{"id":"W4407019191","doi":"10.1080/10903127.2025.2460775","title":"Accuracy of Commercial Large Language Model (ChatGPT) to Predict the Diagnosis for Prehospital Patients Suitable for Ambulance Transport Decisions: Diagnostic Accuracy Study","year":2025,"lang":"en","type":"article","venue":"Prehospital Emergency Care","topic":"Artificial Intelligence in Healthcare and Education","field":"Medicine","cited_by":5,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Alberta","funders":"","keywords":"Medicine; Medical emergency; Emergency medical services; Diagnostic accuracy; Emergency medicine; Triage; Intensive care medicine; Radiology","routes":{"ca_aff":true,"ca_fund":false,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":false},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":["metaresearch","metaepi_narrow"],"consensus_categories":[],"category_scores_codex":[0.0004252635,0.0003104081,0.0005189014,0.0001615398,0.0004329733,0.00001758625,0.0004662444,0.0001522454,0.0000710992],"category_scores_gemma":[0.01285463,0.0002485198,0.0003384999,0.0005161812,0.00004073159,0.0002617438,0.0000977739,0.0002156476,0.000007539933],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001622461,"about_ca_system_score_gemma":0.0004058848,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0007555161,"about_ca_topic_score_gemma":0.001105064,"domain_scores_codex":[0.9972079,0.00005351597,0.001112539,0.0005739041,0.0004333505,0.0006187981],"domain_scores_gemma":[0.9950107,0.00243775,0.0002400608,0.0007797384,0.001338411,0.0001932892],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.0004950064,0.002056137,0.8820401,0.0005610505,0.0001180984,9.481142e-7,0.0481636,0.0004693684,0.00001338628,0.0001796823,0.01048661,0.05541603],"study_design_scores_gemma":[0.001551463,0.005318771,0.919745,0.0009577283,0.0009169807,3.397646e-7,0.05985777,0.002507835,0.003660112,0.001368343,0.003529943,0.0005857345],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9705189,0.00331903,0.008615672,0.001194995,0.001884374,0.01293371,0.001343435,0.00006491596,0.0001249338],"genre_scores_gemma":[0.9894749,0.0005551878,0.0005942275,0.0003228352,0.0002560713,0.008056984,0.0004958154,0.00004757127,0.0001963938],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.05483029,"threshold_uncertainty_score":0.9999967,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.06145698107124997,"score_gpt":0.4222631523210531,"score_spread":0.3608061712498031,"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."}}