The DOI matches its Crossref record; 77 of the 77 references it cites were found in a registry or an index; 43 of the 65 sources its claims cite were checked for support.
MoMA: a mixture-of-multimodal-agents architecture for enhancing clinical prediction modelling
Open accessAbstract
Multimodal electronic health record (EHR) data provide richer, complementary insights into patient health compared to single-modality data. However, effectively integrating diverse data modalities for clinical prediction modeling remains challenging due to the substantial data requirements. We introduce a novel architecture, Mixture-of-Multimodal-Agents (MoMA), designed to leverage multiple large language model (LLM) agents for clinical prediction tasks using multimodal EHR data. MoMA employs specialized LLM agents ("specialist agents") to convert non-textual modalities, such as medical images and laboratory results, into structured textual summaries. These summaries, together with clinical notes, are combined by another LLM ("aggregator agent") to generate a unified multimodal summary, which is then used by a third LLM ("predictor agent") to produce clinical predictions. Evaluating MoMA with different modality combinations and prediction settings, MoMA outperforms existing methods on three prediction tasks using private datasets, highlighting its enhanced accuracy and flexibility across various tasks.
Overview
CoChat Summary
## TL;DR
MoMA is a modular mixture-of-multimodal-agents architecture that converts non-text EHR modalities into text summaries and uses an LLM pipeline for clinical prediction. On three private UW Health tasks, it outperformed baseline methods, with reported gains including macro-F1 near 0.85 and micro-F1 above 0.90 for chest trauma severity, macro-F1 above 0.75 and micro-F1 close to 0.90 for spine trauma in the multitask setting, and AUROC around 0.75 with AUPR near 0.50 for unhealthy alcohol use screening.
## Key findings
- MoMA was evaluated on three private UW Health tasks: chest trauma severity stratification, multitask chest and spine trauma severity stratification, and unhealthy alcohol use screening.
- For chest trauma severity stratification, MoMA achieved macro-F1 scores near 0.85 and micro-F1 scores above 0.90.
- For spine trauma stratification in the multitask setting, MoMA achieved macro-F1 scores above 0.75 and micro-F1 scores close to 0.90.
- For unhealthy alcohol use screening, MoMA achieved AUROC of around 0.75 and AUPR near 0.50; a retrained 1D-CNN on the same small cohort underperformed at AUROC 0.641 and AUPR 0.325.
- In ablation studies, removing the non-text modality reduced performance, indicating the gains were not due only to text understanding but also to use of radiographs or lab measurements.
## Methods
- Design: MoMA uses specialist LLM agents to convert each non-text modality into text, an aggregator LLM to fuse summaries with the original clinical note, and a predictor LLM to produce the final classification; it is framed as a plug-and-play extension of Mixture-of-Experts and Mixture-of-Agents ideas.
- Data/sample: Private UW Health datasets with temporal train/test splits; chest trauma tasks used 2,722 unique patients from January 2015 to December 2019, and unhealthy alcohol use screening used 2,096 consented patients from September 2021 to February 2024.
- Analysis: Compared MoMA against ClinicalBERT, LLaVA-Med, cross-attention fusion, and sparse MoE baselines; evaluated multiclass tasks with macro-/micro-F1 and binary task with AUROC/AUPR; also conducted subgroup and ablation analyses.
## Limitations
- The paper states that agent interactions in MoMA are “relatively simple,” and better communication/coordination among agents may improve performance.
- The authors note inherent LLM issues such as hallucination and omitting necessary signals may still affect classification performance; this is stated in the paper.
- The paper says MoMA may increase inference computational cost because multiple agents must run; this is stated in the paper.
- Evident from the text: the main validation used private datasets, so external researchers cannot directly access the original data.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Figures





Verification
Claims the paper makes
20 Valid · 7 Not valid · 9 Not checked · 97 Own results and interpretations
At least one cited source doesn’t support this claim or points to a different paper, and none confirmed it — see the verdicts below.
In the paper
“This method has consistently demonstrated superior performance compared to early and late fusion strategies across diverse clinical applications14–16.”
§ MoMA: a mixture-of-multimodal-agents architecture for enhancing clinical prediction modelling > Introduction
What the sources say
- Not checked — no full text of the sourcecites [14] Guarrasi, V. et al. A systematic review of intermediate fusion in multimodal deep learning for biomedical applications. Image and Vision Computing 105509 (2025). source
- Disputedcites [15] Stahlschmidt, S. R., Ulfenborg, B. & Synnergren, J. Multimodal deep learning for biomedical data fusion: a review. Brief. Bioinforma. 23, bbab569 (2022). sourceExperimental — accuracy not yet measured
The review’s Discussion on intermediate fusion says that performance advantages are seldom tested and that early fusion performs similarly on some tasks. Its Late fusion section reports a pulmonary-embolism application in which late fusion outperformed intermediate fusion. These findings contradict the claim of consistent superiority across clinical applications. (LLM, conf=0.89)
Checked by a reasoning model — may be wrong.
- Not matchedcites [16] Hayat, N., Geras, K. J. & Shamout, F. E. Medfuse: multi-modal fusion with clinical time-series data and chest x-ray images. In Machine Learning for Healthcare Conference, 479–503 (PMLR, 2022).
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
boostedstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
Some cited sources support this claim and some don’t — read each source’s verdict below.
In the paper
“Machine learning models that integrate task-relevant multimodal EHR have been shown to yield superior performance compared to those relying on a single modality2–6.”
§ MoMA: a mixture-of-multimodal-agents architecture for enhancing clinical prediction modelling > Introduction
What the sources say
- Disputedcites [2] Rohaut, B. et al. Multimodal assessment improves neuroprognosis performance in clinically unresponsive critical-care patients with brain injury. Nat. Med 30.8, 2349–2355 (2024). sourceExperimental — accuracy not yet measured
The Results (‘Multimodal assessment improves neuroprognosis performance’) report that the multivariable classifier trained on multimodal markers performed similarly to individual prognostic markers, not better. The reported accuracy gain with more modalities concerns the specialist team’s prognosis, rather than superior performance by a machine-learning model. (LLM, conf=0.88)
Checked by a reasoning model — may be wrong.
- Supportedcites [3] Soenksen, L. R. et al. Integrated multimodal artificial intelligence framework for healthcare applications. NPJ Digit. Med 5, 149 (2022). source
The Results section (‘Demonstration of HAIM framework on multimodal clinical dataset’; page unavailable) reports that models combining EHR sources and modalities improved predictive performance over single-source and single-modality approaches across the evaluated tasks. The finding is based on this paper’s clinical dataset and tasks, rather than a claim that every multimodal model will outperform every single-modality model. (LLM, conf=0.93)
- Not checked — no full text of the sourcecites [4] Winston, C. et al. Multimodal clinical prediction with unified prompts and pretrained large-language models. In 2024 IEEE 12th International Conference on Healthcare Informatics (ICHI), 679–683 (IEEE, 2024). source
- Partly supportedcites [5] Gao, J. et al. Automated stratification of trauma injury severity across multiple body regions using multi-modal, multi-class machine learning models. J. Am. Med. Inform. Assoc. 31, 1291–1302 (2024). sourceExperimental — accuracy not yet measured
Section 3.3 (page not provided) reports that adding structured EHR data significantly improved performance for arm and extremity injuries. It did not produce a significant improvement in the other three body regions, so this paper does not support the claim of generally superior multimodal performance. (LLM, conf=0.89)
- Partly supportedcites [6] Kline, A. et al. Multimodal machine learning in precision health: a scoping review. NPJ Digit. Med 5, 171 (2022). sourceExperimental — accuracy not yet measured
The Abstract reports that the few studies comparing multimodal with unimodal predictions achieved an average 6.4% increase in predictive accuracy. The Results (Intermediate fusion) describe models combining EHR with other data, but the reported performance gain is for multimodal studies generally, not specifically for task-relevant multimodal EHR models. (LLM, conf=0.82)
boostedstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Disputed
The sources with a verdict support this claim; the rest weren’t checked yet, have no full text we could read, or their verdicts aren’t rated yet.
In the paper
“LLMs are widely recognized for their capability to capture semantic meaning from text28,29, including clinical notes30.”
§ MoMA: a mixture-of-multimodal-agents architecture for enhancing clinical prediction modelling > Introduction
What the sources say
- Supportedcites [28] Zhao, W. X. et al. A survey of large language models. arXiv preprint arXiv:2303.18223 (2023). source
The survey’s “Basic ability” section (p. 23) identifies natural-language understanding as a foundational LLM capability, supporting the claim that LLMs can capture meaning from text. In the citing sentence, the clinical-notes example has a separate citation (30), so that specific example is not attributed to this source. (LLM, conf=0.88)
- Supportedcites [29] Wei, J. et al. Emergent abilities of large language models. Trans. Mach. Learn. Res 2022, 1–30 (2022). source
The citation to this paper supports the general claim about meaning in text; the clinical-notes clause has a separate citation. Appendix A.3 (p. 22) reports emergent performance on word-sense disambiguation, analogical reasoning, and emotional-understanding tasks, supporting that general claim. (LLM, conf=0.86)
- Not checked — no full text of the sourcecites [30] Thirunavukarasu, A. J. et al. Large language models in medicine. Nat. Med. 29, 1930–1940 (2023). source
boostedstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
The cited sources that could be checked couldn’t confirm or dispute this claim.
In the paper
“However, they have yet to demonstrate whether such architectures can effectively leverage their text understanding abilities to improve prediction performance in classification tasks involving multimodal EHR data, which is a critical need in clinical applications.”
§ MoMA: a mixture-of-multimodal-agents architecture for enhancing clinical prediction modelling > Discussion
What the sources say
- Not matchedcites [62] Jiang, B. et al. Multi-modal and multi-agent systems meet rationality: A survey. In ICML 2024 Workshop on LLMs and Cognition (2024).
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- Not matchedcites [63] Du, Y., Li, S., Torralba, A., Tenenbaum, J. B. & Mordatch, I. Improving factuality and reasoning in language models through multiagent debate. In International Conference on Machine Learning (2023).
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- Could not verifycites [64] Tang, X. et al. Medagents: Large language models as collaborators for zero-shot medical reasoning. In Findings of the Association for Computational Linguistics: ACL 2024, pp. 599–621 (2024). sourceExperimental — accuracy not yet measured
The Introduction and §3.2 describe MedAgents’ evaluation on medical question-answering datasets, not multimodal EHR classification. The Limitation section does not discuss that application either. Our copy therefore shows what this framework tested, but does not establish the broader claim that multi-agent architectures have yet to demonstrate improved prediction through text understanding on multimodal EHR data (searched: multimodal EHR classification and electronic health records). (LLM, conf=0.72)
- Not matchedcites [65] Li, J., Zhang, Q., Yu, Y., Fu, Q. & Ye, D. More agents is all you need. Trans. Mach. Learn. Res. (2024).
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Not matched
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“we also demonstrated MoMA using publicly available datasets (MIMIC-IV69 and MIMIC-CXR-JPG70), including admission notes and chest radiograph images, to predict in-hospital mortality in patients admitted to the Trauma Surgical Intensive Care Unit (TSICU).”
§ MoMA: a mixture-of-multimodal-agents architecture for enhancing clinical prediction modelling > Discussion
What the sources say
- Supportedcites [69] Johnson, A. E. et al. Mimic-iv, a freely accessible electronic health record dataset. Sci. Data 10, 1 (2023). source
The Data Records section identifies MIMIC-IV as a dataset accessible through PhysioNet, confirming the dataset citation. The MoMA demonstration and mortality prediction are the citing authors’ own work, not findings attributed to this source. (LLM, conf=0.93)
- Partly supportedcites [70] Johnson, A. E. et al. Mimic-cxr-jpg, a large publicly available database of labeled chest radiographs. arXiv preprint arXiv:1901.07042 (2019). sourceExperimental — accuracy not yet measured
The Background & Summary identifies MIMIC-CXR as a publicly available chest-radiograph dataset, supporting the dataset identification in part. But Methods > Chest radiographs says its images are distributed in DICOM format; our copy does not identify the cited dataset as MIMIC-CXR-JPG. This citation does not establish the citing authors’ MoMA demonstration or TSICU mortality prediction. (LLM, conf=0.84)
boosted- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
42 more claims · 4 with their cited sources checked
Create a free account to see all 133 claims and every source verdict.
- It may state two separate claims as one. 40
- Probably not a substantial claim. 64
Citations
Cites
Huang, L., Yu, W., Ma, W., Zhong, W., Feng, Z., Wang, H., Chen, Q., Peng, W., Feng, X., Qin, B., et al. (2025) · ACM Transactions on Information Systems · doi:10.1145/3703155 · Cited by 4,688
Large language models in medicine
Thirunavukarasu, A. J., Ting, D. S. J., Elangovan, K., Gutierrez, L., Tan, T. F., & Ting, D. S. W. (2023) · Nature Medicine · doi:10.1038/s41591-023-02448-8 · Cited by 3,944 · cited 1×
“LLMsarewidelyrecognized fortheircapabilitytocapturesemanticmeaningfromtext 28,29 , including clinicalnotes 30 .”
MIMIC-IV, a freely accessible electronic health record dataset
Johnson, A. E. W., Bulgarelli, L., Shen, L., Gayles, A., Shammout, A., Horng, S., Pollard, T. J., Hao, S., Moody, B., Gow, B., et al. (2023) · Scientific Data · doi:10.1038/s41597-022-01899-x · Cited by 3,401 · cited 1×
“…summary overwrote textual information (e.g., “ The specialist reports right basal atelectasis, small right pleural effusion, and cardiomegaly, whereas the Aggregator — guided by the original clinical and radiology notes — states no effusion or atelectasis and does not mention cardiomegaly ” ).”
Liu, H., Li, C., Wu, Q., & Lee, Y. J. (2023) · Advances in Neural Information Processing Systems 36 · doi:10.52202/075280-1516 · Cited by 2,428
A Survey of Large Language Models
Zhao, W. X., Zhou, K., Li, J., Tang, T., Dong, Z., Hou, Y., Zhang, B., Min, Y., Zhang, J., Liu, P., et al. (2026) · Frontiers of Computer Science · doi:10.1007/s11704-026-60308-3 · Cited by 2,093
A Survey of Large Language Models
Zhao, W. X., Zhou, K., Li, J., Tang, T., Wang, X., Hou, Y., Min, Y., Zhang, B., Zhang, J., Dong, Z., et al. (2023) · arXiv · doi:10.48550/arxiv.2303.18223 · Cited by 2,090
Savova, G. K., Masanz, J. J., Ogren, P. V., Zheng, J., Sohn, S., Kipper-Schuler, K. C., & Chute, C. G. (2010) · Journal of the American Medical Informatics Association · doi:10.1136/jamia.2009.001560 · Cited by 1,967 · cited 1×
“In the work by Afshar et al. 50 , clinical concepts, such as diseases, symptoms, anatomical sites, medications, and procedures, were extracted fromallavailablenoteswithineachpatientencounterusingtheNLPengine, clinicalTextAndKnowledgeExtractionSystem(cTAKES) 86 .”
MIMIC-CXR, a de-identified publicly available database of chest radiographs with free-text reports
Johnson, A. E. W., Pollard, T. J., Berkowitz, S. J., Greenbaum, N. R., Lungren, M. P., Deng, C.-Y., Mark, R. G., & Horng, S. (2019) · Scientific Data · doi:10.1038/s41597-019-0322-0 · Cited by 1,816
Alsentzer, E., Murphy, J., Boag, W., Weng, W.-H., Jindi, D., Naumann, T., & McDermott, M. (2019) · Proceedings of the 2nd Clinical Natural Language Processing Workshop · doi:10.18653/v1/w19-1909 · Cited by 1,736 · cited 3×
“We validated the MoMA architecture on three clinical prediction tasks using private datasets collected from the University of Wisconsin Hospitals and Clinics (UW Health): chest trauma severity strati fi cation, multitask chest and spine trauma severity strati fi cation, and unhealthy alcohol use screening.”
InfluentialAcosta, J. N., Falcone, G. J., Rajpurkar, P., & Topol, E. J. (2022) · Nature Medicine · doi:10.1038/s41591-022-01981-2 · Cited by 1,365 · cited 1×
“Published in partnership with Seoul National University Bundang Hospital https://doi.org/10.1038/s41746-025-02219-4 MoMA: a mixture-of-multimodal-agents architecture for enhancing clinical prediction modelling”
Influential
Research works printed but not linked: 13
- Li, J., Li, D., Xiong, C. & Hoi, S. Blip: Bootstrapping language-image pre-training for unified vision-language understanding and generation. In International conference on machine learning, 12888–12900 (PMLR, 2022).
- Radford, A. et al. Learning transferable visual models from natural language supervision. In International Conference on Machine Learning, 8748–8763 (2021).
- Hayat, N., Geras, K. J. & Shamout, F. E. Medfuse: multi-modal fusion with clinical time-series data and chest x-ray images. In Machine Learning for Healthcare Conference, 479–503 (PMLR, 2022).
- Li, J., Li, D., Savarese, S. & Hoi, S. Blip-2: Bootstrapping language-image pre-training with frozen image encoders and large language models. In International conference on machine learning, 19730–19742 (PMLR, 2023).
- Peng, Z. et al. Kosmos-2: Grounding multimodal large language models to the world. In International Conference on Learning Representations (2024).
- Driess, D. et al. Palm-e: an embodied multimodal language model. In International Conference on Machine Learning (2023).
- Wang, J., Wang, J., Athiwaratkun, B., Zhang, C. & Zou, J. Mixture-of-agents enhances large language model capabilities. In International Conference on Learning Representations (2025).
- Jiang, B. et al. Multi-modal and multi-agent systems meet rationality: A survey. In ICML 2024 Workshop on LLMs and Cognition (2024).
- Du, Y., Li, S., Torralba, A., Tenenbaum, J. B. & Mordatch, I. Improving factuality and reasoning in language models through multiagent debate. In International Conference on Machine Learning (2023).
- Li, J., Zhang, Q., Yu, Y., Fu, Q. & Ye, D. More agents is all you need. Trans. Mach. Learn. Res. (2024).
- Ben Abacha, A., Hasan, S. A., Datla, V. V., Demner-Fushman, D. & Müller, H. Vqa-med: Overview of the medical visual question answering task at imageclef 2019. In Proceedings of CLEF (Conference and Labs of the Evaluation Forum) 2019 Working Notes (9-12 September 2019, 2019).
- Shazeer, N. et al. Outrageously large neural networks: The sparsely-gated mixture-of-experts layer. In International Conference on Learning Representations (2017)
- Xia, P. et al. Rule: Reliable multimodal rag for factuality in medical vision language models. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, 1081–1093 (2024). · DOI · Paper page
Cited by
Mwangi, B., Hamoudi, H. J. A. S., Sanches, M., et al. (2026) · medRxiv · doi:10.64898/2026.04.17.26351105 · Cited by 1
AdaMoA: Enhancing Mixture-of-Agents via Task-Adaptive Architecture Optimization
Lv, W., Xu, H., Xu, Y., Liu, J., & Xu, Z. (2026) · International Workshop on Quality of Service · doi:10.1109/iwqos70441.2026.11661164
“This method has demonstrated its effectiveness across a broad range of domains such as clinical decision in healthcare [16], [17] and code generation in software engineering [18], [19].”
Metadata
- Authors
- Gao, J., Rahman, M., Caskey, J., Oguss, M., O’Rourke, A., Brown, R., Stey, A., Mayampurath, A., Churpek, M. M., Chen, G., & Afshar, M.
- Affiliations
- University of Wisconsin–Madison (US)
- Northwestern University (US)
- Corresponding author
- Chen, Guanhua
- Afshar, Majid
- ORCID
- Venue
- npj Digital Medicine
- Published
- 2025-12-09
- Volume
- 9
- Issue
- 1
- Pages
- 46-46
- ISSN
- 2398-6352 (ISSN-L)
- PMC
- PMC12804996
- OpenAlex
- W4417156540
- Cited by
- 2
- References
- 52
- Citations by year
- 2026: 2
- FWCI
- 0.80
- Type
- journal article
- Language
- en
- Open access
- gold
- doi.org
- Licence
- cc-by-nc-nd
- creativecommons.org/licenses/by-nc-nd/4.0 (tdm)
- creativecommons.org/licenses/by-nc-nd/4.0 (vor, 36 d)
- Locations
- npj Digital Medicine (publishedVersion, cc-by-nc-nd)
- PubMed (publishedVersion)
- PubMed Central (publishedVersion, cc-by-nc-nd)
- arXiv (Cornell University) (submittedVersion)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
- arXiv (Cornell University) (cc-by)
- Funders
- National Science Foundation, United States (DMS-2054346)
- National Institutes of Health (R01DA051464)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Machine Learning in Healthcare
- Health Sciences › Medicine › Health Informatics › Artificial Intelligence in Healthcare and Education
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Multimodal Machine Learning Applications
- Keywords
- mixture-of-agents, electronic health records, multimodal data integration, clinical notes, large language models, multimodal summarization
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-10-09, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-09 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: MoMA: a mixture-of-multimodal-agents architecture for enhancing clinical prediction modelling. https://app.cochat.ai/papers/10.1038/s41746-025-02219-4
- The paper
- Gao, J., Rahman, M., Caskey, J., Oguss, M., O’Rourke, A., Brown, R., Stey, A., Mayampurath, A., Churpek, M. M., Chen, G., & Afshar, M. (2025). MoMA: a mixture-of-multimodal-agents architecture for enhancing clinical prediction modelling. npj Digital Medicine. https://doi.org/10.1038/s41746-025-02219-4
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.