{"id":"W4410659767","doi":"10.2196/68409","title":"Clinical Performance and Communication Skills of ChatGPT Versus Physicians in Emergency Medicine: Simulated Patient Study","year":2025,"lang":"en","type":"article","venue":"JMIR Medical Informatics","topic":"Artificial Intelligence in Healthcare and Education","field":"Medicine","cited_by":15,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"","keywords":"Preprint; Medicine; MEDLINE; Medical emergency; Medical physics; Computer science; World Wide Web","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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.004068143,0.0004652987,0.000603579,0.0006571051,0.0005844566,0.0009078693,0.0004317506,0.0009413263,0.003763518],"category_scores_gemma":[0.0251137,0.0002538239,0.0004217894,0.0003226573,0.0009918441,0.0009311403,0.001611627,0.0009001245,0.0006443632],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0006724653,"about_ca_system_score_gemma":0.0006608155,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0007882873,"about_ca_topic_score_gemma":0.0007715509,"domain_scores_codex":[0.9979226,0.001233661,0.0001610288,0.0002420664,0.0002316275,0.0002089499],"domain_scores_gemma":[0.9789557,0.01049559,0.002333156,0.00136797,0.001602721,0.005244766],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.03060225,0.05657346,0.7390804,0.0006778312,0.000466135,0.002441402,0.0706339,0.006650806,0.01087293,0.0005112865,0.001986098,0.07950345],"study_design_scores_gemma":[0.002536333,0.1753558,0.769899,0.0001208638,0.0002690374,0.002004611,0.02464148,0.01464169,0.007163118,0.0006169043,0.002495279,0.0002558336],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9997205,0.000008195075,0.00006860226,0.00001098116,0.000002705283,0.00003857365,0.00001397553,0.000002512213,0.0001338544],"genre_scores_gemma":[0.9991885,0.00002413407,0.0003528737,0.00003499542,0.00001156047,0.0001097431,0.00005464135,0.000001784754,0.0002218242],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.004068143,"threshold_uncertainty_score":0.02151459,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.08406932550693881,"score_gpt":0.4844262238204639,"score_spread":0.4003568983135251,"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."}}