{"id":"W3093336514","doi":"10.1186/s12874-020-01129-1","title":"An evaluation of DistillerSR’s machine learning-based prioritization tool for title/abstract screening – impact on reviewer-relevant outcomes","year":2020,"lang":"en","type":"article","venue":"BMC Medical Research Methodology","topic":"Meta-analysis and systematic reviews","field":"Decision Sciences","cited_by":141,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Ottawa; McGill University; Ottawa Hospital","funders":"Canadian Institutes of Health Research","keywords":"Interquartile range; Prioritization; Medicine; Recall; Computer science; Reduction (mathematics); Machine learning; Surgery; Psychology; Mathematics; Management science","routes":{"ca_aff":true,"ca_fund":true,"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","insufficient_payload"],"consensus_categories":["metaresearch"],"category_scores_codex":[0.4938144,0.0001598213,0.002023384,0.0003075932,0.0001000409,0.00008803529,0.000990612,0.000155146,0.03525557],"category_scores_gemma":[0.8683048,0.00007418382,0.0008762632,0.0007921253,0.0001319986,0.00006965977,0.00005971931,0.0004093923,0.0005637506],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00004284471,"about_ca_system_score_gemma":0.0007622634,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00002217283,"about_ca_topic_score_gemma":0.00002105339,"domain_scores_codex":[0.8668057,0.1121272,0.004652982,0.0009470938,0.01505154,0.0004154317],"domain_scores_gemma":[0.8122507,0.1769876,0.002112878,0.00184705,0.006143304,0.000658533],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0002879232,0.0001367853,0.03706897,0.0005452848,0.0001769956,0.000003275101,0.0001951764,0.006944257,0.0003172146,0.001742299,0.03109722,0.9214846],"study_design_scores_gemma":[0.0006530157,0.000559222,0.0452361,0.0001111602,0.0001221086,7.914931e-7,0.00008965871,0.9182074,0.00006352704,0.002263954,0.03257977,0.0001132703],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.01847984,0.0009505139,0.975691,0.002383641,0.0001627631,0.001382995,0.00002940621,0.000008203056,0.0009116292],"genre_scores_gemma":[0.8540188,0.0000877499,0.1441618,0.0005056994,0.0003054648,0.000140573,0.0001193475,0.00002770226,0.0006328098],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.9213713,"threshold_uncertainty_score":0.9656264,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.9578986850993956,"score_gpt":0.7245179146353774,"score_spread":0.2333807704640182,"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."}}