{"id":"W4414489233","doi":"10.21203/rs.3.rs-7578570/v1","title":"Computer Assisted Verbal Autopsy: Comparing Large Language Models to Physicians for Assigning Causes to 6939 Deaths in Sierra Leone from 2019-2022","year":2025,"lang":"en","type":"preprint","venue":"Research Square","topic":"Machine Learning in Healthcare","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":false,"ca_institutions":"St. Michael's Hospital","funders":"Bill and Melinda Gates Foundation","keywords":"Sierra leone; Concordance; Verbal autopsy; Epidemiology; Population; Public health; Health care","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":["metaepi_narrow","open_science","research_integrity"],"consensus_categories":[],"category_scores_codex":[0.00279151,0.0005498232,0.001081711,0.001492543,0.0005099709,0.0008110797,0.003542156,0.0004439552,0.00002250504],"category_scores_gemma":[0.0004787283,0.0006184055,0.0002528572,0.00176568,0.00004060382,0.0002995738,0.008499214,0.003178901,0.00006235071],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001263468,"about_ca_system_score_gemma":0.001043616,"about_ca_topic_candidate":true,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.01323318,"about_ca_topic_score_gemma":0.005803507,"domain_scores_codex":[0.9919306,0.001537932,0.0007738129,0.002185589,0.001714836,0.001857229],"domain_scores_gemma":[0.9944568,0.001909604,0.0001733478,0.002258478,0.0007254397,0.0004763258],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0003053203,0.0009254988,0.02248042,0.003641332,0.0002858905,0.0004263785,0.08283864,0.7225109,0.0006885336,0.01457076,0.030962,0.1203643],"study_design_scores_gemma":[0.0007648467,0.0002560512,0.03406198,0.003693338,0.000009867862,0.000001405677,0.0006345078,0.9557654,0.0001422848,0.001571009,0.002453277,0.0006459806],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.1092084,0.0006958359,0.8773081,0.005375248,0.001171918,0.003835008,0.0009382063,0.0005682596,0.0008990319],"genre_scores_gemma":[0.8621234,0.00001747355,0.1332149,0.001098477,0.0007565076,0.001242923,0.0005471634,0.00008110022,0.000918041],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.7529151,"threshold_uncertainty_score":0.9996268,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1005446132672882,"score_gpt":0.431635171643048,"score_spread":0.3310905583757597,"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."}}