{"id":"W4399673257","doi":"10.2196/58129","title":"Large Language Models Versus Expert Clinicians in Crisis Prediction Among Telemental Health Patients: Comparative Study","year":2024,"lang":"en","type":"article","venue":"JMIR Mental Health","topic":"Mental Health via Writing","field":"Psychology","cited_by":32,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"","keywords":"Preprint; Mental health; Psychology; Medicine; Psychiatry; Computer science; World Wide Web","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":[],"consensus_categories":[],"category_scores_codex":[0.01342109,0.0006895902,0.0006482466,0.001149228,0.0003375227,0.001312378,0.0007397481,0.001029896,0.00312081],"category_scores_gemma":[0.04158263,0.0003392213,0.0007353164,0.0006702959,0.0005196869,0.002166851,0.00118049,0.001263998,0.0008931729],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001251929,"about_ca_system_score_gemma":0.0009477942,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.005094663,"about_ca_topic_score_gemma":0.004590566,"domain_scores_codex":[0.9944021,0.004551794,0.0001922701,0.0004073062,0.0003020662,0.0001445448],"domain_scores_gemma":[0.9581475,0.03688312,0.001118273,0.001316731,0.001340314,0.001194094],"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":[0.02555543,0.009486854,0.6463799,0.000650426,0.001221673,0.0005028692,0.00612588,0.04216437,0.0007257405,0.001510927,0.005025864,0.26065],"study_design_scores_gemma":[0.003106402,0.01640831,0.2412306,0.0003531177,0.001167009,0.0008176049,0.01000537,0.716176,0.001550732,0.005553368,0.003452491,0.000178922],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9940681,0.0005120248,0.002719663,0.0007223861,0.00004221333,0.0001516906,0.0003141521,0.00005744879,0.001412332],"genre_scores_gemma":[0.9966766,0.0002292741,0.002098717,0.0001552639,0.00004122873,0.0000814336,0.0003947993,0.00001135478,0.0003114606],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.01342109,"threshold_uncertainty_score":0.07097834,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.08244833402517719,"score_gpt":0.4919045461106609,"score_spread":0.4094562120854837,"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."}}