{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"institution":[{"name":"medRxiv"}],"indexed":{"date-parts":[[2026,9,24]],"date-time":"2026-09-24T16:04:36Z","timestamp":1790265876430,"version":"4.1.0"},"posted":{"date-parts":[[2026,9,21]]},"group-title":"Health Informatics","reference-count":68,"publisher":"openRxiv","license":[{"start":{"date-parts":[[2026,9,21]],"date-time":"2026-09-21T00:00:00Z","timestamp":1789948800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/www.medrxiv.org\/about\/FAQ#license"}],"funder":[{"name":"National Human Genome Research Institute","award":["HG011052"],"award-info":[{"award-number":["HG011052"]}],"id":[{"id":"https:\/\/ror.org\/00baak391","id-type":"ROR","asserted-by":"crossref"}]},{"name":"National Human Genome Research Institute","award":["HG011052-03S1"],"award-info":[{"award-number":["HG011052-03S1"]}],"id":[{"id":"https:\/\/ror.org\/00baak391","id-type":"ROR","asserted-by":"crossref"}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"accepted":{"date-parts":[[2026,9,21]]},"abstract":"<jats:title>Abstract<\/jats:title>\n                <jats:p>\n                  The truthful use of large language models (LLMs) is a growing challenge in safeguarding sensitive patient data from leakage. We introduce and evaluate a privacy-preserving knowledge distillation framework for LLM-based clinical modeling, using multimorbidity scoring as a healthcare task. Although LLMs can encode rich clinical knowledge and improve upon traditional rule-based comorbidity scoring, their direct evaluation on large-scale biobank data remains constrained by patient privacy. In our framework, multimorbidity reasoning is distilled from state-of-the-art LLM teacher models into compact student models (CoLLMs) using synthetic cohorts that preserve UK Biobank distributions, without exposing real patient data. This approach achieves high-fidelity knowledge transfer (Spearman \u03c1 = 0.75\u20130.89). Independent LLM-as-Judge evaluation confirms the clinical significance of the distilled knowledge and reveals substantial variability among teacher models. When applied to real UK Biobank data, CoLLM-derived multimorbidity scores improve survival prediction (C-index up to 0.91) and exhibit higher SNP heritability (h\n                  <jats:sup>2<\/jats:sup>\n                  \u2248 0.05). Our work establishes a trustworthy, privacy-compliant pathway for large-scale healthcare applications of LLMs.\n                <\/jats:p>","DOI":"10.64898\/2026.09.19.26363476","type":"posted-content","created":{"date-parts":[[2026,9,22]],"date-time":"2026-09-22T01:10:45Z","timestamp":1790039445000},"source":"Crossref","is-referenced-by-count":0,"title":["Privacy-Preserving Distilled Large Language Models Enhance Multimorbidity Scoring"],"prefix":"10.64898","author":[{"given":"Raghav","family":"Awasthi","sequence":"first","affiliation":[{"name":"Department of Population and Quantitative Health Sciences, Case Western Reserve University School of Medicine, Cleveland, OH, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Yihe","family":"Yang","sequence":"additional","affiliation":[{"name":"Department of Population and Quantitative Health Sciences, Case Western Reserve University School of Medicine, Cleveland, OH, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Mengxuan","family":"Li","sequence":"additional","affiliation":[{"name":"Department of Population and Quantitative Health Sciences, Case Western Reserve University School of Medicine, Cleveland, OH, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Xiaofeng","family":"Zhu","sequence":"additional","affiliation":[{"name":"Department of Population and Quantitative Health Sciences, Case Western Reserve University School of Medicine, Cleveland, OH, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"54368","reference":[{"key":"2026092408150623000_2026.09.19.26363476v1.1","doi-asserted-by":"crossref","first-page":"101860","DOI":"10.1016\/j.eclinm.2023.101860","article-title":"Global and regional prevalence of multimorbidity in the adult population in community settings: a systematic review and meta-analysis","volume":"57","year":"2023","journal-title":"EClinicalMedicine"},{"key":"2026092408150623000_2026.09.19.26363476v1.2","first-page":"1","article-title":"High-risk multimorbidity patterns on the road to cardiovascular mortality","volume":"18","year":"2020","journal-title":"BMC Medicine"},{"key":"2026092408150623000_2026.09.19.26363476v1.3","doi-asserted-by":"crossref","first-page":"48","DOI":"10.1038\/s41572-022-00376-4","article-title":"Multimorbidity","volume":"8","year":"2022","journal-title":"Nature reviews. Disease primers"},{"key":"2026092408150623000_2026.09.19.26363476v1.4","doi-asserted-by":"publisher","DOI":"10.1093\/eurpub\/cky098"},{"key":"2026092408150623000_2026.09.19.26363476v1.5","doi-asserted-by":"publisher","DOI":"10.1370\/afm.983"},{"key":"2026092408150623000_2026.09.19.26363476v1.6","doi-asserted-by":"crossref","first-page":"407","DOI":"10.1186\/s12871-025-03246-0","article-title":"Associations of multimorbidity with mortality, hospital stay, and hospitalization costs in Chinese surgical patients: a retrospective cohort study","volume":"25","year":"2025","journal-title":"BMC Anesthesiology"},{"key":"2026092408150623000_2026.09.19.26363476v1.7","first-page":"1","article-title":"Unraveling cradle-to-grave disease trajectories from multilayer comorbidity networks","volume":"7","year":"2024","journal-title":"npj Digital Medicine"},{"key":"2026092408150623000_2026.09.19.26363476v1.8","doi-asserted-by":"crossref","first-page":"957","DOI":"10.3390\/v17070957","article-title":"The Role of Comorbidities in COVID-19 Severity","volume":"17","year":"2025","journal-title":"Viruses"},{"key":"2026092408150623000_2026.09.19.26363476v1.9","doi-asserted-by":"crossref","unstructured":"Nigatu, B. Z. & Dessie, N. T . Prevalence of comorbidities and their association with disease severity and mortality in COVID-19 patients: A systematic review and meta-analysis. Journal of multimorbidity and comorbidity 15, (2025).","DOI":"10.1177\/26335565251371256"},{"key":"2026092408150623000_2026.09.19.26363476v1.10","doi-asserted-by":"crossref","first-page":"e0329701","DOI":"10.1371\/journal.pone.0329701","article-title":"Comorbidity patterns associated with severe COVID-19 outcomes: A cohort study based on the UK Biobank","volume":"20","year":"2025","journal-title":"PLOS ONE"},{"key":"2026092408150623000_2026.09.19.26363476v1.11","doi-asserted-by":"crossref","unstructured":"Rashid, M. et al. Impact of co-morbid burden on mortality in patients with coronary heart disease, heart failure, and cerebrovascular accident: a systematic review and meta-analysis. European heart journal. Quality of care & clinical outcomes 3, (2017).","DOI":"10.1093\/ehjqcco\/qcw025"},{"key":"2026092408150623000_2026.09.19.26363476v1.12","doi-asserted-by":"crossref","first-page":"1215458","DOI":"10.3389\/fcvm.2024.1215458","article-title":"Comorbidity patterns in cardiovascular diseases: the role of life-stage and socioeconomic status","volume":"11","year":"2024","journal-title":"Front. Cardiovasc. Med"},{"key":"2026092408150623000_2026.09.19.26363476v1.13","doi-asserted-by":"crossref","unstructured":"S\u00f8gaard, M. , Thomsen, R. W. , Bossen, K. S. , S\u00f8rensen, H. T. & N\u00f8rgaard, M . The impact of comorbidity on cancer survival: a review. Clinical epidemiology 5, (2013).","DOI":"10.2147\/CLEP.S47150"},{"key":"2026092408150623000_2026.09.19.26363476v1.14","doi-asserted-by":"crossref","first-page":"e26425","DOI":"10.2196\/26425","article-title":"Physical Comorbidities and Their Relationship with Cancer Treatment and Its Outcomes in Older Adult Populations: Systematic Review","volume":"7","year":"2021","journal-title":"JMIR Cancer"},{"key":"2026092408150623000_2026.09.19.26363476v1.15","doi-asserted-by":"crossref","first-page":"111869","DOI":"10.1016\/j.jclinepi.2025.111869","article-title":"Comparison of established comorbidity scores using administrative data of patients undergoing surgery or interventional procedures in Massachusetts","volume":"185","year":"2025","journal-title":"Journal of Clinical Epidemiology"},{"key":"2026092408150623000_2026.09.19.26363476v1.16","doi-asserted-by":"crossref","unstructured":"Sharma, N. , Schwendimann, R. , Endrich, O. , Ausserhofer, D. & Simon, M . Comparing Charlson and Elixhauser comorbidity indices with different weightings to predict in-hospital mortality: an analysis of national inpatient data. BMC health services research 21, (2021).","DOI":"10.1186\/s12913-020-05999-5"},{"key":"2026092408150623000_2026.09.19.26363476v1.17","doi-asserted-by":"publisher","DOI":"10.1016\/j.wneu.2020.08.138"},{"key":"2026092408150623000_2026.09.19.26363476v1.18","unstructured":"Br\u00e4mer, G. R. International statistical classification of diseases and related health problems. Tenth revision. World health statistics quarterly. Rapport trimestriel de statistiques sanitaires mondiales 41, (1988)."},{"key":"2026092408150623000_2026.09.19.26363476v1.19","doi-asserted-by":"crossref","first-page":"e767","DOI":"10.1016\/S2589-7500(24)00151-1","article-title":"The potential for large language models to transform cardiovascular medicine","volume":"6","year":"2024","journal-title":"Lancet Digit Health"},{"key":"2026092408150623000_2026.09.19.26363476v1.20","doi-asserted-by":"crossref","first-page":"100868","DOI":"10.1016\/j.bj.2025.100868","article-title":"Roles and potential of Large language models in healthcare: A comprehensive review","volume":"48","year":"2025","journal-title":"Biomedical Journal"},{"key":"2026092408150623000_2026.09.19.26363476v1.21","doi-asserted-by":"publisher","DOI":"10.1038\/s41586-023-06291-2"},{"key":"2026092408150623000_2026.09.19.26363476v1.22","doi-asserted-by":"crossref","first-page":"1418","DOI":"10.1038\/s42256-024-00944-1","article-title":"LLM-based agentic systems in medicine and healthcare","volume":"6","year":"2024","journal-title":"Nature Machine Intelligence"},{"key":"2026092408150623000_2026.09.19.26363476v1.23","doi-asserted-by":"crossref","first-page":"e0000198","DOI":"10.1371\/journal.pdig.0000198","article-title":"Performance of ChatGPT on USMLE: Potential for AI-assisted medical education using large language models","volume":"2","year":"2023","journal-title":"PLOS Digital Health"},{"key":"2026092408150623000_2026.09.19.26363476v1.24","doi-asserted-by":"crossref","first-page":"263","DOI":"10.1038\/s41746-025-01684-1","article-title":"Evaluating large language model workflows in clinical decision support for triage and referral and diagnosis","volume":"8","year":"2025","journal-title":"NPJ Digital Medicine"},{"key":"2026092408150623000_2026.09.19.26363476v1.25","doi-asserted-by":"publisher","DOI":"10.1038\/s41591-023-02344-7"},{"key":"2026092408150623000_2026.09.19.26363476v1.26","doi-asserted-by":"crossref","first-page":"120","DOI":"10.1038\/s41746-023-00873-0","article-title":"The imperative for regulatory oversight of large language models (or generative AI) in healthcare","volume":"6","year":"2023","journal-title":"NPJ Digit Med"},{"key":"2026092408150623000_2026.09.19.26363476v1.27","doi-asserted-by":"crossref","first-page":"257","DOI":"10.1038\/s41746-024-01233-2","article-title":"Privacy-preserving large language models for structured medical information retrieval","volume":"7","year":"2024","journal-title":"npj Digital Medicine"},{"key":"2026092408150623000_2026.09.19.26363476v1.28","doi-asserted-by":"crossref","first-page":"e76571","DOI":"10.2196\/76571","article-title":"Considerations for Patient Privacy of Large Language Models in Health Care: Scoping Review","volume":"27","year":"2025","journal-title":"Journal of Medical Internet Research"},{"key":"2026092408150623000_2026.09.19.26363476v1.29","unstructured":"Website. UK Biobank. (2023, October 26). Material transfer agreement: New applicant MTA \u2013 Data only (ACCESS_031_F v1.4). UK Biobank Limited. https:\/\/www.ukbiobank.ac.uk\/wp-content\/uploads\/2025\/01\/Material-Transfer-Agreement.pdf."},{"key":"2026092408150623000_2026.09.19.26363476v1.30","doi-asserted-by":"crossref","unstructured":"Mo\u00ebll, B. , Sand Aronsson, F. & Akbar, S . Medical reasoning in LLMs: an in-depth analysis of DeepSeek R1. Front Artif Intell 8, 1616145 (2025).","DOI":"10.3389\/frai.2025.1616145"},{"key":"2026092408150623000_2026.09.19.26363476v1.31","doi-asserted-by":"crossref","first-page":"e2526021","DOI":"10.1001\/jamanetworkopen.2025.26021","article-title":"Fidelity of Medical Reasoning in Large Language Models","volume":"8","year":"2025","journal-title":"JAMA Netw Open"},{"key":"2026092408150623000_2026.09.19.26363476v1.32","doi-asserted-by":"crossref","unstructured":"Shojaee, P. , et al. The Illusion of Thinking: Understanding the Strengths and Limitations of Reasoning Models via the Lens of Problem Complexity. (2025).","DOI":"10.70777\/si.v2i6.15919"},{"key":"2026092408150623000_2026.09.19.26363476v1.33","unstructured":"Mirzadeh, I. , et al. GSM-Symbolic: Understanding the Limitations of Mathematical Reasoning in Large Language Models. (2024)."},{"key":"2026092408150623000_2026.09.19.26363476v1.34","unstructured":"Hinton, G. , Vinyals, O. & Dean, J. Distilling the Knowledge in a Neural Network. (2015)."},{"key":"2026092408150623000_2026.09.19.26363476v1.35","unstructured":"Iyer, S. When Babies Teach Babies: Can student knowledge sharing outperform Teacher-Guided Distillation on small datasets? in The 2nd BabyLM Challenge at the 28th Conference on Computational Natural Language Learning 197\u2013211 (2024)."},{"key":"2026092408150623000_2026.09.19.26363476v1.36","unstructured":"Can Students Beyond The Teacher? Distilling Knowledge from Teacher\u2019s Bias. https:\/\/arxiv.org\/html\/2412.09874v1."},{"key":"2026092408150623000_2026.09.19.26363476v1.37","unstructured":"Xu, L. , Skoularidou, M. , Cuesta-Infante, A. & Veeramachaneni, K . Modeling Tabular data using Conditional GAN. Advances in Neural Information Processing Systems 32, (2019)."},{"key":"2026092408150623000_2026.09.19.26363476v1.38","unstructured":"OpenAI et al. GPT-4 Technical Report. (2023)."},{"key":"2026092408150623000_2026.09.19.26363476v1.39","unstructured":"Anil, R. et al. Gemini: A Family of Highly Capable Multimodal Models. (2023)."},{"key":"2026092408150623000_2026.09.19.26363476v1.40","doi-asserted-by":"publisher","DOI":"10.1038\/s41586-025-09422-z"},{"key":"2026092408150623000_2026.09.19.26363476v1.41","doi-asserted-by":"publisher","DOI":"10.1109\/dsaa.2016.49"},{"key":"2026092408150623000_2026.09.19.26363476v1.42","doi-asserted-by":"crossref","unstructured":"Bi, W. et al. Efficient mixed model approach for large-scale genome-wide association studies of ordinal categorical phenotypes. American journal of human genetics 108, (2021).","DOI":"10.1101\/2020.10.09.333146"},{"key":"2026092408150623000_2026.09.19.26363476v1.43","unstructured":"Genome-Wide Robust Analysis for Biobank Data (GRAB) [R package GRAB version 0.2.3]. (2025)."},{"key":"2026092408150623000_2026.09.19.26363476v1.44","doi-asserted-by":"publisher","DOI":"10.2174\/138920207783591690"},{"key":"2026092408150623000_2026.09.19.26363476v1.45","doi-asserted-by":"publisher","DOI":"10.1038\/sj.gene.6364061"},{"key":"2026092408150623000_2026.09.19.26363476v1.46","doi-asserted-by":"publisher","DOI":"10.1056\/NEJMoa0902604"},{"key":"2026092408150623000_2026.09.19.26363476v1.47","doi-asserted-by":"publisher","DOI":"10.1016\/j.jacasi.2025.04.012"},{"key":"2026092408150623000_2026.09.19.26363476v1.48","doi-asserted-by":"crossref","unstructured":"U\u011furlu-\u00c7imen, D. et al. AF10 (MLLT10) prevents somatic cell reprogramming through regulation of DOT1L-mediated H3K79 methylation. Epigenetics & chromatin 14, (2021).","DOI":"10.1186\/s13072-021-00406-7"},{"key":"2026092408150623000_2026.09.19.26363476v1.49","doi-asserted-by":"crossref","unstructured":"Deutsch, J. L. & Heath, J. L . MLLT10 in benign and malignant hematopoiesis. Experimental hematology 87, (2020).","DOI":"10.1016\/j.exphem.2020.06.002"},{"key":"2026092408150623000_2026.09.19.26363476v1.50","doi-asserted-by":"crossref","unstructured":"Jing, X. et al. MLLT10 promotes tumor migration, invasion, and metastasis in human colorectal cancer. Scandinavian journal of gastroenterology 53, (2018).","DOI":"10.1080\/00365521.2018.1481521"},{"key":"2026092408150623000_2026.09.19.26363476v1.51","doi-asserted-by":"publisher","DOI":"10.1016\/0895-4356(95)00564-1"},{"key":"2026092408150623000_2026.09.19.26363476v1.52","doi-asserted-by":"publisher","DOI":"10.1097\/MLR.0b013e31819432e5"},{"key":"2026092408150623000_2026.09.19.26363476v1.53","first-page":"1","article-title":"Reasoning-based LLMs surpass average human performance on medical social skills","volume":"15","year":"2025","journal-title":"Scientific Reports"},{"key":"2026092408150623000_2026.09.19.26363476v1.54","doi-asserted-by":"crossref","first-page":"305","DOI":"10.1038\/s41562-024-02046-9","article-title":"Large language models surpass human experts in predicting neuroscience results","volume":"9","year":"2024","journal-title":"Nature Human Behaviour"},{"key":"2026092408150623000_2026.09.19.26363476v1.55","doi-asserted-by":"crossref","first-page":"104675","DOI":"10.1016\/j.jml.2025.104675","article-title":"LLMs model how humans induce logically structured rules","volume":"146","year":"2026","journal-title":"Journal of Memory and Language"},{"key":"2026092408150623000_2026.09.19.26363476v1.56","doi-asserted-by":"crossref","first-page":"1677","DOI":"10.1038\/s41592-025-02748-6","article-title":"GeneAgent: self-verification language agent for gene-set analysis using domain databases","volume":"22","year":"2025","journal-title":"Nature Methods"},{"key":"2026092408150623000_2026.09.19.26363476v1.57","first-page":"22","article-title":"Enhancing functional gene set analysis with large language models","volume":"22","year":"2024","journal-title":"Nature Methods"},{"key":"2026092408150623000_2026.09.19.26363476v1.58","first-page":"vbaf054","article-title":"Enhancing gene set overrepresentation analysis with large language models","volume":"5","year":"2025","journal-title":"Bioinform Adv"},{"key":"2026092408150623000_2026.09.19.26363476v1.59","doi-asserted-by":"crossref","first-page":"e1001779","DOI":"10.1371\/journal.pmed.1001779","article-title":"UK Biobank: An Open Access Resource for Identifying the Causes of a Wide Range of Complex Diseases of Middle and Old Age","volume":"12","year":"2015","journal-title":"PLOS Medicine"},{"key":"2026092408150623000_2026.09.19.26363476v1.60","first-page":"6","article-title":"The Effectiveness of ICD-10-CM in Capturing Public Health Diseases","volume":"4","year":"2007","journal-title":"Perspectives in Health Information Management \/ AHIMA, American Health Information Management Association"},{"key":"2026092408150623000_2026.09.19.26363476v1.61","doi-asserted-by":"crossref","unstructured":"Quan, H. et al. Coding algorithms for defining comorbidities in ICD-9-CM and ICD-10 administrative data. Medical care 43, (2005).","DOI":"10.1097\/01.mlr.0000182534.19832.83"},{"key":"2026092408150623000_2026.09.19.26363476v1.62","unstructured":"Cui, E. H. , Li, Y. & Liu, Z. The Kolmogorov-Smirnov Statistic Revisited. (2025)."},{"key":"2026092408150623000_2026.09.19.26363476v1.63","doi-asserted-by":"crossref","first-page":"640","DOI":"10.1038\/s41746-025-02005-2","article-title":"Evaluating clinical AI summaries with large language models as judges","volume":"8","year":"2025","journal-title":"npj Digital Medicine"},{"key":"2026092408150623000_2026.09.19.26363476v1.64","doi-asserted-by":"publisher","DOI":"10.1101\/2025.10.27.25338910"},{"key":"2026092408150623000_2026.09.19.26363476v1.65","unstructured":"Li, H. , et al. LLMs-as-Judges: A Comprehensive Survey on LLM-based Evaluation Methods. (2024)."},{"key":"2026092408150623000_2026.09.19.26363476v1.66","doi-asserted-by":"publisher","DOI":"10.1007\/s12110-014-9209-0"},{"key":"2026092408150623000_2026.09.19.26363476v1.67","unstructured":"Website. https:\/\/www.biorxiv.org\/content\/10.1101\/518027v4."},{"key":"2026092408150623000_2026.09.19.26363476v1.68","doi-asserted-by":"crossref","unstructured":"Cuellar-Partida, Gabriel , et al. \"Complex-Traits Genetics Virtual Lab: A community-driven web platform for post-GWAS analyses.\" BioRxiv (2019): 518027.","DOI":"10.1101\/518027"}],"container-title":[],"original-title":[],"link":[{"URL":"https:\/\/syndication.highwire.org\/content\/doi\/10.64898\/2026.09.19.26363476","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,9,24]],"date-time":"2026-09-24T15:17:52Z","timestamp":1790263072000},"score":1,"resource":{"primary":{"URL":"http:\/\/medrxiv.org\/lookup\/doi\/10.64898\/2026.09.19.26363476"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,9,21]]},"references-count":68,"URL":"https:\/\/doi.org\/10.64898\/2026.09.19.26363476","relation":{},"subject":[],"published":{"date-parts":[[2026,9,21]]},"subtype":"preprint"}}