{"id":"W4408688552","doi":"10.2196/65263","title":"Large Language Models for Pediatric Differential Diagnoses in Rural Health Care: Multicenter Retrospective Cohort Study Comparing GPT-3 With Pediatrician Performance","year":2025,"lang":"en","type":"article","venue":"JMIRx Med","topic":"Clinical Reasoning and Diagnostic Skills","field":"Medicine","cited_by":13,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"","keywords":"Medical diagnosis; Medicine; Retrospective cohort study; Pediatrics; Cohort; Health care; Cohort study; Family medicine; Surgery; Internal medicine; Radiology","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":[0.0003646343,0.0002359058,0.0008053111,0.0002504714,0.0001261355,0.00002912981,0.0001234445,0.00008597571,0.00001446982],"category_scores_gemma":[0.002487367,0.0001825,0.00009828447,0.0004395326,0.00002920915,0.00007568466,0.00007493232,0.0003532984,0.000004512817],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0003420318,"about_ca_system_score_gemma":0.0002364263,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0005148089,"about_ca_topic_score_gemma":0.001146876,"domain_scores_codex":[0.9980671,0.00007400774,0.0005035916,0.0004361419,0.0003791915,0.0005399378],"domain_scores_gemma":[0.9975245,0.001648791,0.0001495105,0.0003355029,0.000167198,0.0001745099],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.0007962785,0.001608065,0.9909644,0.0001671023,0.00006348833,0.00001871204,0.004236131,0.00004691817,7.388479e-7,0.00003344403,0.001332497,0.0007322136],"study_design_scores_gemma":[0.01247367,0.001067978,0.9764984,0.0004260505,0.0002858546,0.000001107772,0.00475846,0.004278577,0.00001061352,0.000006033663,0.00002356873,0.0001696904],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9937286,0.0006364286,0.00060741,0.0006705563,0.0003386197,0.003504581,0.00003397428,0.00009219031,0.0003875921],"genre_scores_gemma":[0.9976985,0.0001891193,0.0002043004,0.0003452139,0.0003691732,0.0006305287,0.0001260961,0.00002838946,0.0004087019],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.01446602,"threshold_uncertainty_score":0.7442135,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01130311277027957,"score_gpt":0.3343810915467513,"score_spread":0.3230779787764717,"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."}}