{"id":"W4327862079","doi":"10.1101/2023.03.14.23287202","title":"Characterizing the limitations of using diagnosis codes in the context of machine learning for healthcare","year":2023,"lang":"en","type":"preprint","venue":"medRxiv","topic":"Machine Learning in Healthcare","field":"Computer Science","cited_by":1,"is_retracted":false,"has_abstract":true,"ca_institutions":"Institute for Clinical Evaluative Sciences; SickKids Foundation; Hospital for Sick Children","funders":"","keywords":"Medicine; Context (archaeology); Confidence interval; Medical diagnosis; Pediatrics; Odds ratio; Gold standard (test); Concordance; Kappa; 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":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.003006462,0.0002242472,0.0005039054,0.0002525213,0.0002542162,0.00007506109,0.001926272,0.0001550403,0.000001300786],"category_scores_gemma":[0.002899648,0.0001519101,0.0001927021,0.0005521203,0.00009892345,0.00008258721,0.0007815716,0.001215457,0.000001524145],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00006082363,"about_ca_system_score_gemma":0.0002367616,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.004669549,"about_ca_topic_score_gemma":0.001938445,"domain_scores_codex":[0.9966905,0.001280353,0.0008035314,0.0004705199,0.0004221046,0.000333022],"domain_scores_gemma":[0.9915106,0.006325114,0.0008912557,0.0009398356,0.0002908932,0.0000422616],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.00001577678,0.00006280219,0.9066335,0.001695151,0.00004912582,0.00000614111,0.04148192,0.009215681,0.0002953729,0.01080705,0.00001901626,0.02971849],"study_design_scores_gemma":[0.000309107,0.0001844875,0.3485146,0.001441511,0.00003726019,0.000007284077,0.002446233,0.6409485,0.0005817868,0.002802911,0.002414877,0.0003114524],"study_design_candidate":"observational","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9226421,0.001354769,0.02258573,0.05055276,0.0009098514,0.001720697,0.00009743486,0.0001171934,0.00001943217],"genre_scores_gemma":[0.9928772,0.0007130307,0.005336876,0.0004823006,0.00008622198,0.0004139146,0.00004346111,0.00003279979,0.00001416884],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.6317328,"threshold_uncertainty_score":0.7058992,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.2051099917473286,"score_gpt":0.3675477546853069,"score_spread":0.1624377629379783,"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."}}