{"id":"W4416969343","doi":"10.2196/78309","title":"Unsupervised Characterization of Temporal Dataset Shifts as an Early Indicator of AI Performance Variations: Evaluation Study Using the Medical Information Mart for Intensive Care-IV Dataset","year":2025,"lang":"en","type":"article","venue":"JMIR Medical Informatics","topic":"Healthcare Technology and Patient Monitoring","field":"Medicine","cited_by":1,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"","keywords":"Training set; Clinical Practice; Characterization (materials science); Health informatics; Change detection","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":[0.001939506,0.0001601891,0.0003849336,0.000397328,0.0001729232,0.00001921865,0.0003693521,0.0003928642,0.000079014],"category_scores_gemma":[0.001965799,0.0001183239,0.00003812989,0.0004928142,0.0001999592,0.00108125,0.0001424752,0.0005276894,0.000006445324],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001074973,"about_ca_system_score_gemma":0.001800308,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0001206245,"about_ca_topic_score_gemma":0.0000147255,"domain_scores_codex":[0.9965734,0.0001444391,0.001479801,0.0001097899,0.001475669,0.0002169014],"domain_scores_gemma":[0.9973227,0.000163116,0.0005600675,0.0005877728,0.001191384,0.0001749565],"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":[0.002325523,0.001456827,0.634296,0.008644072,0.0007193412,0.000006183815,0.1230657,0.00002105983,0.0001857891,0.001260871,0.0038111,0.2242076],"study_design_scores_gemma":[0.0158413,0.006378124,0.420824,0.003583388,0.001160814,0.00004157129,0.08997593,0.4491238,0.002056824,0.0001371157,0.0103204,0.0005567889],"study_design_candidate":"observational","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9897247,0.0000109711,0.003894763,0.0008897244,0.0003395759,0.003119441,0.001970041,0.0000359813,0.00001478902],"genre_scores_gemma":[0.9564675,0.00001604369,0.0004277749,0.002491758,0.0000606435,0.0002416256,0.04028673,0.000007020101,9.000448e-7],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.4491027,"threshold_uncertainty_score":0.4825107,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.05172330938516912,"score_gpt":0.4106883063506143,"score_spread":0.3589649969654452,"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."}}