{"id":"W4417223250","doi":"10.48550/arxiv.2504.00052","title":"Assessing Validity of ICD-10 Administrative Data in Coding Comorbidities","year":2025,"lang":"en","type":"preprint","venue":"ArXiv.org","topic":"Medical Coding and Health Information","field":"Health Professions","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"","keywords":"Comorbidity; Coding (social sciences); Chart; Health informatics; Diagnosis code; Informatics; Diabetes mellitus","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":["metaepi_narrow","research_integrity","insufficient_payload"],"consensus_categories":[],"category_scores_codex":[0.004233558,0.0002586444,0.0007704159,0.0002600125,0.000476533,0.00002320172,0.00083023,0.0007254625,0.001071765],"category_scores_gemma":[0.003680913,0.0002465502,0.0000594814,0.0002222409,0.0001334854,0.0003990694,0.001829511,0.002929481,0.0001359098],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0003076506,"about_ca_system_score_gemma":0.003838742,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.001006494,"about_ca_topic_score_gemma":0.0003972279,"domain_scores_codex":[0.9957293,0.001108531,0.001755543,0.0004599354,0.0004318339,0.00051482],"domain_scores_gemma":[0.9947035,0.002459524,0.001169405,0.00122638,0.0002803652,0.0001607983],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.00005764131,0.00009380901,0.9594904,0.01261566,0.00003667274,0.000007938918,0.006048691,0.00002218449,0.00001259872,0.001294633,0.01922244,0.001097285],"study_design_scores_gemma":[0.001275674,0.00008333513,0.9338378,0.01871875,0.000101622,7.985283e-7,0.01217643,0.006709105,0.0001666904,0.001246893,0.02516523,0.0005177283],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9289884,0.0001436603,0.002010079,0.00253281,0.00279078,0.00103955,0.0006958958,0.0001213373,0.06167747],"genre_scores_gemma":[0.9931648,0.0002540351,0.000736044,0.0009310854,0.0003923454,0.0001017346,0.001563645,0.00001333328,0.002842928],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.06417643,"threshold_uncertainty_score":0.9999987,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.7906320805313973,"score_gpt":0.5919478072159695,"score_spread":0.1986842733154278,"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."}}