{"id":"W4312974298","doi":"10.2196/38150","title":"An End-to-End Natural Language Processing Application for Prediction of Medical Case Coding Complexity: Algorithm Development and Validation","year":2022,"lang":"en","type":"article","venue":"JMIR Medical Informatics","topic":"Machine Learning in Healthcare","field":"Computer Science","cited_by":6,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"","keywords":"Computer science; Medical classification; Coding (social sciences); Metadata; Documentation; Machine learning; Artificial intelligence; Data mining; Medicine","routes":{"ca_aff":false,"ca_fund":false,"ca_venue":true,"about_ca":false,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.001860785,0.000105011,0.0001704305,0.0001583479,0.0004117367,0.00006291511,0.000508376,0.0000867792,0.00003808708],"category_scores_gemma":[0.0001760668,0.00009965869,0.00001910657,0.0003280616,0.00006227507,0.0003939052,0.0003728086,0.0004264977,0.000001112847],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001138345,"about_ca_system_score_gemma":0.0004268473,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00005095297,"about_ca_topic_score_gemma":0.00001589115,"domain_scores_codex":[0.9973242,0.0000922586,0.0006476543,0.0001584631,0.001564367,0.0002130926],"domain_scores_gemma":[0.9989725,0.0001395863,0.000250734,0.0002189099,0.0001038012,0.0003144991],"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":[0.000004961602,0.00004748019,0.000215136,0.0003942098,0.000005436334,0.00001481763,0.03354024,0.00004724747,0.000008241395,0.002031281,0.00006964395,0.9636213],"study_design_scores_gemma":[0.0003823031,0.0001284735,0.0004681483,0.00005437357,0.000003436484,0.0006425172,0.00339819,0.9917013,0.0001300773,0.00006639663,0.002931391,0.00009341089],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.1554232,0.00004296175,0.8424039,0.001155241,0.0001527058,0.0006286558,0.00002968958,0.0001344831,0.000029124],"genre_scores_gemma":[0.7710081,0.000001389013,0.2276952,0.0006566375,0.00007618702,0.0003126067,0.0002386895,0.00000714849,0.000004124388],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.991654,"threshold_uncertainty_score":0.4063963,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03208591677609966,"score_gpt":0.357616337529728,"score_spread":0.3255304207536284,"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."}}