pith. sign in

arxiv: 2606.21885 · v1 · pith:JHRJHHDGnew · submitted 2026-06-20 · 💻 cs.LG · cs.AI

Cohort-Anchored Foundation Models for Electronic Health Records: From Risk Scores to Auditable Peer Cohorts

Kaiping Zheng This is my paper

Pith reviewed 2026-06-26 12:29 UTC · model grok-4.3

classification 💻 cs.LG cs.AI
keywords foundation modelselectronic health recordspatient cohortsclinical AImultimodal alignmentinterpretabilityauditable decisionscohort-conditioned pretraining
0
0 comments X

The pith

Anchoring foundation models to explicit patient cohorts organizes EHR representations around clinically meaningful groups for auditable decisions.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper argues that existing EHR foundation models treat patient comparison as an emergent byproduct of representation learning, which creates problems with interpretability, distribution shift, and alignment to clinical reasoning. It proposes instead to treat patient cohorts as first-class objects by inserting four stages into the pipeline: deviation-aware data curation, cohort-conditioned pretraining, multimodal cohort alignment, and clinician-in-the-loop refinement. These stages are meant to improve data quality, structure representations around comparable patients, keep modality-specific relationships intact, and let clinicians audit outputs against peer groups. The framework is presented as composable on top of existing encoders and is demonstrated on four clinical tasks spanning prediction, risk stratification, imaging triage, and report generation. A sympathetic reader would conclude that making cohorts primary changes the source of clinical evidence from opaque embeddings to explicit peer comparisons.

Core claim

The central claim is that explicitly elevating patient cohorts to first-class objects throughout the learning pipeline, via deviation-aware data curation, cohort-conditioned pretraining, multimodal cohort alignment, and clinician-in-the-loop refinement, yields representations that preserve modality-specific relationships, organize around clinically meaningful cohort structure, and support auditable clinical decision-making. The resulting Cohort-Anchored Foundation Model framework augments existing EHR encoders without modifying them and is illustrated across acute kidney injury prediction, cardiovascular risk from ECGs, optic neuropathy triage from imaging, and electroretinogram-grounded rep

What carries the argument

The Cohort-Anchored Foundation Model (CAFM) framework, which inserts four cohort-centric stages into the training and refinement pipeline to make patient comparison a primary rather than emergent source of clinical evidence.

If this is right

  • The framework augments existing EHR foundation models without changing their underlying encoders.
  • Representations remain organized around clinically meaningful cohort structure while keeping modality-specific relationships intact.
  • Clinician-in-the-loop refinement produces outputs that support direct audit against peer patient groups.
  • The same four-stage structure applies across prediction, risk stratification, imaging triage, and report generation tasks.
  • Five empirically testable hypotheses follow directly from the claim that cohort anchoring improves trustworthiness over standard representation learning.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • If cohort conditioning reduces effective distribution shift within matched groups, the same stages could be applied to other high-stakes domains where peer comparison is already a standard of care.
  • The clinician-in-the-loop stage implicitly treats human feedback as a source of cohort labels, suggesting a route to iterative refinement that does not require full retraining.
  • Open challenges listed in the paper around irregular temporality and evaluation beyond accuracy point to the need for new metrics that score how well a prediction is justified by its retrieved cohort rather than by aggregate accuracy alone.

Load-bearing premise

The four stages can be composed on top of existing encoders while preserving modality-specific relationships and yielding clinically meaningful cohort structure.

What would settle it

A controlled comparison on one of the four case studies in which the cohort-anchored version shows no gain in clinician-rated auditability or agreement with reasoning compared with the unmodified base encoder.

Figures

Figures reproduced from arXiv: 2606.21885 by Kaiping Zheng.

Figure 1
Figure 1. Figure 1: The proposed shift: the same backbone, but a different output unit. Instead of producing a scalar risk score, [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: The cafm framework. Four stages ((1) deviation-aware cohort construction, (2) cohort-conditioned pretraining, (3) multimodal cohort alignment, (4) clinician-in-the-loop refinement) expose cohorts as first-class objects. the model should consume the mask explicitly rather than be asked to ignore it. Counterfactual decision-support frameworks [91] that distinguish “what was observed” from “what would have ha… view at source ↗
Figure 3
Figure 3. Figure 3: Cohort-conditioned pretraining. Anchor xa pulls toward augmentation positives and cohort co-members, while pushing away near-miss negatives that share similar phenotypes but differ in outcomes. InfoNCE, we normalise it over the in-batch comparison set: s˜π(xi , xj ) = sπ(xi , xj ) P xk∈B\{xi} sπ(xi , xk) + ε , (2) such that P xj∈B\{xi} s˜π(xi , xj ) = 1 , thereby yielding a valid cross-entropy interpretati… view at source ↗
Figure 4
Figure 4. Figure 4: Multimodal cohort alignment. Per-modality encoders for text, structured records, imaging, and physiological [PITH_FULL_IMAGE:figures/full_fig_p009_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: The four case studies are illustrated, spanning the principal modalities of EHR data: (a) acute kidney injury [PITH_FULL_IMAGE:figures/full_fig_p014_5.png] view at source ↗
read the original abstract

Foundation models have achieved remarkable performance across medical question answering, imaging, and electronic health record (EHR) tasks, yet reliable clinical deployment remains challenging due to limited interpretability, vulnerability to distribution shift, and weak alignment with clinician reasoning. We argue that these limitations arise because existing approaches prioritize representation learning while treating patient comparison as an emergent property rather than a primary source of clinical evidence. To address this gap, we propose CAFM, a Cohort-Anchored Foundation Model framework that elevates patient cohorts to a first-class object throughout the learning pipeline. The framework consists of four stages: deviation-aware data curation, cohort-conditioned pretraining, multimodal cohort alignment, and clinician-in-the-loop refinement. Together, these stages improve data quality, organize representations around clinically meaningful cohort structure, preserve modality-specific relationships, and support auditable clinical decision-making. The framework is compositional and can augment existing EHR foundation models without modifying their underlying encoders. We illustrate CAFM through four clinical case studies spanning acute kidney injury prediction, cardiovascular risk stratification from electrocardiograms, optic neuropathy triage from orbital imaging, and electroretinogram-grounded report generation. We further present five empirically testable hypotheses and identify open challenges in data quality, irregular temporality, multimodal learning, distribution shift, and evaluation beyond predictive accuracy. We argue that explicitly anchoring foundation models to patient cohorts provides a principled path toward trustworthy clinical AI.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 0 minor

Summary. The manuscript proposes the Cohort-Anchored Foundation Model (CAFM) framework for electronic health records. It claims that elevating patient cohorts to a first-class object via four stages—deviation-aware data curation, cohort-conditioned pretraining, multimodal cohort alignment, and clinician-in-the-loop refinement—addresses limitations in interpretability, distribution shift, and clinical alignment of existing foundation models. The framework is described as compositional, augmenting existing EHR encoders without modification, and is illustrated through four clinical case studies (acute kidney injury prediction, cardiovascular risk from ECGs, optic neuropathy triage, and electroretinogram-grounded report generation). Five empirically testable hypotheses are listed along with open challenges in data quality, temporality, multimodality, shift, and evaluation.

Significance. If the four stages can be implemented to produce auditable cohort structure while preserving modality-specific relationships, the framework could supply a structured route to more trustworthy clinical AI by treating cohort comparison as a primary modeling objective rather than an emergent property. The explicit listing of testable hypotheses and open challenges is a constructive element. The absence of any quantitative results, derivations, or validation data means the significance remains prospective.

major comments (2)
  1. [Abstract] Abstract: the central composability claim—that the four stages 'can augment existing EHR foundation models without modifying their underlying encoders'—is load-bearing for the entire proposal, yet no mechanism is supplied for introducing cohort conditioning into the pretraining objective or performing multimodal alignment without encoder updates, new parameters, or changes to the forward pass or loss.
  2. [Abstract] Abstract (case studies paragraph): the four clinical case studies are presented only as illustrations with no quantitative results, performance metrics, error bars, or cohort-structure validation, leaving the claim that the stages 'yield clinically meaningful cohort structure' untested and preventing assessment of whether modality-specific relationships are preserved.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive review and for recognizing the prospective value of the CAFM framework along with the explicit hypotheses and challenges. We address the two major comments point by point below. Both comments correctly identify that the abstract presents high-level claims without supporting implementation details or empirical data; we agree that revisions are required to clarify the manuscript's conceptual scope.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central composability claim—that the four stages 'can augment existing EHR foundation models without modifying their underlying encoders'—is load-bearing for the entire proposal, yet no mechanism is supplied for introducing cohort conditioning into the pretraining objective or performing multimodal alignment without encoder updates, new parameters, or changes to the forward pass or loss.

    Authors: We agree that the abstract states the composability claim without supplying concrete mechanisms. The manuscript frames CAFM as a compositional framework at the conceptual level, with the four stages intended to operate via modular additions (e.g., auxiliary cohort embeddings or conditioning signals) rather than encoder modification. Because no explicit mechanism, derivation, or pseudocode is provided in the current text, this constitutes a genuine gap. We will revise the abstract and add a dedicated subsection under the framework description to outline candidate mechanisms (such as input-level cohort tokens or auxiliary losses that leave the base encoder unchanged) while preserving the claim that no encoder weights are altered. This revision will be marked clearly as prospective. revision: yes

  2. Referee: [Abstract] Abstract (case studies paragraph): the four clinical case studies are presented only as illustrations with no quantitative results, performance metrics, error bars, or cohort-structure validation, leaving the claim that the stages 'yield clinically meaningful cohort structure' untested and preventing assessment of whether modality-specific relationships are preserved.

    Authors: We agree that the case studies are presented solely as illustrations and contain no quantitative results, metrics, or validation of cohort structure or modality preservation. This is consistent with the manuscript's stated purpose as a framework proposal that lists five empirically testable hypotheses and open challenges rather than reporting experiments. The abstract's phrasing that the stages 'yield clinically meaningful cohort structure' is therefore unsupported by data in the current version. We will revise the abstract and case-study descriptions to explicitly label them as conceptual illustrations, remove any implication of empirical validation, and direct readers to the listed hypotheses for future testing. This change will be made in the next version. revision: yes

Circularity Check

0 steps flagged

Conceptual framework proposal exhibits no circularity

full rationale

The paper advances a high-level architectural proposal (CAFM) consisting of four named stages whose composability on unmodified encoders is asserted at the framework level. No equations, fitted parameters, or derivation steps appear in the provided text. No self-citations are invoked as load-bearing uniqueness theorems, and the central claim does not reduce by construction to any input data or prior result. The argument remains an independent design claim whose empirical testability is explicitly deferred to future work and hypotheses.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 1 invented entities

The framework rests on domain assumptions about the clinical value of cohort structure and the feasibility of compositional application to existing encoders; no free parameters or invented physical entities are introduced.

axioms (2)
  • domain assumption Existing approaches prioritize representation learning while treating patient comparison as an emergent property rather than a primary source of clinical evidence.
    Stated directly in the abstract as the root cause of current limitations.
  • domain assumption Cohort structure can organize representations while preserving modality-specific relationships.
    Invoked in the description of the multimodal cohort alignment stage.
invented entities (1)
  • Cohort-Anchored Foundation Model (CAFM) no independent evidence
    purpose: Framework that elevates patient cohorts to first-class objects in the EHR learning pipeline.
    Newly introduced construct whose value is argued but not independently evidenced in the abstract.

pith-pipeline@v0.9.1-grok · 5774 in / 1416 out tokens · 32611 ms · 2026-06-26T12:29:46.017104+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

151 extracted references · 7 linked inside Pith

  1. [1]

    Scalable and accurate deep learning with electronic health records

    Alvin Rajkomar, Eyal Oren, Kai Chen, Andrew M Dai, Nissan Hajaj, Michaela Hardt, Peter J Liu, Xiaobing Liu, Jake Marcus, Mimi Sun, et al. Scalable and accurate deep learning with electronic health records. NPJ digital medicine, 1(1):18, 2018

    2018

  2. [2]

    Retain: An interpretable predictive model for healthcare using reverse time attention mechanism

    Edward Choi, Mohammad Taha Bahadori, Jimeng Sun, Joshua Kulas, Andy Schuetz, and Walter Stew- art. Retain: An interpretable predictive model for healthcare using reverse time attention mechanism. Advances in neural information processing systems, 29, 2016

    2016

  3. [3]

    Doctor ai: Predicting clinical events via recurrent neural net- works

    Edward Choi, Mohammad Taha Bahadori, Andy Schuetz, Walter F Stewart, and Jimeng Sun. Doctor ai: Predicting clinical events via recurrent neural net- works. InMachine learning for healthcare conference, pages 301–318. PMLR, 2016

    2016

  4. [4]

    Benjamin Shickel, Patrick James Tighe, Azra Bihorac, and Parisa Rashidi. Deep ehr: a survey of recent ad- vances in deep learning techniques for electronic health record (ehr) analysis.IEEE journal of biomedical and health informatics, 22(5):1589–1604, 2017

    2017

  5. [5]

    Attention is all you need

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin. Attention is all you need. 21 Advances in neural information processing systems, 30, 2017

    2017

  6. [6]

    Bert: Pre-training of deep bidi- rectional transformers for language understanding

    Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. Bert: Pre-training of deep bidi- rectional transformers for language understanding. InProceedings of the 2019 conference of the North American chapter of the association for computational linguistics: human language technologies, volume 1 (long and short papers), pages 4171–4186, 2019

    2019

  7. [7]

    Language models are few-shot learners

    Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, et al. Language models are few-shot learners. Advances in neural information processing systems, 33:1877–1901, 2020

    1901

  8. [8]

    Llama: Open and effi- cient foundation language models.arXiv preprint arXiv:2302.13971, 2023

    Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Timothée Lacroix, Baptiste Rozière, Naman Goyal, Eric Ham- bro, Faisal Azhar, et al. Llama: Open and effi- cient foundation language models.arXiv preprint arXiv:2302.13971, 2023

    Pith/arXiv arXiv 2023

  9. [9]

    Gpt-4 technical report.arXiv preprint arXiv:2303.08774, 2023

    Josh Achiam, Steven Adler, Sandhini Agarwal, Lama Ahmad, Ilge Akkaya, Florencia Leoni Aleman, Diogo Almeida, Janko Altenschmidt, Sam Altman, Shyamal Anadkat, et al. Gpt-4 technical report.arXiv preprint arXiv:2303.08774, 2023

    Pith/arXiv arXiv 2023

  10. [10]

    Learning transferable visual models from natu- ral language supervision

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sas- try, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learning transferable visual models from natu- ral language supervision. InInternational conference on machine learning, pages 8748–8763. PmLR, 2021

    2021

  11. [11]

    Dinov2: Learning robust visual fea- tures without supervision.Transactions on Machine Learning Research Journal, 2024

    Maxime Oquab, Timothée Darcet, Théo Moutakanni, Huy Vo, Marc Szafraniec, Vasil Khalidov, Pierre Fer- nandez, Daniel Haziza, Francisco Massa, Alaaeldin El-Nouby, et al. Dinov2: Learning robust visual fea- tures without supervision.Transactions on Machine Learning Research Journal, 2024

    2024

  12. [12]

    Large language models encode clinical knowl- edge.Nature, 620(7972):172–180, 2023

    Karan Singhal, Shekoofeh Azizi, Tao Tu, S Sara Mah- davi, Jason Wei, Hyung Won Chung, Nathan Scales, Ajay Tanwani, Heather Cole-Lewis, Stephen Pfohl, et al. Large language models encode clinical knowl- edge.Nature, 620(7972):172–180, 2023

    2023

  13. [13]

    Toward expert-level medical question answering with large language models.Nature medicine, 31(3):943– 950, 2025

    Karan Singhal, Tao Tu, Juraj Gottweis, Rory Sayres, Ellery Wulczyn, Mohamed Amin, Le Hou, Kevin Clark, Stephen R Pfohl, Heather Cole-Lewis, et al. Toward expert-level medical question answering with large language models.Nature medicine, 31(3):943– 950, 2025

    2025

  14. [14]

    Foundation models for generalist medical artificial intelligence.Nature, 616 (7956):259–265, 2023

    Michael Moor, Oishi Banerjee, Zahra Shakeri Hossein Abad, Harlan M Krumholz, Jure Leskovec, Eric J Topol, and Pranav Rajpurkar. Foundation models for generalist medical artificial intelligence.Nature, 616 (7956):259–265, 2023

    2023

  15. [15]

    Towards generalist biomedical ai.Nejm Ai, 1(3): AIoa2300138, 2024

    Tao Tu, Shekoofeh Azizi, Danny Driess, Mike Schaek- ermann, Mohamed Amin, Pi-Chuan Chang, Andrew Carroll, Charles Lau, Ryutaro Tanno, Ira Ktena, et al. Towards generalist biomedical ai.Nejm Ai, 1(3): AIoa2300138, 2024

    2024

  16. [16]

    A large language model for electronic health records.NPJ digital medicine, 5(1): 194, 2022

    Xi Yang, Aokun Chen, Nima PourNejatian, Hoo Chang Shin, Kaleb E Smith, Christopher Parisien, Colin Compas, Cheryl Martin, Anthony B Costa, Mona G Flores, et al. A large language model for electronic health records.NPJ digital medicine, 5(1): 194, 2022

    2022

  17. [17]

    A visual-language foundation model for compu- tational pathology.Nature medicine, 30(3):863–874, 2024

    Ming Y Lu, Bowen Chen, Drew FK Williamson, Richard J Chen, Ivy Liang, Tong Ding, Guillaume Jaume, Igor Odintsov, Long Phi Le, Georg Gerber, et al. A visual-language foundation model for compu- tational pathology.Nature medicine, 30(3):863–874, 2024

    2024

  18. [18]

    Ecg-fm: An open elec- trocardiogram foundation model.Jamia Open, 8(5): ooaf122, 2025

    Kaden McKeen, Sameer Masood, Augustin Toma, Barry Rubin, and Bo Wang. Ecg-fm: An open elec- trocardiogram foundation model.Jamia Open, 8(5): ooaf122, 2025

    2025

  19. [19]

    Self-supervised learning in medicine and health- care.Nature Biomedical Engineering, 6(12):1346–1352, 2022

    Rayan Krishnan, Pranav Rajpurkar, and Eric J Topol. Self-supervised learning in medicine and health- care.Nature Biomedical Engineering, 6(12):1346–1352, 2022

    2022

  20. [20]

    Capabilities of gpt-4 on medical challenge problems.arXiv preprint arXiv:2303.13375, 2023

    Harsha Nori, Nicholas King, Scott Mayer McKinney, Dean Carignan, and Eric Horvitz. Capabilities of gpt-4 on medical challenge problems.arXiv preprint arXiv:2303.13375, 2023

    Pith/arXiv arXiv 2023

  21. [21]

    Can large language models reason about medical questions?Pat- terns, 5(3), 2024

    Valentin Liévin, Christoffer Egeberg Hother, An- dreas Geert Motzfeldt, and Ole Winther. Can large language models reason about medical questions?Pat- terns, 5(3), 2024

    2024

  22. [22]

    Survey of hallucination in natural language generation.ACM computing surveys, 55(12):1–38, 2023

    Ziwei Ji, Nayeon Lee, Rita Frieske, Tiezheng Yu, Dan Su, Yan Xu, Etsuko Ishii, Ye Jin Bang, Andrea Madotto, and Pascale Fung. Survey of hallucination in natural language generation.ACM computing surveys, 55(12):1–38, 2023

    2023

  23. [23]

    Med-halt: Medical domain hallu- cination test for large language models

    Ankit Pal, Logesh Kumar Umapathi, and Malaikan- nan Sankarasubbu. Med-halt: Medical domain hallu- cination test for large language models. InProceedings of the 27th Conference on Computational Natural Lan- guage Learning (CoNLL), pages 314–334, 2023

    2023

  24. [24]

    Variable generalization performance of a deep learning model to detect pneumonia in chest radio- graphs: a cross-sectional study.PLoS medicine, 15 (11):e1002683, 2018

    John R Zech, Marcus A Badgeley, Manway Liu, An- thony B Costa, Joseph J Titano, and Eric Karl Oer- mann. Variable generalization performance of a deep learning model to detect pneumonia in chest radio- graphs: a cross-sectional study.PLoS medicine, 15 (11):e1002683, 2018

    2018

  25. [25]

    The clini- cian and dataset shift in artificial intelligence.New England Journal of Medicine, 385(3):283–286, 2021

    Samuel G Finlayson, Adarsh Subbaswamy, Karan- deep Singh, John Bowers, Annabel Kupke, Jonathan Zittrain, Isaac S Kohane, and Suchi Saria. The clini- cian and dataset shift in artificial intelligence.New England Journal of Medicine, 385(3):283–286, 2021

    2021

  26. [26]

    The limits of fair medicalimagingaiinreal-worldgeneralization.Nature medicine, 30(10):2838–2848, 2024

    Yuzhe Yang, Haoran Zhang, Judy W Gichoya, Dina Katabi, and Marzyeh Ghassemi. The limits of fair medicalimagingaiinreal-worldgeneralization.Nature medicine, 30(10):2838–2848, 2024

    2024

  27. [27]

    The myth of generalisability in clinical research and machine learning in health care.The Lancet Digital Health, 2 (9):e489–e492, 2020

    Joseph Futoma, Morgan Simons, Trishan Panch, Fi- nale Doshi-Velez, and Leo Anthony Celi. The myth of generalisability in clinical research and machine learning in health care.The Lancet Digital Health, 2 (9):e489–e492, 2020. 22

    2020

  28. [28]

    The false hope of current approaches to explainable artificial intelligence in health care.The lancet digital health, 3(11):e745–e750, 2021

    Marzyeh Ghassemi, Luke Oakden-Rayner, and An- drew L Beam. The false hope of current approaches to explainable artificial intelligence in health care.The lancet digital health, 3(11):e745–e750, 2021

    2021

  29. [29]

    Stop explaining black box machine learning models for high stakes decisions and use inter- pretable models instead.Nature machine intelligence, 1(5):206–215, 2019

    Cynthia Rudin. Stop explaining black box machine learning models for high stakes decisions and use inter- pretable models instead.Nature machine intelligence, 1(5):206–215, 2019

    2019

  30. [30]

    Behrt: transformer for electronic health records.Scientific reports, 10(1):7155, 2020

    Yikuan Li, Shishir Rao, José Roberto Ayala Solares, Abdelaali Hassaine, Rema Ramakrishnan, Dexter Canoy, Yajie Zhu, Kazem Rahimi, and Gholamreza Salimi-Khorshidi. Behrt: transformer for electronic health records.Scientific reports, 10(1):7155, 2020

    2020

  31. [31]

    Med-bert: pretrained contextualized em- beddings on large-scale structured electronic health records for disease prediction.NPJ digital medicine, 4(1):86, 2021

    Laila Rasmy, Yang Xiang, Ziqian Xie, Cui Tao, and Degui Zhi. Med-bert: pretrained contextualized em- beddings on large-scale structured electronic health records for disease prediction.NPJ digital medicine, 4(1):86, 2021

    2021

  32. [32]

    Cehr-bert: Incorporating tempo- ral information from structured ehr data to improve prediction tasks

    Chao Pang, Xinzhuo Jiang, Krishna S Kalluri, Matthew Spotnitz, RuiJun Chen, Adler Perotte, and Karthik Natarajan. Cehr-bert: Incorporating tempo- ral information from structured ehr data to improve prediction tasks. InMachine learning for health, pages 239–260. PMLR, 2021

    2021

  33. [33]

    MOTOR: A time-to-event foundation model for structured medical records

    Ethan Steinberg, Jason Fries, Yizhe Xu, and Nigam Shah. MOTOR: A time-to-event foundation model for structured medical records. InICLR. OpenReview.net, 2024

    2024

  34. [34]

    The shaky founda- tions of large language models and foundation models for electronic health records.npj digital medicine, 6 (1):135, 2023

    Michael Wornow, Yizhe Xu, Rahul Thapa, Birju Patel, Ethan Steinberg, Scott Fleming, Michael A Pfeffer, Jason Fries, and Nigam H Shah. The shaky founda- tions of large language models and foundation models for electronic health records.npj digital medicine, 6 (1):135, 2023

    2023

  35. [35]

    Ehrshot: An ehr bench- mark for few-shot evaluation of foundation models

    Michael Wornow, Rahul Thapa, Ethan Steinberg, Ja- son Fries, and Nigam Shah. Ehrshot: An ehr bench- mark for few-shot evaluation of foundation models. Advances in Neural Information Processing Systems, 36:67125–67137, 2023

    2023

  36. [36]

    Holistic evaluation of large language models for medical tasks with medhelm.Nature Medicine, pages 1–9, 2026

    Suhana Bedi, Hejie Cui, Miguel Fuentes, Alyssa Unell, Michael Wornow, Juan M Banda, Nikesh Kotecha, Timothy Keyes, Yifan Mai, Mert Oez, et al. Holistic evaluation of large language models for medical tasks with medhelm.Nature Medicine, pages 1–9, 2026

    2026

  37. [37]

    Medmcqa: A large-scale multi- subject multi-choice dataset for medical domain ques- tion answering

    Ankit Pal, Logesh Kumar Umapathi, and Malaikan- nan Sankarasubbu. Medmcqa: A large-scale multi- subject multi-choice dataset for medical domain ques- tion answering. InConference on health, inference, and learning, pages 248–260. PMLR, 2022

    2022

  38. [38]

    Next-generation phenotyping of electronic health records.Journal of the American Medical Informatics Association, 20(1): 117–121, 2013

    George Hripcsak and David J Albers. Next-generation phenotyping of electronic health records.Journal of the American Medical Informatics Association, 20(1): 117–121, 2013

    2013

  39. [39]

    Phewas: demonstrating the feasibility of a phenome-wide scan to discover gene– disease associations.Bioinformatics, 26(9):1205–1210, 2010

    Joshua C Denny, Marylyn D Ritchie, Melissa A Bas- ford, Jill M Pulley, Lisa Bastarache, Kristin Brown- Gentry, Deede Wang, Dan R Masys, Dan M Roden, and Dana C Crawford. Phewas: demonstrating the feasibility of a phenome-wide scan to discover gene– disease associations.Bioinformatics, 26(9):1205–1210, 2010

    2010

  40. [40]

    Observational health data sciences and informatics (ohdsi): opportunities for observational researchers.Studies in health technology and informatics, 216:574, 2015

    George Hripcsak, Jon D Duke, Nigam H Shah, Chris- tian G Reich, Vojtech Huser, Martijn J Schuemie, Marc A Suchard, Rae Woong Park, Ian Chi Kei Wong, Peter R Rijnbeek, et al. Observational health data sciences and informatics (ohdsi): opportunities for observational researchers.Studies in health technology and informatics, 216:574, 2015

    2015

  41. [41]

    Cohortnet: Em- powering cohort discovery for interpretable healthcare analytics.Proceedings of the VLDB Endowment, 17 (10):2487–2500, 2024

    Qingpeng Cai, Kaiping Zheng, HV Jagadish, Beng Chin Ooi, and James Yip. Cohortnet: Em- powering cohort discovery for interpretable healthcare analytics.Proceedings of the VLDB Endowment, 17 (10):2487–2500, 2024

    2024

  42. [42]

    Neuralcohort: Cohort-aware neural representa- tion learning for healthcare analytics

    Changshuo Liu, Lingze Zeng, Kaiping Zheng, Shaofeng Cai, Beng Chin Ooi, and James Wei Luen Yip. Neuralcohort: Cohort-aware neural representa- tion learning for healthcare analytics. InForty-second International Conference on Machine Learning, 2025

    2025

  43. [43]

    Exploiting negative samples: a catalyst for co- hort discovery in healthcare analytics

    Kaiping Zheng, Horng-Ruey Chua, Melanie Herschel, HV Jagadish, Beng Chin Ooi, and James Wei Luen Yip. Exploiting negative samples: a catalyst for co- hort discovery in healthcare analytics. InForty-first International Conference on Machine Learning, 2024

    2024

  44. [44]

    Benchmarking retrieval-augmented generation for medicine

    Guangzhi Xiong, Qiao Jin, Zhiyong Lu, and Aidong Zhang. Benchmarking retrieval-augmented generation for medicine. InFindings of the Association for Com- putational Linguistics: ACL 2024, pages 6233–6251, 2024

    2024

  45. [45]

    Ram-ehr: Retrieval augmentation meets clinical pre- dictions on electronic health records

    Ran Xu, Wenqi Shi, Yue Yu, Yuchen Zhuang, Bowen Jin, May Dongmei Wang, Joyce Ho, and Carl Yang. Ram-ehr: Retrieval augmentation meets clinical pre- dictions on electronic health records. InProceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers), pages 754–765, 2024

    2024

  46. [46]

    Multimodal biomedical ai.Nature medicine, 28(9):1773–1784, 2022

    Julián N Acosta, Guido J Falcone, Pranav Rajpurkar, and Eric J Topol. Multimodal biomedical ai.Nature medicine, 28(9):1773–1784, 2022

    2022

  47. [47]

    Ai in health and medicine.Nature medicine, 28(1):31–38, 2022

    Pranav Rajpurkar, Emma Chen, Oishi Banerjee, and Eric J Topol. Ai in health and medicine.Nature medicine, 28(1):31–38, 2022

    2022

  48. [48]

    High-performance medicine: the con- vergence of human and artificial intelligence.Nature medicine, 25(1):44–56, 2019

    Eric J Topol. High-performance medicine: the con- vergence of human and artificial intelligence.Nature medicine, 25(1):44–56, 2019

    2019

  49. [49]

    Large language models in medicine.Nature medicine, 29(8):1930–1940, 2023

    Arun James Thirunavukarasu, Darren Shu Jeng Ting, Kabilan Elangovan, Laura Gutierrez, Ting Fang Tan, and Daniel Shu Wei Ting. Large language models in medicine.Nature medicine, 29(8):1930–1940, 2023

    1930

  50. [50]

    Publicly available clinical bert embed- dings

    Emily Alsentzer, John Murphy, William Boag, Wei- Hung Weng, Di Jindi, Tristan Naumann, and Matthew McDermott. Publicly available clinical bert embed- dings. InProceedings of the 2nd clinical natural lan- guage processing workshop, pages 72–78, 2019

    2019

  51. [51]

    Biobert: a pre-trained biomedical lan- guage representation model for biomedical text mining

    Jinhyuk Lee, Wonjin Yoon, Sungdong Kim, Donghyeon Kim, Sunkyu Kim, Chan Ho So, and Jaewoo Kang. Biobert: a pre-trained biomedical lan- guage representation model for biomedical text mining. Bioinformatics, 36(4):1234–1240, 2020. 23

    2020

  52. [52]

    Domain-specific language model pretraining for biomedical natural language processing.ACM Transactions on Comput- ing for Healthcare (HEALTH), 3(1):1–23, 2021

    Yu Gu, Robert Tinn, Hao Cheng, Michael Lucas, Naoto Usuyama, Xiaodong Liu, Tristan Naumann, Jianfeng Gao, and Hoifung Poon. Domain-specific language model pretraining for biomedical natural language processing.ACM Transactions on Comput- ing for Healthcare (HEALTH), 3(1):1–23, 2021

    2021

  53. [53]

    Biogpt: gen- erative pre-trained transformer for biomedical text generation and mining.Briefings in bioinformatics, 23(6):bbac409, 2022

    Renqian Luo, Liai Sun, Yingce Xia, Tao Qin, Sheng Zhang, Hoifung Poon, and Tie-Yan Liu. Biogpt: gen- erative pre-trained transformer for biomedical text generation and mining.Briefings in bioinformatics, 23(6):bbac409, 2022

    2022

  54. [54]

    Capa- bilities of gemini models in medicine.arXiv preprint arXiv:2404.18416, 2024

    Khaled Saab, Tao Tu, Wei-Hung Weng, Ryutaro Tanno, David Stutz, Ellery Wulczyn, Fan Zhang, Tim Strother, Chunjong Park, Elahe Vedadi, et al. Capa- bilities of gemini models in medicine.arXiv preprint arXiv:2404.18416, 2024

    Pith/arXiv arXiv 2024

  55. [55]

    Do we still need clinical language models? InConference on health, inference, and learning, pages 578–597

    Eric Lehman, Evan Hernandez, Diwakar Mahajan, Jonas Wulff, Micah J Smith, Zachary Ziegler, Daniel Nadler, Peter Szolovits, Alistair Johnson, and Emily Alsentzer. Do we still need clinical language models? InConference on health, inference, and learning, pages 578–597. PMLR, 2023

    2023

  56. [56]

    Cehr-gpt: Generating electronic health records with chronological patient timelines.arXiv preprint arXiv:2402.04400, 2024

    Chao Pang, Xinzhuo Jiang, Nishanth Parameshwar Pavinkurve, Krishna S Kalluri, Elise L Minto, Jason Patterson, Linying Zhang, George Hripcsak, Gamze Gürsoy, Noémie Elhadad, et al. Cehr-gpt: Generating electronic health records with chronological patient timelines.arXiv preprint arXiv:2402.04400, 2024

    arXiv 2024

  57. [57]

    Biomedclip: a multimodal biomedical foundation model pretrained from fifteen million scientific image-text pairs.arXiv preprint arXiv:2303.00915, 2023

    Sheng Zhang, Yanbo Xu, Naoto Usuyama, Hanwen Xu, Jaspreet Bagga, Robert Tinn, Sam Preston, Ra- jesh Rao, Mu Wei, Naveen Valluri, et al. Biomedclip: a multimodal biomedical foundation model pretrained from fifteen million scientific image-text pairs.arXiv preprint arXiv:2303.00915, 2023

    Pith/arXiv arXiv 2023

  58. [58]

    A visual–language foundation model for pathology image analysis using medical twitter.Nature medicine, 29(9):2307–2316, 2023

    Zhi Huang, Federico Bianchi, Mert Yuksekgonul, Thomas J Montine, and James Zou. A visual–language foundation model for pathology image analysis using medical twitter.Nature medicine, 29(9):2307–2316, 2023

    2023

  59. [59]

    Med- flamingo: a multimodal medical few-shot learner

    Michael Moor, Qian Huang, Shirley Wu, Michihiro Yasunaga, Yash Dalmia, Jure Leskovec, Cyril Zakka, Eduardo Pontes Reis, and Pranav Rajpurkar. Med- flamingo: a multimodal medical few-shot learner. In Machine learning for health (ML4H), pages 353–367. PMLR, 2023

    2023

  60. [60]

    Guiding masked representation learn- ing to capture spatio-temporal relationship of electro- cardiogram

    Yeongyeon Na, Minje Park, Yunwon Tae, and Sunghoon Joo. Guiding masked representation learn- ing to capture spatio-temporal relationship of electro- cardiogram. InICLR. OpenReview.net, 2024

    2024

  61. [61]

    Deep learning for ecg analysis: Benchmarks and insights from ptb-xl.IEEE journal of biomedical and health informatics, 25(5):1519–1528, 2020

    Nils Strodthoff, Patrick Wagner, Tobias Schaeffter, and Wojciech Samek. Deep learning for ecg analysis: Benchmarks and insights from ptb-xl.IEEE journal of biomedical and health informatics, 25(5):1519–1528, 2020

    2020

  62. [62]

    Scaling wear- able foundation models

    Girish Narayanswamy, Xin Liu, Kumar Ayush, Yuzhe Yang, Xuhai Xu, Shun Liao, Jake Garrison, Shyam Tailor, Jacob Sunshine, Yun Liu, et al. Scaling wear- able foundation models. InICLR. OpenReview.net, 2025

    2025

  63. [63]

    Screen- ing for cardiac contractile dysfunction using an arti- ficial intelligence–enabled electrocardiogram.Nature medicine, 25(1):70–74, 2019

    Zachi I Attia, Suraj Kapa, Francisco Lopez-Jimenez, Paul M McKie, Dorothy J Ladewig, Gaurav Satam, Patricia A Pellikka, Maurice Enriquez-Sarano, Pe- ter A Noseworthy, Thomas M Munger, et al. Screen- ing for cardiac contractile dysfunction using an arti- ficial intelligence–enabled electrocardiogram.Nature medicine, 25(1):70–74, 2019

    2019

  64. [64]

    Zachi I Attia, Peter A Noseworthy, Francisco Lopez- Jimenez, Samuel J Asirvatham, Abhishek J Desh- mukh, Bernard J Gersh, Rickey E Carter, Xiaoxi Yao, Alejandro A Rabinstein, Brad J Erickson, et al. An artificial intelligence-enabled ecg algorithm for the identification of patients with atrial fibrillation during sinus rhythm: a retrospective analysis of ...

    2019

  65. [65]

    Diyhealth suite: Dataset, model, and bench- mark for health management at home

    Changshuo Liu, Junran Wu, Zhongle Xie, Wenqiao Zhang, Kaiping Zheng, Jiaqi Zhu, Qingpeng Cai, Ooi Gene Anne, Marcus Chun Jin Tan, Jianwei Yin, et al. Diyhealth suite: Dataset, model, and bench- mark for health management at home. InForty-third International Conference on Machine Learning, 2026

    2026

  66. [66]

    On the opportunities and risks of foundation models.arXiv preprint arXiv:2108.07258, 2021

    Rishi Bommasani, Drew A Hudson, Ehsan Adeli, Russ Altman, Simran Arora, Sydney von Arx, Michael S Bernstein, Jeannette Bohg, Antoine Bosselut, Emma Brunskill, et al. On the opportunities and risks of foundation models.arXiv preprint arXiv:2108.07258, 2021

    Pith/arXiv arXiv 2021

  67. [67]

    Transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (tripod): the tripod statement

    Gary S Collins, Johannes B Reitsma, Douglas G Alt- man, and Karel GM Moons. Transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (tripod): the tripod statement. Journal of British Surgery, 102(3):148–158, 2015

    2015

  68. [68]

    Tripod+ ai state- ment: updated guidance for reporting clinical predic- tion models that use regression or machine learning methods.bmj, 385, 2024

    Gary S Collins, Karel GM Moons, Paula Dhiman, Richard D Riley, Andrew L Beam, Ben Van Calster, Marzyeh Ghassemi, Xiaoxuan Liu, Johannes B Re- itsma, Maarten Van Smeden, et al. Tripod+ ai state- ment: updated guidance for reporting clinical predic- tion models that use regression or machine learning methods.bmj, 385, 2024

    2024

  69. [69]

    The tripod-llm reporting guideline for studies using large language models.Nature medicine, 31(1):60–69, 2025

    Jack Gallifant, Majid Afshar, Saleem Ameen, Yin- dalon Aphinyanaphongs, Shan Chen, Giovanni Cac- ciamani, Dina Demner-Fushman, Dmitriy Dligach, Roxana Daneshjou, Chrystinne Fernandes, et al. The tripod-llm reporting guideline for studies using large language models.Nature medicine, 31(1):60–69, 2025

    2025

  70. [70]

    Gram: graph- based attention model for healthcare representation learning

    Edward Choi, Mohammad Taha Bahadori, Le Song, Walter F Stewart, and Jimeng Sun. Gram: graph- based attention model for healthcare representation learning. InProceedings of the 23rd ACM SIGKDD international conference on knowledge discovery and data mining, pages 787–795, 2017

    2017

  71. [71]

    Patient subtyping via time- aware lstm networks

    Inci M Baytas, Cao Xiao, Xi Zhang, Fei Wang, Anil K Jain, and Jiayu Zhou. Patient subtyping via time- aware lstm networks. InProceedings of the 23rd ACM SIGKDD international conference on knowledge dis- covery and data mining, pages 65–74, 2017

    2017

  72. [72]

    Learn- ing tasks for multitask learning: Heterogenous patient populations in the icu

    Harini Suresh, Jen J Gong, and John V Guttag. Learn- ing tasks for multitask learning: Heterogenous patient populations in the icu. InProceedings of the 24th 24 ACM SIGKDD International Conference on Knowl- edge Discovery & Data Mining, pages 802–810, 2018

    2018

  73. [73]

    Protosteer: Steering deep sequence model with prototypes.IEEE transactions on visualization and computer graphics, 26(1):238–248, 2019

    Yao Ming, Panpan Xu, Furui Cheng, Huamin Qu, and Liu Ren. Protosteer: Steering deep sequence model with prototypes.IEEE transactions on visualization and computer graphics, 26(1):238–248, 2019

    2019

  74. [74]

    Mime: Multilevel medical embedding of elec- tronic health records for predictive healthcare.Ad- vances in neural information processing systems, 31, 2018

    Edward Choi, Cao Xiao, Walter Stewart, and Jimeng Sun. Mime: Multilevel medical embedding of elec- tronic health records for predictive healthcare.Ad- vances in neural information processing systems, 31, 2018

    2018

  75. [75]

    Electronic medical record phenotyping using the anchor and learn framework.Journal of the American Medical Informatics Association, 23(4): 731–740, 2016

    Yoni Halpern, Steven Horng, Youngduck Choi, and David Sontag. Electronic medical record phenotyping using the anchor and learn framework.Journal of the American Medical Informatics Association, 23(4): 731–740, 2016

    2016

  76. [76]

    A simple framework for con- trastive learning of visual representations

    Ting Chen, Simon Kornblith, Mohammad Norouzi, and Geoffrey Hinton. A simple framework for con- trastive learning of visual representations. InInter- national conference on machine learning, pages 1597–

  77. [77]

    Momentum contrast for unsupervised visual representation learning

    Kaiming He, Haoqi Fan, Yuxin Wu, Saining Xie, and Ross Girshick. Momentum contrast for unsupervised visual representation learning. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 9729–9738, 2020

    2020

  78. [78]

    Emerging properties in self-supervised vision transformers

    Mathilde Caron, Hugo Touvron, Ishan Misra, Hervé Jégou, Julien Mairal, Piotr Bojanowski, and Armand Joulin. Emerging properties in self-supervised vision transformers. InProceedings of the IEEE/CVF inter- national conference on computer vision, pages 9650– 9660, 2021

    2021

  79. [79]

    Elda: learning explicit dual- interactions for healthcare analytics

    Qingpeng Cai, Kaiping Zheng, Beng Chin Ooi, Wei Wang, and Chang Yao. Elda: learning explicit dual- interactions for healthcare analytics. In2022 IEEE 38th International Conference on Data Engineering (ICDE), pages 393–406. IEEE, 2022

    2022

  80. [80]

    Arm-net: Adaptive relation modeling network for structured data

    Shaofeng Cai, Kaiping Zheng, Gang Chen, HV Ja- gadish, Beng Chin Ooi, and Meihui Zhang. Arm-net: Adaptive relation modeling network for structured data. InProceedings of the 2021 international confer- ence on management of data, pages 207–220, 2021

    2021

Showing first 80 references.