{"id":"W4412034422","doi":"10.2196/70314","title":"Methods for Analytical Validation of Novel Digital Clinical Measures: Implementation Feasibility Evaluation Using Real-World Datasets","year":2025,"lang":"en","type":"article","venue":"Journal of Medical Internet Research","topic":"Digital Mental Health Interventions","field":"Psychology","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Toronto","funders":"National Heart, Lung, and Blood Institute","keywords":"Preprint; Computer science; Digital health; Data science; Data mining; World Wide Web; Health care","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":["metaresearch"],"consensus_categories":[],"category_scores_codex":[0.2078938,0.001591668,0.0009672113,0.003118453,0.001209235,0.00303681,0.003711493,0.002120468,0.003610769],"category_scores_gemma":[0.4298357,0.001052823,0.002843329,0.002806773,0.002284552,0.003155566,0.004201793,0.002514674,0.0005878676],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.002238607,"about_ca_system_score_gemma":0.006495564,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00463047,"about_ca_topic_score_gemma":0.003720711,"domain_scores_codex":[0.8886773,0.09159999,0.006349124,0.004758341,0.008025953,0.0005893987],"domain_scores_gemma":[0.4938924,0.4127155,0.0268189,0.03379026,0.03129934,0.001483502],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"observational","study_design_scores_codex":[0.006662471,0.004651933,0.2085393,0.00459798,0.004099443,0.0005222765,0.002799848,0.2909538,0.006850614,0.0411512,0.007423443,0.4217477],"study_design_scores_gemma":[0.003277228,0.005538417,0.03613626,0.001555162,0.0008434902,0.0004904078,0.001104994,0.9030907,0.008378429,0.02738835,0.01192716,0.0002694465],"study_design_candidate":"observational","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.09007652,0.0004257905,0.8932983,0.0008837986,0.000151129,0.0110348,0.001520467,0.0008620371,0.001747215],"genre_scores_gemma":[0.213166,0.0001926861,0.7710494,0.00027758,0.0000465348,0.01397299,0.001060298,0.00007585988,0.0001586645],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.2078938,"threshold_uncertainty_score":0.976808,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.7033936381253428,"score_gpt":0.7594059642282517,"score_spread":0.05601232610290885,"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."}}