Evaluating Clinical Foundation Models for Early Alzheimer's Disease and Related Dementia Prediction from Longitudinal EHRs
- 1Institute for Population and Precision Health, University of Chicago, Chicago, IL, USA
Abstract
Early identification of Alzheimers disease and related dementias (ADRD) remains challenging despite its importance for timely intervention, management of modifiable risk factors, and care planning. We developed and evaluated ADRD onset prediction models using longitudinal electronic health records (EHRs) from the All of Us Research Program at clinically meaningful lead times of 6, 12, 24, and 36 months before diagnosis, benchmarking interpretable count-based representations against four publicly available pretrained clinical foundation models (CLMBR-T, GPT-style, LLaMA-style, and Mamba) across multiple ADRD phenotype definitions. Count-based models consistently achieved the highest discrimination and calibration across all cohorts and prediction horizons. Predictive performance declined with increasing lead time for all approaches; however, the performance gap between count-based and pretrained representations progressively narrowed, with foundation models achieving comparable AUROC of 0.719 (compared to the AUROC of 0.738 of count-based model) at the 36-month horizon while providing higher sensitivity and F1 scores under a fixed operating threshold. External validation with zero-shot evaluation on UChicago EHRs exhibited limited generalizability for count-based and pretrained clinical foundation model based representations. These findings demonstrate that transparent count-based EHR representations remain the strongest overall approach for ADRD onset prediction, while pretrained clinical foundation models provide complementary advantages for long-term risk identification and establish a benchmark for evaluating transferable clinical representations in temporal ADRD risk prediction.
Related
- No evidence that serum metabolite slopes improve held-out prediction of clinical progression in the Alzheimer’s disease neuroimaging initiative
- FERMAT: An Autoregressive Clinical Foundation Model for Disease Risk and Clinical Event Prediction
- Prediction of cognitive impairment in Parkinson’s disease using TabPFN: an integrative analysis of clinical data, longitudinal trajectories, and lipidomics
- Rethinking Input Complexity in Transformer-Based Clinical Prediction: Implications for Feature Dimensionality and Sequence Length in Longitudinal EHR Data
- Real‐world clinical, biomarker, and imaging outcomes of lecanemab in Alzheimer's disease: a multicenter longitudinal cohort study in Jiangsu, China
BibTeX
@misc{farzana2026evaluating,
title = {{Evaluating Clinical Foundation Models for Early Alzheimer's Disease and Related Dementia Prediction from Longitudinal EHRs}},
author = {S. Farzana and A. Arian and T. Rundek and M. Desvarieux and H. Ahsan},
howpublished = {medRxiv},
year = {2026},
month = sep,
doi = {10.64898/2026.09.01.26361933},
url = {https://www.medrxiv.org/content/10.64898/2026.09.01.26361933v1},
note = {CC BY-NC-ND},
}
RIS
TY - UNPB
TI - Evaluating Clinical Foundation Models for Early Alzheimer's Disease and Related Dementia Prediction from Longitudinal EHRs
AU - S. Farzana
AU - A. Arian
AU - T. Rundek
AU - M. Desvarieux
AU - H. Ahsan
PY - 2026
DA - 2026/09/01
PB - medRxiv
DO - 10.64898/2026.09.01.26361933
UR - https://www.medrxiv.org/content/10.64898/2026.09.01.26361933v1
AB - Early identification of Alzheimers disease and related dementias (ADRD) remains challenging despite its importance for timely intervention, management of modifiable risk factors, and care planning. We developed and evaluated ADRD onset prediction models using longitudinal electronic health records (EHRs) from the All of Us Research Program at clinically meaningful lead times of 6, 12, 24, and 36 months before diagnosis, benchmarking interpretable count-based representations against four publicly available pretrained clinical foundation models (CLMBR-T, GPT-style, LLaMA-style, and Mamba) across multiple ADRD phenotype definitions. Count-based models consistently achieved the highest discrimination and calibration across all cohorts and prediction horizons. Predictive performance declined with increasing lead time for all approaches; however, the performance gap between count-based and pretrained representations progressively narrowed, with foundation models achieving comparable AUROC of 0.719 (compared to the AUROC of 0.738 of count-based model) at the 36-month horizon while providing higher sensitivity and F1 scores under a fixed operating threshold. External validation with zero-shot evaluation on UChicago EHRs exhibited limited generalizability for count-based and pretrained clinical foundation model based representations. These findings demonstrate that transparent count-based EHR representations remain the strongest overall approach for ADRD onset prediction, while pretrained clinical foundation models provide complementary advantages for long-term risk identification and establish a benchmark for evaluating transferable clinical representations in temporal ADRD risk prediction.
ER -