{"id":"W2081069584","doi":"10.1097/01.sla.0000154258.30305.df","title":"Evaluating Meta-analyses in the General Surgical Literature","year":2005,"lang":"en","type":"article","venue":"Annals of Surgery","topic":"Meta-analysis and systematic reviews","field":"Decision Sciences","cited_by":131,"is_retracted":false,"has_abstract":true,"ca_institutions":"McMaster University; University of Calgary","funders":"","keywords":"Medicine; Meta-analysis; MEDLINE; Concordance; Critical appraisal; Systematic review; Evidence-based medicine; Family medicine; Alternative medicine; Internal 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":[{"model":"gpt","categories":["metaresearch","metaepi_broad"],"domain":"methods","study_design":"systematic_review","genre":"empirical","about_ca_system":false,"about_ca_topic":false,"confidence":"high","status":"direct model label, unvalidated"},{"model":"grok","categories":["metaresearch","metaepi_broad"],"domain":"methods","study_design":"systematic_review","genre":"empirical","about_ca_system":false,"about_ca_topic":false,"confidence":"high","status":"direct model label, unvalidated"},{"model":"opus","categories":["metaresearch","metaepi_broad"],"domain":"methods","study_design":"observational","genre":"empirical","about_ca_system":false,"about_ca_topic":false,"confidence":"high","status":"direct model label, unvalidated"}],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":["metaresearch","insufficient_payload"],"consensus_categories":["metaresearch"],"category_scores_codex":[0.2731131,0.0002597481,0.004710898,0.0006318925,0.00007430485,0.000606087,0.00126476,0.00008586672,0.009595282],"category_scores_gemma":[0.03606727,0.00008852242,0.006882219,0.002865985,0.00005551543,0.0003757047,0.00006852684,0.0001925113,0.0004994694],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00000392645,"about_ca_system_score_gemma":0.00005679932,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00001708199,"about_ca_topic_score_gemma":0.00001659714,"domain_scores_codex":[0.9572694,0.02380348,0.0105428,0.0007513728,0.007278417,0.0003544836],"domain_scores_gemma":[0.9675193,0.02418369,0.003779335,0.002940191,0.001489401,0.00008811528],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"not_applicable","study_design_gemma":"not_applicable","study_design_scores_codex":[0.00003859975,0.0004027984,0.02967573,0.00008537232,0.005848264,0.0001345957,0.002938794,0.006120134,0.0003831092,0.004073814,0.8113729,0.1389259],"study_design_scores_gemma":[0.0002695895,0.00006099486,0.05701112,0.0001320574,0.005283858,0.0001597964,0.00108577,0.02720219,0.00138117,0.01486777,0.8917685,0.0007772485],"study_design_candidate":"not_applicable","study_design_consensus":"not_applicable","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9401754,0.03870529,0.00007147048,0.01403335,0.0001072789,0.000283581,0.00002297956,0.000003053748,0.006597555],"genre_scores_gemma":[0.9916838,0.0002390217,0.001405058,0.002077,0.0002809549,0.00003586739,0.00001115346,0.000007374644,0.004259737],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.2370458,"threshold_uncertainty_score":0.9913101,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.9842320020590651,"score_gpt":0.6970802859876226,"score_spread":0.2871517160714425,"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."}}