{"id":"W4386304371","doi":"10.1186/s12909-023-04600-x","title":"Medical specialty certification exams studied according to the Ottawa Quality Criteria: a systematic review","year":2023,"lang":"en","type":"review","venue":"BMC Medical Education","topic":"Innovations in Medical Education","field":"Medicine","cited_by":6,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"","keywords":"Specialty; Certification; Summative assessment; Medical education; Medicine; Family medicine; Quality (philosophy); Inclusion (mineral); Psychology; Political science; Formative assessment; Pedagogy","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":["metaresearch","metaepi_narrow","insufficient_payload"],"consensus_categories":["insufficient_payload"],"category_scores_codex":[0.02121145,0.0007767554,0.004497761,0.0006585585,0.0004038491,0.0001044328,0.001564643,0.001099962,0.002192828],"category_scores_gemma":[0.2843252,0.0004827406,0.0007306383,0.004565734,0.0002662435,0.0001229477,0.0003199634,0.001931764,0.003601878],"about_ca_system_candidate":true,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001433349,"about_ca_system_score_gemma":0.02316135,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0001272574,"about_ca_topic_score_gemma":0.0002201458,"domain_scores_codex":[0.9844217,0.002835064,0.00538918,0.001234253,0.0054079,0.0007118651],"domain_scores_gemma":[0.9865287,0.004210985,0.002623158,0.003831851,0.001503523,0.001301767],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"systematic_review","study_design_gemma":"systematic_review","study_design_scores_codex":[0.000003940264,0.0003608358,0.000006718398,0.5594988,0.0001435338,0.000002995886,0.0001311808,1.013646e-8,2.368984e-8,0.0006610051,0.1665489,0.2726421],"study_design_scores_gemma":[0.0001490576,0.00003812772,0.00005146756,0.5682557,0.002585805,0.0001580224,0.0006553482,0.00002566785,9.134518e-8,0.00003115034,0.4277305,0.0003190862],"study_design_candidate":"systematic_review","study_design_consensus":"systematic_review","genre_codex":"review","genre_gemma":"review","genre_scores_codex":[0.00001465005,0.9501714,0.001548917,0.0279692,0.007769591,0.01132482,0.00001389638,0.0003186301,0.0008689353],"genre_scores_gemma":[0.0000131313,0.9592482,0.0007771446,0.0199268,0.00611761,0.009989866,0.001560935,0.0001623479,0.002203909],"genre_candidate":"review","genre_consensus":"review","teacher_disagreement_score":0.272323,"threshold_uncertainty_score":0.9997624,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.2085994000249314,"score_gpt":0.528740869479074,"score_spread":0.3201414694541426,"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."}}