{"id":"W6977540911","doi":"10.6084/m9.figshare.12850223","title":"The diagnostic value of Red Flags in thoracolumbar pain: a systematic review","year":2020,"lang":"en","type":"article","venue":"Figshare","topic":"Health, Technology, Consumer Behavior","field":"Health Professions","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"","keywords":"Observational study; Systematic review; Cohen's kappa; Low back pain; MEDLINE; Back pain; Gold standard (test); Chest pain","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","insufficient_payload"],"consensus_categories":["insufficient_payload"],"category_scores_codex":[0.001592756,0.0001677661,0.0007514993,0.0000497696,0.0002982759,0.000005775929,0.0006590377,0.0003158707,0.01175339],"category_scores_gemma":[0.1262677,0.0001241991,0.00009245811,0.0006650695,0.0000179101,0.00005475791,0.0002440614,0.001069682,0.002807892],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001334562,"about_ca_system_score_gemma":0.0005508734,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00005599138,"about_ca_topic_score_gemma":0.0002068233,"domain_scores_codex":[0.9946091,0.002986926,0.001313664,0.0002948424,0.0002696655,0.0005258645],"domain_scores_gemma":[0.9906527,0.007561618,0.0006058638,0.0007770668,0.0002282232,0.0001745266],"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":[0.000004531631,0.00002557742,0.001630706,0.659133,0.00001284378,0.00003226631,0.0004894907,3.282662e-7,0.00001113583,0.00004238422,0.3379029,0.0007148862],"study_design_scores_gemma":[0.000332603,0.00008805945,0.00982497,0.9655266,0.00009476837,0.000002018885,0.0007784959,0.0001940617,0.00001119465,0.00003977202,0.02291104,0.0001964272],"study_design_candidate":"systematic_review","study_design_consensus":"systematic_review","genre_codex":"review","genre_gemma":"empirical","genre_scores_codex":[0.0109633,0.4811194,0.00001584353,0.2205894,0.0009003415,0.1024196,0.1783594,0.002629667,0.003003039],"genre_scores_gemma":[0.8451576,0.008048781,0.0002796337,0.05814939,0.0003568914,0.06237604,0.02461775,0.0003336818,0.000680274],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.8341942,"threshold_uncertainty_score":0.9979686,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1353443287988047,"score_gpt":0.4329198461723284,"score_spread":0.2975755173735236,"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."}}