{"id":"W2803387226","doi":"10.1186/s13063-018-2654-z","title":"Outcome pre-specification requires sufficient detail to guard against outcome switching in clinical trials: a case study","year":2018,"lang":"en","type":"article","venue":"Trials","topic":"Meta-analysis and systematic reviews","field":"Decision Sciences","cited_by":16,"is_retracted":false,"has_abstract":true,"ca_institutions":"Western University; University Hospital","funders":"","keywords":"Medicine; Outcome (game theory); Confidence interval; Clinical trial; Odds ratio; Guard (computer science); Odds; Internal medicine; Logistic regression","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":"gemma","categories":["metaresearch","research_integrity"],"domain":"reporting","study_design":"case_report","genre":"empirical","about_ca_system":false,"about_ca_topic":false,"confidence":"low","status":"direct model label, unvalidated"},{"model":"gpt","categories":["metaresearch"],"domain":"reporting","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","scholarly_communication","insufficient_payload"],"consensus_categories":["metaresearch","insufficient_payload"],"category_scores_codex":[0.8457862,0.0005013412,0.01437462,0.001130738,0.0002655185,0.001569524,0.002204126,0.0001992701,0.003033901],"category_scores_gemma":[0.7327273,0.000236816,0.003945106,0.002468092,0.00006748258,0.0003020315,0.0003314092,0.0003801656,0.006301441],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001176823,"about_ca_system_score_gemma":0.0001191762,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0001414009,"about_ca_topic_score_gemma":0.001396476,"domain_scores_codex":[0.5713759,0.2660154,0.1480044,0.003869434,0.009803327,0.0009315178],"domain_scores_gemma":[0.8102284,0.1352558,0.03830762,0.01268875,0.002651151,0.000868289],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.0004346348,0.001239315,0.8024684,0.00002745229,0.000451923,0.0003285178,0.01020299,0.0001401608,0.0002427192,0.0003311756,0.01124761,0.1728851],"study_design_scores_gemma":[0.02310545,0.004686238,0.6736854,0.0003759377,0.005686044,0.0003344943,0.1135413,0.01884541,0.0001887274,0.004348011,0.1517813,0.003421591],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9669095,0.0001924589,0.02060969,0.0008737469,0.00220195,0.008291474,0.00001796539,0.00002109574,0.0008821021],"genre_scores_gemma":[0.9926426,0.000007930387,0.002960706,0.0005913076,0.001112354,0.0002880965,0.000002261832,0.00002333544,0.002371427],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.2388525,"threshold_uncertainty_score":0.999467,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.9745720513411243,"score_gpt":0.7350387175842116,"score_spread":0.2395333337569128,"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."}}