EHR Foundation Models Improve Robustness in the Presence of Temporal Distribution Shift
Bibliographic record
Abstract
ABSTRACT Background Temporal distribution shift negatively impacts the performance of clinical prediction models over time. Pretraining foundation models using self-supervised learning on electronic health records (EHR) may be effective in acquiring informative global patterns that can improve the robustness of task-specific models. Objective To evaluate the utility of EHR foundation models in improving the in-distribution (ID) and out-of-distribution (OOD) performance of clinical prediction models. Methods The cohort consisted of adult inpatients admitted between 2009-2021. Gated recurrent unit (GRU)- and transformer (TRANS)-based foundation models were pretrained on EHR of patients admitted between 2009-2012 and were subsequently used to construct patient representations (CLMBR). These representations were used to learn logistic regression models (CLMBR GRU and CLMBR TRANS ) to predict hospital mortality, long length of stay, 30-day readmission, and ICU admission. We compared CLMBR GRU and CLMBR TRANS with baseline logistic regression models learned on count-based representations (count-LR) and end-to-end (ETE) GRU and transformer models in ID (2009-2012) and OOD (2013-2021) year groups. Performance was measured using area-under-the-receiver-operating-characteristic curve, area- under-the-precision-recall curve, and absolute calibration error. Results Models trained on CLMBR generally showed better discrimination relative to count-LR in both ID and OOD year groups. In addition, they often matched or were better than their ETE counterparts. Finally, foundation models’ performance in the self-supervised learning task tracked closely with the ID and OOD performance of the downstream models. Conclusions These results suggest that pretraining foundation models on electronic health records is a useful approach for developing clinical prediction models that perform well in the presence of temporal distribution shift.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.003 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.003 | 0.002 |
| Research integrity | 0.000 | 0.001 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".