{"id":"W3212889536","doi":"10.2196/32832","title":"Diagnostic Performance of an App-Based Symptom Checker in Mental Disorders: Comparative Study in Psychotherapy Outpatients","year":2021,"lang":"en","type":"article","venue":"JMIR Mental Health","topic":"Digital Mental Health Interventions","field":"Psychology","cited_by":34,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"","keywords":"Medical diagnosis; Mental health; Usability; Context (archaeology); Inter-rater reliability; Psychology; Triage; Health care; Clinical psychology; Psychiatry; Medicine; Rating scale; Computer science","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":["metaepi_narrow"],"consensus_categories":[],"category_scores_codex":[0.0003491919,0.0002500988,0.0004941456,0.0002659355,0.00009052314,0.00001777757,0.0002163559,0.00006576259,0.0003673718],"category_scores_gemma":[0.000003778164,0.0002696643,0.0000794219,0.0006049942,0.00009143777,0.0002687142,0.00003251877,0.0002589998,0.00009059047],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.000539406,"about_ca_system_score_gemma":0.0001551825,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.001173409,"about_ca_topic_score_gemma":0.01122712,"domain_scores_codex":[0.996938,0.0005448645,0.001041151,0.0005608539,0.0003517513,0.0005633105],"domain_scores_gemma":[0.9989923,0.00008521602,0.0002822003,0.0004127088,0.00002624503,0.0002013167],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.001429951,0.05725161,0.8487725,0.0003129544,0.00004496563,0.00001470766,0.05642851,0.00002673696,0.00001388641,0.0002669238,0.0005369632,0.03490027],"study_design_scores_gemma":[0.01009199,0.009845539,0.9583265,0.0007201816,0.000002671281,0.000005009233,0.01950185,0.0006100305,0.0001994259,0.00009422507,0.0003487244,0.0002538153],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9932578,0.0003723079,0.000002700725,0.0006023513,0.0009968058,0.003421214,0.0002084397,0.00003460326,0.001103737],"genre_scores_gemma":[0.9977755,0.00002473373,0.0001103336,0.0005023092,0.00001814981,0.0009506957,0.0003110712,0.00003180246,0.000275439],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.109554,"threshold_uncertainty_score":0.9999756,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.04310502899517276,"score_gpt":0.4441968307297203,"score_spread":0.4010918017345476,"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."}}