{"id":"W4392104875","doi":"10.1097/sla.0000000000006250","title":"Distinguishing Clinical From Statistical Significances in Contemporary Comparative Effectiveness Research","year":2024,"lang":"en","type":"article","venue":"Annals of Surgery","topic":"Health Systems, Economic Evaluations, Quality of Life","field":"Economics, Econometrics and Finance","cited_by":22,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Toronto","funders":"National Cancer Institute","keywords":"Medicine; Clinical significance; Statistical significance; Observational study; Clinical trial; Sample size determination; Comparative effectiveness research; MEDLINE; Clinical study design; Internal medicine; Intensive care medicine; 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":"codex-gemma-dda1882f352a","candidate_categories":["metaresearch"],"consensus_categories":["metaresearch"],"category_scores_codex":[0.08768897,0.0001438927,0.001638274,0.0005622967,0.00007922302,0.0001170819,0.0001961581,0.000146123,0.0001966216],"category_scores_gemma":[0.01818566,0.0001701269,0.0001758038,0.0004483806,0.0003411592,0.0004539767,0.00005999214,0.0004842567,0.0006296689],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00009726036,"about_ca_system_score_gemma":0.0005065258,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.006050556,"about_ca_topic_score_gemma":0.0001126583,"domain_scores_codex":[0.9911803,0.002954442,0.004602305,0.0006869312,0.0001732272,0.0004028011],"domain_scores_gemma":[0.9174721,0.08126491,0.0006625316,0.0002988421,0.000147441,0.0001542341],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.0001268294,0.0002233248,0.7839133,0.000881433,0.0001600977,0.00003926853,0.001559328,0.000035312,0.000003879357,0.08756285,0.1246868,0.0008076702],"study_design_scores_gemma":[0.0002203061,0.00007537859,0.8573321,0.001558226,0.000002293981,6.170665e-7,0.00132994,0.004157698,0.00005353173,0.1062115,0.02875822,0.0003002865],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9647856,0.01692449,0.001403787,0.008110037,0.001331363,0.0004735437,0.001426251,0.00004447616,0.005500472],"genre_scores_gemma":[0.9980732,0.000323547,0.0003147732,0.0006965494,0.0003586855,0.00007238174,0.0001030956,0.00002071421,0.00003712007],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.09592856,"threshold_uncertainty_score":0.9900846,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.9704443140674623,"score_gpt":0.676357735808296,"score_spread":0.2940865782591664,"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."}}