{"id":"W4411181503","doi":"10.1177/20552076251348850","title":"Answering real-world clinical questions using large language model, retrieval-augmented generation, and agentic systems","year":2025,"lang":"en","type":"article","venue":"Digital Health","topic":"Artificial Intelligence in Healthcare and Education","field":"Medicine","cited_by":33,"is_retracted":false,"has_abstract":true,"ca_institutions":"Institute for Clinical Evaluative Sciences; SickKids Foundation; Hospital for Sick Children","funders":"","keywords":"Question answering; Computer science; Natural language processing; Psychology; Information retrieval","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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.02663529,0.001730762,0.001027896,0.002624159,0.0008297205,0.003920905,0.002274428,0.002639974,0.005208817],"category_scores_gemma":[0.08449626,0.0007178416,0.001821653,0.001062091,0.001365535,0.00307289,0.003716618,0.001861507,0.001439519],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001967668,"about_ca_system_score_gemma":0.002866902,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.002774024,"about_ca_topic_score_gemma":0.004206572,"domain_scores_codex":[0.9792982,0.01519338,0.001544837,0.002103633,0.001612897,0.0002470393],"domain_scores_gemma":[0.8997135,0.08684385,0.004886299,0.003917679,0.00382149,0.0008173028],"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":[0.004126356,0.001921359,0.02462132,0.006823096,0.001241135,0.00185405,0.009523885,0.1504083,0.03108518,0.01528195,0.03099598,0.7221174],"study_design_scores_gemma":[0.001621823,0.001647315,0.005347813,0.0008355317,0.0007344536,0.0009339141,0.00163982,0.8803695,0.02524799,0.0506031,0.03063335,0.000385393],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.1016859,0.002185029,0.8598198,0.005275519,0.0003112117,0.004583597,0.002490665,0.01803709,0.005611181],"genre_scores_gemma":[0.2799351,0.0004188591,0.7117692,0.001771024,0.0001786788,0.002162938,0.002107918,0.0003323881,0.001323892],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.02663529,"threshold_uncertainty_score":0.1408626,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.2412290545891537,"score_gpt":0.5220397402381639,"score_spread":0.2808106856490102,"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."}}