{"id":"W6920720692","doi":"10.60692/gkg8d-rgv05","title":"EHR foundation models improve robustness in the presence of temporal distribution shift","year":2023,"lang":"en","type":"article","venue":"Greater South Information System","topic":"Machine Learning in Healthcare","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Institute for Clinical Evaluative Sciences; SickKids Foundation; Hospital for Sick Children","funders":"","keywords":"Robustness (evolution); Logistic regression; Foundation (evidence); Transformer; Predictive modelling; Regression; Health records","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":[0.001479617,0.0001143662,0.0001530013,0.0002028615,0.0001215639,0.0002122494,0.0007058634,0.00007789189,0.000001200111],"category_scores_gemma":[0.0000616994,0.00008282694,0.0000422504,0.0009936219,0.00002346455,0.001927275,0.000145079,0.0001534142,0.00009770484],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00009159681,"about_ca_system_score_gemma":0.00006780979,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0001446341,"about_ca_topic_score_gemma":0.000002064571,"domain_scores_codex":[0.9982271,0.0002493125,0.0006359508,0.00015785,0.0004957903,0.0002339792],"domain_scores_gemma":[0.9988053,0.00004156897,0.0003797043,0.0005819115,0.0001560705,0.00003548317],"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":[0.00003843016,0.000005537109,0.288213,0.001831507,0.00001358104,0.000006644667,0.2974708,0.3406032,6.874002e-7,0.06630241,0.00014591,0.005368215],"study_design_scores_gemma":[0.0002354089,0.00002819258,0.09273765,0.00008849301,0.000001813436,0.000005862992,0.002921002,0.9037918,0.00001640629,0.00005497356,0.00003541578,0.00008291074],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.3417346,7.71243e-7,0.6565545,0.0004530953,0.0003305897,0.000434127,0.00003111357,0.0002064345,0.0002547887],"genre_scores_gemma":[0.9992046,7.881162e-8,0.0005182594,0.00003681339,0.00003129079,0.00009600492,0.00009462333,0.000003628989,0.00001469579],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.65747,"threshold_uncertainty_score":0.3377585,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.04488758857166815,"score_gpt":0.2602870451111151,"score_spread":0.215399456539447,"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."}}