{"id":"W4404887887","doi":"10.2196/65454","title":"Classifying Unstructured Text in Electronic Health Records for Mental Health Prediction Models: Large Language Model Evaluation Study","year":2024,"lang":"en","type":"article","venue":"JMIR Medical Informatics","topic":"Machine Learning in Healthcare","field":"Computer Science","cited_by":22,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"National Institute of Mental Health","keywords":"Preprint; Health records; Computer science; Mental health; Language model; Mental model; Natural language processing; Electronic health record; Unstructured data; Artificial intelligence; Data mining; Psychology; World Wide Web; Psychiatry; Health care; Cognitive science; Big data","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":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.00580413,0.0002064933,0.000336525,0.0003592347,0.0002426969,0.0001568251,0.0005661993,0.0001541921,0.00002069343],"category_scores_gemma":[0.0001385186,0.0001817085,0.00006771238,0.0006332732,0.00002122591,0.0009594052,0.0001824454,0.0009969581,0.000007260716],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001483931,"about_ca_system_score_gemma":0.003552302,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0001500469,"about_ca_topic_score_gemma":0.0008183015,"domain_scores_codex":[0.9958116,0.0003374432,0.00118607,0.0003026714,0.001548513,0.0008136687],"domain_scores_gemma":[0.9988183,0.0001240528,0.0002395682,0.0004255718,0.00007639849,0.0003160636],"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":[0.00002996334,0.0002837329,0.0006066269,0.001798596,0.0000409011,0.000002163167,0.267621,0.02124192,5.898036e-7,0.02538084,0.008372712,0.674621],"study_design_scores_gemma":[0.001165955,0.0006390313,0.0002434178,0.0003157827,0.000003229111,0.0000163071,0.004919199,0.9885904,4.094923e-7,0.002974578,0.001001878,0.0001298216],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.07180942,0.001409143,0.9105208,0.009918543,0.0009827927,0.004438528,0.00006656413,0.0005984913,0.000255755],"genre_scores_gemma":[0.9797368,0.0001358619,0.01676778,0.002438648,0.0001307818,0.000515214,0.0002099448,0.00002432096,0.00004066803],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.9673485,"threshold_uncertainty_score":0.7409859,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03635449339094467,"score_gpt":0.3996838041880542,"score_spread":0.3633293107971095,"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."}}