{"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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.005474109,0.001106109,0.0008381291,0.0007978699,0.0003715028,0.00132986,0.001119669,0.0009094211,0.001157759],"category_scores_gemma":[0.02028947,0.0004836611,0.001116557,0.0005461172,0.0004757613,0.002122568,0.001807843,0.002238256,0.0007313368],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0008588409,"about_ca_system_score_gemma":0.001407828,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.01099682,"about_ca_topic_score_gemma":0.01047121,"domain_scores_codex":[0.9984111,0.0005545497,0.0001327425,0.0005038187,0.0002191044,0.0001787419],"domain_scores_gemma":[0.9919373,0.005250398,0.0006850531,0.0009607603,0.0009620022,0.0002045285],"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":[0.0008465584,0.000495167,0.06758431,0.0001603237,0.0004096094,0.0002902555,0.0004175476,0.6351278,0.006018311,0.002086148,0.005792227,0.2807717],"study_design_scores_gemma":[0.00002257773,0.0001489251,0.004360904,0.00001837746,0.0000360874,0.00005347393,0.00003899081,0.9914299,0.00130184,0.001957522,0.0006123573,0.00001906468],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.6789638,0.0009823447,0.3077769,0.001782631,0.0001746123,0.0001741175,0.001342501,0.005376595,0.003426509],"genre_scores_gemma":[0.973581,0.0001690195,0.02323518,0.0002999732,0.00004998772,0.00006830244,0.001483976,0.00009965146,0.001012996],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.01099682,"threshold_uncertainty_score":0.02895015,"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."}}