006721363,0.00001219881,0.003656887,0.00002038716,0.0002250862,0.001012858,0.005534759,0.04426049],"study_design_scores_gemma":[0.0008540948,0.0004329231,0.9788356,0.0009008842,0.00005994764,0.00001174158,0.00289456,0.0001771375,0.0003912998,0.005805365,0.009580293,0.00005613939],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9558722,0.00155066,0.0003383495,0.03515819,0.005344685,0.0002455716,0.000002066769,0.000001631958,0.001486621],"genre_scores_gemma":[0.9915373,0.007444674,0.00003234723,0.0004911501,0.0003657708,0.000004526049,0.00001071313,0.000003777523,0.0001097806],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.04420435,"threshold_uncertainty_score":0.4516965,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1087583195405758,"score_gpt":0.4874386001978604,"score_spread":0.3786802806572845,"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."}}