{"id":"W4409476426","doi":"10.4300/jgme-d-24-00524.1","title":"Construct Validity Evidence for ACGME Milestones in Surgical Specialties: A Systematic Review","year":2025,"lang":"en","type":"review","venue":"Journal of Graduate Medical Education","topic":"Innovations in Medical Education","field":"Medicine","cited_by":2,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Toronto","funders":"","keywords":"Graduate medical education; Context (archaeology); Construct validity; Formative assessment; Accreditation; Inter-rater reliability; MEDLINE; Systematic review; Medical education; Evidence-based medicine; Medicine; Psychology; External validity; Evidence-based practice; Validity; Nursing; Rating scale; Psychometrics; Clinical psychology; Patient satisfaction; Alternative medicine; Pathology","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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.0575171,0.00221938,0.0106824,0.01914226,0.001397662,0.006076347,0.0028671,0.002791406,0.00406966],"category_scores_gemma":[0.2553104,0.001834043,0.01753897,0.01647131,0.003157919,0.004958404,0.003314878,0.002086441,0.0003370583],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.009223551,"about_ca_system_score_gemma":0.02671091,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.01358575,"about_ca_topic_score_gemma":0.03112688,"domain_scores_codex":[0.948553,0.01709452,0.01905699,0.003750284,0.01082267,0.0007225616],"domain_scores_gemma":[0.7147317,0.2232373,0.03842982,0.004430896,0.018107,0.00106329],"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":[0.0002034154,0.00002745208,0.004822026,0.9397834,0.0250472,0.00006236244,0.0004547547,0.0001329128,0.00007326441,0.0004525101,0.0006968993,0.02824386],"study_design_scores_gemma":[0.0002554594,0.000154301,0.01007265,0.8643413,0.1178549,0.0002130863,0.0005770355,0.0001837939,0.0001764031,0.0006303408,0.005481061,0.00005981158],"study_design_candidate":"systematic_review","study_design_consensus":"systematic_review","genre_codex":"review","genre_gemma":"review","genre_scores_codex":[0.003665303,0.9922461,0.0007616779,0.0005294631,0.0001614091,0.001337265,0.0007168084,0.00001758672,0.0005644292],"genre_scores_gemma":[0.1205221,0.8677449,0.004736576,0.001182756,0.0001928461,0.004511254,0.0009096091,0.00002398868,0.0001759988],"genre_candidate":"review","genre_consensus":"review","teacher_disagreement_score":0.0575171,"threshold_uncertainty_score":0.3041831,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.2265361997761885,"score_gpt":0.518192283470297,"score_spread":0.2916560836941085,"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."}}