{"id":"W4323359744","doi":"10.1038/s41598-023-30820-8","title":"EHR foundation models improve robustness in the presence of temporal distribution shift","year":2023,"lang":"en","type":"article","venue":"Scientific Reports","topic":"Machine Learning in Healthcare","field":"Computer Science","cited_by":47,"is_retracted":false,"has_abstract":true,"ca_institutions":"Institute for Clinical Evaluative Sciences; SickKids Foundation; Hospital for Sick Children","funders":"","keywords":"Logistic regression; Robustness (evolution); Receiver operating characteristic; Transformer; Computer science; Artificial intelligence; Statistics; Machine learning; Medicine; Mathematics; Engineering","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":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.005313234,0.00009391714,0.0001276458,0.0001580272,0.0002494158,0.0003405446,0.0007000671,0.00005206092,0.000004710135],"category_scores_gemma":[0.0003946743,0.00007100355,0.00004943074,0.002180567,0.0001377171,0.000733004,0.0002464759,0.0001880599,0.000009080046],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00005044857,"about_ca_system_score_gemma":0.0002341264,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0005032562,"about_ca_topic_score_gemma":0.0001253197,"domain_scores_codex":[0.9974573,0.0002312132,0.0005332744,0.0006501788,0.0008237321,0.0003043074],"domain_scores_gemma":[0.9979548,0.000139822,0.0003659547,0.001327235,0.0001658411,0.00004630412],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.00001735542,0.000275648,0.1671168,0.0005684609,0.00001390352,0.0009558807,0.01872683,0.6473722,0.001371488,0.09557161,0.01225647,0.05575332],"study_design_scores_gemma":[0.00005591565,0.00002423295,0.02412336,0.00004033254,0.000001698156,0.00003152905,0.00009347826,0.9123712,0.0002954603,0.06166993,0.001193245,0.00009966658],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.6606913,0.00002610857,0.3296143,0.002723409,0.005890031,0.000588053,0.000003503409,0.0001829438,0.0002804333],"genre_scores_gemma":[0.9983103,9.757118e-7,0.001213096,0.00001224729,0.0000310037,0.00004117137,0.000129827,0.000004636695,0.0002567731],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.337619,"threshold_uncertainty_score":0.328388,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.0317654811580866,"score_gpt":0.3032715348931921,"score_spread":0.2715060537351056,"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."}}