{"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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.006400666,0.0005463561,0.0003624302,0.001065955,0.0003869743,0.001003693,0.0008801269,0.0004714253,0.001512169],"category_scores_gemma":[0.02767293,0.0003564225,0.0007626128,0.0008943647,0.0004571347,0.001084353,0.001047751,0.0007769961,0.0004142419],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001155136,"about_ca_system_score_gemma":0.0009403891,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.009311935,"about_ca_topic_score_gemma":0.008803071,"domain_scores_codex":[0.9974974,0.00113365,0.0001765807,0.0006277765,0.0003883269,0.0001762363],"domain_scores_gemma":[0.9849612,0.006472382,0.004804835,0.001577658,0.001432394,0.0007516209],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.0001062361,0.00004264034,0.9979386,0.000006412087,0.00003821928,0.00004395544,0.0001388678,0.0003265916,0.00004399293,0.00001728334,0.0001679715,0.001129246],"study_design_scores_gemma":[0.00005315629,0.0006697582,0.9820805,0.00003566744,0.00009776285,0.0006803025,0.000980645,0.014561,0.0002118956,0.0001331239,0.0004779522,0.00001834651],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9990545,0.00005122182,0.0003654317,0.00004061822,0.000003057701,0.00002623805,0.0002989694,0.00001698665,0.0001429343],"genre_scores_gemma":[0.9990097,0.00002789818,0.0004511634,0.0000198032,0.0000041008,0.00002474569,0.0004131108,0.000008968879,0.00004051172],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.009311935,"threshold_uncertainty_score":0.03385031,"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."}}