{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,7,8]],"date-time":"2026-07-08T18:45:40Z","timestamp":1783536340779,"version":"3.55.0"},"reference-count":41,"publisher":"Public Library of Science (PLoS)","issue":"7","license":[{"start":{"date-parts":[[2026,7,8]],"date-time":"2026-07-08T00:00:00Z","timestamp":1783468800000},"content-version":"vor","delay-in-days":0,"URL":"http:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"DOI":"10.13039\/501100002347","name":"Bundesministerium f\u00fcr Bildung und Forschung","doi-asserted-by":"publisher","award":["01KD2420C"],"award-info":[{"award-number":["01KD2420C"]}],"id":[{"id":"10.13039\/501100002347","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/501100001659","name":"Deutsche Forschungsgemeinschaft","doi-asserted-by":"publisher","award":["FU 356\/12-2"],"award-info":[{"award-number":["FU 356\/12-2"]}],"id":[{"id":"10.13039\/501100001659","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/501100001659","name":"Deutsche Forschungsgemeinschaft","doi-asserted-by":"publisher","award":["352692197"],"award-info":[{"award-number":["352692197"]}],"id":[{"id":"10.13039\/501100001659","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":["www.plosdigitalhealth.org"],"crossmark-restriction":false},"short-container-title":["PLOS Digit Health"],"abstract":"<jats:p>In medical documentation, vast amounts of unstructured text are generated that are still underutilized in current prognostic models. We investigate the potential of self-hosted large language models (LLM) to extract clinically meaningful, patient-specific information from routine clinical notes for personalized risk stratification in cancer care. We collected real-world medical notes from 2,708 non-small cell lung cancer (NSCLC) patients and 814 colon cancer patients documented before treatment at a large comprehensive cancer center. LLMs extracted key prognostic indicators, including comorbidities, metastatic sites, and qualitative descriptors of patient condition, in a zero-shot manner without prior task-specific training. Integrating these LLM-derived features into machine learning models significantly improved the prediction of overall survival compared to TNM staging alone (C-Index: NSCLC, 0.72 vs 0.64; colon cancer, 0.70 vs 0.59), and surpassed models using text embeddings. Based on the LLM-informed risk scores, patients were stratified into four distinct risk groups, enabling reclassification of 61.4% of NSCLC and 68.3% of colon cancer patients. Analysis of model drivers revealed that LLM-derived factors, such as the physical condition, substantially modulated the prognostic impact of TNM stage. These findings highlight the potential of self-hosted LLM to derive prognostically relevant information from unstructured clinical documentation and support clinical decision-making.<\/jats:p>","DOI":"10.1371\/journal.pdig.0001546","type":"journal-article","created":{"date-parts":[[2026,7,8]],"date-time":"2026-07-08T17:30:32Z","timestamp":1783531832000},"page":"e0001546","update-policy":"https:\/\/doi.org\/10.1371\/journal.pdig.corrections_policy","source":"Crossref","is-referenced-by-count":0,"title":["Large language models enable prognostic stratification of cancer patients using real-world clinical notes"],"prefix":"10.1371","volume":"5","author":[{"given":"Niklas","family":"Kiermeyer","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0009-0009-0960-2802","authenticated-orcid":true,"given":"Tim","family":"Lenfers","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Amin","family":"Dada","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Julian","family":"Friedrich","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Sameh","family":"Khattab","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Eric","family":"Knop","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Jan","family":"Egger","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Markus","family":"Pauly","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Andreas","family":"Jung","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Gr\u00e9goire","family":"Montavon","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-8772-4778","authenticated-orcid":true,"given":"Jens T.","family":"Siveke","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Marcel","family":"Wiesweg","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Stefan","family":"Kasper","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Ulf P.","family":"Neumann","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Frederick","family":"Klauschen","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Sylvia","family":"Hartmann","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Martin","family":"Schuler","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Philipp","family":"Keyl","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Jens","family":"Kleesiek","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-5617-091X","authenticated-orcid":true,"given":"Julius","family":"Keyl","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"340","published-online":{"date-parts":[[2026,7,8]]},"reference":[{"issue":"8","key":"pdig.0001546.ref001","article-title":"Pan-cancer integrative histology-genomic analysis via multimodal deep learning","volume":"40","author":"RJ Chen","year":"2022","journal-title":"Cancer Cell"},{"issue":"2","key":"pdig.0001546.ref002","doi-asserted-by":"crossref","first-page":"307","DOI":"10.1038\/s43018-024-00891-1","article-title":"Decoding pan-cancer treatment outcomes using multimodal real-world data and explainable artificial intelligence","volume":"6","author":"J Keyl","year":"2025","journal-title":"Nat Cancer"},{"issue":"5","key":"pdig.0001546.ref003","doi-asserted-by":"crossref","first-page":"100555","DOI":"10.1016\/j.esmoop.2022.100555","article-title":"Multimodal survival prediction in advanced pancreatic cancer using machine learning","volume":"7","author":"J Keyl","year":"2022","journal-title":"ESMO Open"},{"key":"pdig.0001546.ref004","doi-asserted-by":"crossref","first-page":"788740","DOI":"10.3389\/fonc.2021.788740","article-title":"Multimodal deep learning for prognosis prediction in renal cancer","volume":"11","author":"S Schulz","year":"2021","journal-title":"Front Oncol"},{"issue":"1","key":"pdig.0001546.ref005","doi-asserted-by":"crossref","first-page":"44","DOI":"10.1038\/s43856-023-00276-y","article-title":"Multimodal deep learning to predict prognosis in adult and pediatric brain tumors","volume":"3","author":"S Steyaert","year":"2023","journal-title":"Commun Med (Lond)"},{"issue":"4","key":"pdig.0001546.ref006","doi-asserted-by":"crossref","first-page":"351","DOI":"10.1038\/s42256-023-00633-5","article-title":"Multimodal data fusion for cancer biomarker discovery with deep learning","volume":"5","author":"S Steyaert","year":"2023","journal-title":"Nat Mach Intell"},{"issue":"1","key":"pdig.0001546.ref007","doi-asserted-by":"crossref","first-page":"181","DOI":"10.1038\/s41698-024-00670-2","article-title":"Biologically interpretable multi-task deep learning pipeline predicts molecular alterations, grade, and prognosis in glioma patients","volume":"8","author":"X Wu","year":"2024","journal-title":"NPJ Precis Oncol"},{"key":"pdig.0001546.ref008","author":"C Wong","year":"2025"},{"key":"pdig.0001546.ref009","article-title":"LLM-AIx: an open source pipeline for Information Extraction from unstructured medical text based on privacy preserving Large Language Models","author":"IC Wiest","year":"2024","journal-title":"medRxiv"},{"issue":"1","key":"pdig.0001546.ref010","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","author":"IC Wiest","year":"2024","journal-title":"NPJ Digit Med"},{"key":"pdig.0001546.ref011","author":"H Touvron","year":"2023"},{"key":"pdig.0001546.ref012","author":"G Team","year":"2024"},{"issue":"1","key":"pdig.0001546.ref013","doi-asserted-by":"crossref","first-page":"135","DOI":"10.1038\/s41746-023-00879-8","article-title":"The shaky foundations of large language models and foundation models for electronic health records","volume":"6","author":"M Wornow","year":"2023","journal-title":"NPJ Digit Med"},{"issue":"1","key":"pdig.0001546.ref014","doi-asserted-by":"crossref","first-page":"54","DOI":"10.1038\/s41698-024-00553-6","article-title":"All models are wrong and yours are useless: making clinical prediction models impactful for patients","volume":"8","author":"F Markowetz","year":"2024","journal-title":"NPJ Precis Oncol"},{"key":"pdig.0001546.ref015","author":"J Chen","year":"2024"},{"key":"pdig.0001546.ref016"},{"key":"pdig.0001546.ref017","doi-asserted-by":"crossref","first-page":"841","DOI":"10.1214\/08-AOAS169","article-title":"Random survival forests","volume":"2","author":"H Ishwaran","year":"2008","journal-title":"Ann Appl Stat"},{"issue":"4","key":"pdig.0001546.ref018","doi-asserted-by":"crossref","first-page":"757","DOI":"10.1093\/biostatistics\/kxu010","article-title":"Random survival forests for competing risks","volume":"15","author":"H Ishwaran","year":"2014","journal-title":"Biostatistics"},{"key":"pdig.0001546.ref019","author":"S Lundberg","year":"2017"},{"issue":"7972","key":"pdig.0001546.ref020","doi-asserted-by":"crossref","first-page":"172","DOI":"10.1038\/s41586-023-06291-2","article-title":"Large language models encode clinical knowledge","volume":"620","author":"K Singhal","year":"2023","journal-title":"Nature"},{"key":"pdig.0001546.ref021","author":"K Saab","year":"2024"},{"key":"pdig.0001546.ref022","doi-asserted-by":"crossref","first-page":"728","DOI":"10.1038\/s41586-024-08167-5","article-title":"Automated real-world data integration improves cancer outcome prediction","volume":"636","author":"J Jee","year":"2024","journal-title":"Nat"},{"issue":"3","key":"pdig.0001546.ref023","doi-asserted-by":"crossref","first-page":"917","DOI":"10.1038\/s41591-024-03408-6","article-title":"Nationwide real-world implementation of AI for cancer detection in population-based mammography screening","volume":"31","author":"N Eisemann","year":"2025","journal-title":"Nat Med"},{"key":"pdig.0001546.ref024","author":"JS Boyle","year":"2023"},{"issue":"1","key":"pdig.0001546.ref025","doi-asserted-by":"crossref","first-page":"96","DOI":"10.1038\/s43856-025-00808-8","article-title":"Comprehensive testing of large language models for extraction of structured data in pathology","volume":"5","author":"B Grothey","year":"2025","journal-title":"Commun Med (Lond)"},{"issue":"1","key":"pdig.0001546.ref026","doi-asserted-by":"crossref","first-page":"222","DOI":"10.1038\/s41746-024-01219-0","article-title":"A scoping review of large language model based approaches for information extraction from radiology reports","volume":"7","author":"D Reichenpfader","year":"2024","journal-title":"NPJ Digit Med"},{"issue":"2","key":"pdig.0001546.ref027","doi-asserted-by":"crossref","first-page":"109022","DOI":"10.1016\/j.isci.2024.109022","article-title":"Evaluation of GPT-4 for 10-year cardiovascular risk prediction: insights from the UK Biobank and KoGES data","volume":"27","author":"C Han","year":"2024","journal-title":"iScience"},{"issue":"1","key":"pdig.0001546.ref028","doi-asserted-by":"crossref","first-page":"21728","DOI":"10.1038\/s41598-025-06591-9","article-title":"Comparison of AI chatbot predicted and realworld survival outcomes in hepatocellular carcinoma","volume":"15","author":"EE Kavak","year":"2025","journal-title":"Sci Rep"},{"issue":"7969","key":"pdig.0001546.ref029","doi-asserted-by":"crossref","first-page":"357","DOI":"10.1038\/s41586-023-06160-y","article-title":"Health system-scale language models are all-purpose prediction engines","volume":"619","author":"LY Jiang","year":"2023","journal-title":"Nature"},{"key":"pdig.0001546.ref030","first-page":"197","article-title":"Contextual embeddings from clinical notes improves prediction of sepsis","volume":"2020","author":"F Amrollahi","year":"2021","journal-title":"AMIA Annu Symp Proc"},{"key":"pdig.0001546.ref031","doi-asserted-by":"crossref","first-page":"747250","DOI":"10.3389\/fonc.2021.747250","article-title":"Deep-learning-based natural language processing of serial free-text radiological reports for predicting rectal cancer patient survival","volume":"11","author":"S Kim","year":"2021","journal-title":"Front Oncol"},{"issue":"6","key":"pdig.0001546.ref032","doi-asserted-by":"crossref","first-page":"488","DOI":"10.1016\/j.trecan.2020.12.010","article-title":"How comorbidities shape cancer biology and survival","volume":"7","author":"G Panigrahi","year":"2021","journal-title":"Trends Cancer"},{"issue":"5","key":"pdig.0001546.ref033","doi-asserted-by":"crossref","first-page":"365","DOI":"10.1038\/s43587-022-00231-x","article-title":"The importance of aging in cancer research","volume":"2","year":"2022","journal-title":"Nat Aging"},{"issue":"1","key":"pdig.0001546.ref034","doi-asserted-by":"crossref","first-page":"102219","DOI":"10.1016\/j.esmoop.2023.102219","article-title":"Prognostic value of deep learning-derived body composition in advanced pancreatic cancer-a retrospective multicenter study","volume":"9","author":"J Keyl","year":"2024","journal-title":"ESMO Open"},{"issue":"1","key":"pdig.0001546.ref035","doi-asserted-by":"crossref","first-page":"545","DOI":"10.1002\/jcsm.13158","article-title":"Deep learning-based assessment of body composition and liver tumour burden for survival modelling in advanced colorectal cancer","volume":"14","author":"J Keyl","year":"2023","journal-title":"J Cachexia Sarcopenia Muscle"},{"issue":"1","key":"pdig.0001546.ref036","doi-asserted-by":"crossref","first-page":"60","DOI":"10.1038\/s41591-024-03425-5","article-title":"The TRIPOD-LLM reporting guideline for studies using large language models","volume":"31","author":"J Gallifant","year":"2025","journal-title":"Nat Med"},{"key":"pdig.0001546.ref037","first-page":"2825","article-title":"Scikit-learn: machine learning in Python","volume":"12","author":"F Pedregosa","year":"2011","journal-title":"J Mach Learn Res"},{"key":"pdig.0001546.ref038","volume-title":"R: a language and environment for statistical computing","author":"R Core Team","year":"2022"},{"key":"pdig.0001546.ref039","volume-title":"A package for survival analysis in R","author":"TM Therneau","year":"2023"},{"key":"pdig.0001546.ref040","first-page":"1","article-title":"Scikit-survival: a library for time-to-event analysis built on top of scikit-learn","volume":"21","author":"S P\u00f6lsterl","year":"2020","journal-title":"J Mach Learn Res"},{"issue":"3","key":"pdig.0001546.ref041","doi-asserted-by":"crossref","first-page":"261","DOI":"10.1038\/s41592-019-0686-2","article-title":"SciPy 1.0: fundamental algorithms for scientific computing in Python","volume":"17","author":"P Virtanen","year":"2020","journal-title":"Nat Methods"}],"container-title":["PLOS Digital Health"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/dx.plos.org\/10.1371\/journal.pdig.0001546","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,7,8]],"date-time":"2026-07-08T17:30:42Z","timestamp":1783531842000},"score":1,"resource":{"primary":{"URL":"https:\/\/dx.plos.org\/10.1371\/journal.pdig.0001546"}},"subtitle":[],"editor":[{"given":"Matthew","family":"Watson","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"editor"}]}],"short-title":[],"issued":{"date-parts":[[2026,7,8]]},"references-count":41,"journal-issue":{"issue":"7","published-online":{"date-parts":[[2026,7,8]]}},"URL":"https:\/\/doi.org\/10.1371\/journal.pdig.0001546","relation":{},"ISSN":["2767-3170"],"issn-type":[{"value":"2767-3170","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,7,8]]}}}