{"id":"W4403123652","doi":"10.2196/64143","title":"Practical Aspects of Using Large Language Models to Screen Abstracts for Cardiovascular Drug Development: Cross-Sectional Study","year":2024,"lang":"en","type":"article","venue":"JMIR Medical Informatics","topic":"Health Systems, Economic Evaluations, Quality of Life","field":"Economics, Econometrics and Finance","cited_by":7,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"","keywords":"Drug development; Computer science; Drug; Medicine; Process (computing); Intensive care medicine; Risk analysis (engineering); Data science; Pharmacology; Programming language","routes":{"ca_aff":false,"ca_fund":false,"ca_venue":true,"about_ca":false,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"metacan-v3-hybrid-931329e0061c","candidate_categories":["metaresearch"],"consensus_categories":[],"category_scores_codex":[0.08428775,0.0009539403,0.0007919182,0.002600415,0.0008176398,0.002874974,0.001393419,0.001503831,0.005542044],"category_scores_gemma":[0.2433794,0.0007642199,0.002533931,0.001900134,0.0007576235,0.003051529,0.002140461,0.001944541,0.0011],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0009657522,"about_ca_system_score_gemma":0.002344728,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.007203425,"about_ca_topic_score_gemma":0.007301395,"domain_scores_codex":[0.9583419,0.03493947,0.002275259,0.001939164,0.00203118,0.0004730591],"domain_scores_gemma":[0.4501337,0.4902196,0.0288089,0.01575869,0.01265009,0.002428992],"domain_codex":null,"domain_gemma":"methods","domain_candidate":"methods","domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.005312521,0.002518573,0.8964659,0.0004109864,0.003040212,0.0003772862,0.001627379,0.02546622,0.00100376,0.002584525,0.003160639,0.05803195],"study_design_scores_gemma":[0.001244548,0.008622522,0.5552925,0.0006895452,0.00592385,0.001152755,0.004029472,0.3921973,0.005121395,0.01872896,0.006650992,0.0003461428],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9286371,0.001707343,0.06091332,0.002685567,0.00009867204,0.0006561824,0.003034368,0.0003238488,0.001943733],"genre_scores_gemma":[0.976288,0.0003008833,0.020521,0.0002964014,0.0000678041,0.0003736939,0.001380877,0.00003577648,0.0007355452],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.9157122,"threshold_uncertainty_score":0.4457616,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.3344917901473083,"score_gpt":0.4836248899399951,"score_spread":0.1491330997926868,"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."}}