{"id":"W7117138703","doi":"10.1186/s12916-025-04584-z","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":"article","venue":"BMC Medicine","topic":"Machine Learning in Healthcare","field":"Computer Science","cited_by":2,"is_retracted":false,"has_abstract":true,"ca_institutions":"Centre for Global Health Research; St. Michael's Hospital","funders":"Canadian Institutes of Health Research; Bill and Melinda Gates Foundation","keywords":"Sierra leone; Coding (social sciences); Quality (philosophy); MEDLINE; Quality management","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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.03781495,0.001069686,0.0009496681,0.002009153,0.0006576577,0.001858159,0.001820594,0.000959569,0.002514246],"category_scores_gemma":[0.06555098,0.000551946,0.002366896,0.0009027175,0.0006931745,0.001269265,0.001552565,0.001386294,0.0005691305],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.003163906,"about_ca_system_score_gemma":0.002412951,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.04001706,"about_ca_topic_score_gemma":0.02825711,"domain_scores_codex":[0.989639,0.008347248,0.0004399446,0.001122398,0.0002520152,0.0001993532],"domain_scores_gemma":[0.8898725,0.1022405,0.003304983,0.0017422,0.002078954,0.0007608857],"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":[0.006296869,0.0007668944,0.5963443,0.0005495098,0.003434839,0.0003252535,0.001708578,0.3043396,0.0002294856,0.001756732,0.005719657,0.07852833],"study_design_scores_gemma":[0.0003736546,0.001046494,0.0671519,0.0003073481,0.0007359158,0.0001801078,0.0009760203,0.9241204,0.0003023091,0.003543749,0.001173543,0.00008857479],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.987493,0.0007516194,0.007078769,0.001135904,0.00008049819,0.0001723728,0.001998367,0.0003096949,0.0009797955],"genre_scores_gemma":[0.9894893,0.0002080545,0.006832995,0.0002082578,0.00005633966,0.0001444367,0.002571588,0.00004506619,0.0004439582],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.04001706,"threshold_uncertainty_score":0.199987,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.04734636144092669,"score_gpt":0.3481439785694647,"score_spread":0.300797617128538,"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."}}