{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,7,9]],"date-time":"2026-07-09T16:09:38Z","timestamp":1783613378369,"version":"3.55.0"},"reference-count":43,"publisher":"Springer Science and Business Media LLC","issue":"1","license":[{"start":{"date-parts":[[2025,12,13]],"date-time":"2025-12-13T00:00:00Z","timestamp":1765584000000},"content-version":"tdm","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0"},{"start":{"date-parts":[[2025,12,19]],"date-time":"2025-12-19T00:00:00Z","timestamp":1766102400000},"content-version":"vor","delay-in-days":6,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0"}],"funder":[{"DOI":"10.13039\/501100021171","name":"Basic and Applied Basic Research Foundation of Guangdong Province","doi-asserted-by":"publisher","award":["2024A1515220067"],"award-info":[{"award-number":["2024A1515220067"]}],"id":[{"id":"10.13039\/501100021171","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":["link.springer.com"],"crossmark-restriction":false},"short-container-title":["npj Digit. Med."],"abstract":"<jats:title>Abstract<\/jats:title>\n                  <jats:p>\n                    Perioperative complications are a major global concern, yet manual detection suffers from 27% under\u2011reporting and frequent misclassification. Clinical LLM deployment is constrained by data sovereignty, compute cost, and limited locally deployable model performance. We show targeted prompt engineering plus Low\u2011Rank Adaptation (LoRA) fine\u2011tuning converts smaller open\u2011source LLMs into expert\u2011level diagnostic tools. In dual\u2011center validation, we built a framework simultaneously identifying and grading 22 complication severities. State\u2011of\u2011the\u2011art models outperformed human experts; Chain\u2011of\u2011Thought prompting significantly improved general models (\n                    <jats:italic>p<\/jats:italic>\n                    \u2009&lt;\u20090.001) while preserving reasoning models\u2019 performance. Across documentation length quartiles, AI models maintained F1\u2009&gt;\u20090.64, whereas human performance declined from 0.73 to 0.45, demonstrating superior robustness to documentation complexity. Our targeted strategy\u2014decomposing detection into focused single\u2011complication assessments\u2014improved small models, with further gains from LoRA. On external validation (Center 2), the optimized 4B model\u2019s micro\u2011F1 rose from 0.28 to 0.64, approaching human experts (F1\u2009=\u20090.69), driven by the targeted strategy (\u0394F1\u2009=\u20090.256, 95% CI [0.181, 0.336]) and LoRA (\u0394F1\u2009=\u20090.103, 95% CI [0.023, 0.186]). Concurrently, the 8B model surpassed human experts (F1\u2009&gt;\u20090.70). Optimized small models enable expert\u2011level accuracy with local deployment and preserved data sovereignty, offering a practical path for resource\u2011limited healthcare.\n                  <\/jats:p>","DOI":"10.1038\/s41746-025-02139-3","type":"journal-article","created":{"date-parts":[[2025,12,13]],"date-time":"2025-12-13T09:36:23Z","timestamp":1765618583000},"update-policy":"https:\/\/doi.org\/10.1007\/springer_crossmark_policy","source":"Crossref","is-referenced-by-count":6,"title":["Enhancing privacy-preserving deployable large language models for perioperative complication detection: a targeted strategy with LoRA fine-tuning"],"prefix":"10.1038","volume":"8","author":[{"given":"Shaowei","family":"Gao","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Xu","family":"Zhao","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Lihui","family":"Chen","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Junrong","family":"Yu","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Shuning","family":"Tian","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Huaqiang","family":"Zhou","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Jingru","family":"Chen","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Sizhe","family":"Long","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Qiulan","family":"He","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Xia","family":"Feng","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"297","published-online":{"date-parts":[[2025,12,13]]},"reference":[{"key":"2139_CR1","doi-asserted-by":"publisher","first-page":"139","DOI":"10.1016\/S0140-6736(08)60878-8","volume":"372","author":"TG Weiser","year":"2008","unstructured":"Weiser, T. G. et al. An estimation of the global volume of surgery: a modelling strategy based on available data. Lancet 372, 139\u2013144 (2008).","journal-title":"Lancet"},{"key":"2139_CR2","doi-asserted-by":"publisher","first-page":"1059","DOI":"10.1016\/S0140-6736(12)61148-9","volume":"380","author":"RM Pearse","year":"2012","unstructured":"Pearse, R. M. et al. Mortality after surgery in Europe: a 7 day cohort study. Lancet 380, 1059\u20131065 (2012).","journal-title":"Lancet"},{"key":"2139_CR3","unstructured":"WHO Guidelines for Safe Surgery 2009: Safe Surgery Saves Lives. (World Health Organization, Geneva, 2009)."},{"key":"2139_CR4","doi-asserted-by":"publisher","first-page":"292.e1","DOI":"10.1016\/j.urolonc.2024.04.010","volume":"42","author":"JY Hsueh","year":"2024","unstructured":"Hsueh, J. Y., Nethala, D., Singh, S., Linehan, W. M. & Ball, M. W. Investigating the clinical reasoning abilities of large language model GPT-4: an analysis of postoperative complications from renal surgeries. Urol. Oncol. 42, 292.e1\u2013292.e7 (2024).","journal-title":"Urol. Oncol."},{"key":"2139_CR5","doi-asserted-by":"publisher","first-page":"543","DOI":"10.18203\/2319-2003.ijbcp20241657","volume":"13","author":"SB Vampana","year":"2024","unstructured":"Vampana, S. B., Jayanthi, E. S. S., Mary, D. A. & Sriniketh, C. Artificial intelligence-driven patient monitoring for adverse event detection in clinical trials. Int. J. Basic Clin. Pharm. 13, 543\u2013550 (2024).","journal-title":"Int. J. Basic Clin. Pharm."},{"key":"2139_CR6","doi-asserted-by":"publisher","first-page":"54","DOI":"10.1016\/j.jss.2010.08.042","volume":"173","author":"EJ Veen","year":"2012","unstructured":"Veen, E. J., Janssen-Heijnen, M. L. G., Bosma, E., de Jongh, M. A. C. & Roukema, J. A. The accuracy of complications documented in a prospective complication registry. J. Surg. Res 173, 54\u201359 (2012).","journal-title":"J. Surg. Res"},{"key":"2139_CR7","unstructured":"Vaswani, A. et al. Attention is All You Need. In Advances in Neural Information Processing Systems, Vol. 30 (Curran Associates, Inc., 2017)."},{"key":"2139_CR8","doi-asserted-by":"publisher","first-page":"928","DOI":"10.1001\/jamasurg.2024.1621","volume":"159","author":"P Chung","year":"2024","unstructured":"Chung, P. et al. Large language model capabilities in perioperative risk prediction and prognostication. JAMA Surg. 159, 928\u2013937 (2024).","journal-title":"JAMA Surg."},{"key":"2139_CR9","doi-asserted-by":"publisher","first-page":"664","DOI":"10.1097\/ICO.0000000000003492","volume":"43","author":"M Delsoz","year":"2024","unstructured":"Delsoz, M. et al. Performance of ChatGPT in diagnosis of corneal eye diseases. Cornea 43, 664\u2013670 (2024).","journal-title":"Cornea"},{"key":"2139_CR10","doi-asserted-by":"publisher","first-page":"e0000198","DOI":"10.1371\/journal.pdig.0000198","volume":"2","author":"TH Kung","year":"2023","unstructured":"Kung, T. H. et al. Performance of ChatGPT on USMLE: potential for AI-assisted medical education using large language models. PLoS Digit. Health 2, e0000198 (2023).","journal-title":"PLoS Digit. Health"},{"key":"2139_CR11","doi-asserted-by":"publisher","first-page":"943","DOI":"10.1038\/s41591-024-03423-7","volume":"31","author":"K Singhal","year":"2025","unstructured":"Singhal, K. et al. Toward expert-level medical question answering with large language models. Nat. Med. 31, 943\u2013950 (2025).","journal-title":"Nat. Med."},{"key":"2139_CR12","doi-asserted-by":"publisher","first-page":"315","DOI":"10.1001\/jama.2023.9651","volume":"330","author":"T Minssen","year":"2023","unstructured":"Minssen, T., Vayena, E. & Cohen, I. G. The challenges for regulating medical use of ChatGPT and other large language models. JAMA 330, 315\u2013316 (2023).","journal-title":"JAMA"},{"key":"2139_CR13","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1038\/s41746-023-00873-0","volume":"6","author":"B Mesk\u00f3","year":"2023","unstructured":"Mesk\u00f3, B. & Topol, E. J. The imperative for regulatory oversight of large language models (or generative AI) in healthcare. npj Digit. Med. 6, 1\u20136 (2023).","journal-title":"npj Digit. Med."},{"key":"2139_CR14","doi-asserted-by":"publisher","first-page":"257","DOI":"10.1038\/s41746-024-01233-2","volume":"7","author":"IC Wiest","year":"2024","unstructured":"Wiest, I. C. et al. Privacy-preserving large language models for structured medical information retrieval. NPJ Digit. Med. 7, 257 (2024).","journal-title":"NPJ Digit. Med."},{"key":"2139_CR15","unstructured":"Hu, E. J. et al. LoRA: Low-rank adaptation of large language models. In International Conference on Learning Representations (2022)."},{"key":"2139_CR16","unstructured":"Wei, J. et al. Chain-of-thought prompting elicits reasoning in large language models. In Advances in Neural Information Processing Systems (eds Koyejo, S. et al.) vol. 35, 24824\u201324837 (Curran Associates, Inc., 2022)."},{"key":"2139_CR17","doi-asserted-by":"publisher","first-page":"e55499","DOI":"10.2196\/55499","volume":"12","author":"E Asgari","year":"2024","unstructured":"Asgari, E. et al. Impact of electronic health record use on cognitive load and burnout among clinicians: narrative review. JMIR Med. Inf. 12, e55499 (2024).","journal-title":"JMIR Med. Inf."},{"key":"2139_CR18","doi-asserted-by":"publisher","first-page":"e999","DOI":"10.1097\/PTS.0000000000001002","volume":"18","author":"S Nijor","year":"2022","unstructured":"Nijor, S., Rallis, G., Lad, N. & Gokcen, E. Patient safety issues from information overload in electronic medical records. J. Patient Saf. 18, e999\u2013e1003 (2022).","journal-title":"J. Patient Saf."},{"key":"2139_CR19","unstructured":"Khot, T. et al. Decomposed prompting: a modular approach for solving complex tasks. In The Eleventh International Conference on Learning Representations (2023)."},{"key":"2139_CR20","doi-asserted-by":"publisher","unstructured":"Stein, A., Naik, A., Velingker, N., Naik, M. & Wong, E. The road to generalizable neuro-symbolic learning should be paved with foundation models. Preprint at https:\/\/doi.org\/10.48550\/arXiv.2505.24874 (2025).","DOI":"10.48550\/arXiv.2505.24874"},{"key":"2139_CR21","doi-asserted-by":"crossref","unstructured":"Oh, M.-Y. et al. Machine learning\u2013based explainable automated nonlinear computation scoring system for health score and an application for prediction of perioperative stroke: retrospective study. J Med. Int. Res. https:\/\/www.jmir.org\/2025\/1\/e58021 (2025).","DOI":"10.2196\/58021"},{"key":"2139_CR22","doi-asserted-by":"crossref","unstructured":"Huang, P. et al. Machine learning\u2013based prediction of early complications following surgery for intestinal obstruction: multicenter retrospective study. J. Med. Int. https:\/\/www.jmir.org\/2025\/1\/e68354 (2025).","DOI":"10.2196\/68354"},{"key":"2139_CR23","doi-asserted-by":"publisher","DOI":"10.1186\/s40537-025-01155-0","volume":"12","author":"W Jiang","year":"2025","unstructured":"Jiang, W. et al. An artificial intelligence platform for predicting postoperative complications in metastatic spinal surgery: development and validation study. J. Big Data 12, 120 (2025).","journal-title":"J. Big Data"},{"key":"2139_CR24","doi-asserted-by":"publisher","unstructured":"Shoham, G. et al. Utilizing artificial intelligence for predicting postoperative complications in breast reduction surgery: a comprehensive retrospective analysis of predictive features and outcomes. https:\/\/doi.org\/10.1093\/asj\/sjaf021.","DOI":"10.1093\/asj\/sjaf021"},{"key":"2139_CR25","doi-asserted-by":"publisher","first-page":"7957","DOI":"10.1245\/s10434-024-15861-1","volume":"31","author":"M Angelucci","year":"2024","unstructured":"Angelucci, M. et al. Surgical outcome in retroperitoneal sarcoma surgery: accuracy of P-POSSUM, ACS-NSQIP, and inflammatory biomarkers prognostic index (IBPI) risk-calculators for prediction of severe and overall morbidity. Ann. Surg. Oncol. 31, 7957\u20137966 (2024).","journal-title":"Ann. Surg. Oncol."},{"key":"2139_CR26","doi-asserted-by":"publisher","unstructured":"Bai, J. et al. Qwen Technical Report. Preprint at https:\/\/doi.org\/10.48550\/arXiv.2309.16609 (2023).","DOI":"10.48550\/arXiv.2309.16609"},{"key":"2139_CR27","doi-asserted-by":"publisher","unstructured":"Rust, P., Pfeiffer, J., Vuli\u0107, I., Ruder, S. & Gurevych, I. How Good is Your Tokenizer? On the Monolingual Performance of Multilingual Language Models. In Proc. 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers) (eds Zong, C., Xia, F., Li, W. & Navigli, R.)3118\u20133135 (Association for Computational Linguistics, Online, 2021). https:\/\/doi.org\/10.18653\/v1\/2021.acl-long.243.","DOI":"10.18653\/v1\/2021.acl-long.243"},{"key":"2139_CR28","doi-asserted-by":"publisher","first-page":"145","DOI":"10.1055\/a-2405-2489","volume":"63","author":"S Yada","year":"2024","unstructured":"Yada, S., Nakamura, Y., Wakamiya, S. & Aramaki, E. Cross-lingual natural language processing on limited annotated case\/radiology reports in English and Japanese: insights from the real-MedNLP Workshop. Methods Inf. Med. 63, 145\u2013163 (2024).","journal-title":"Methods Inf. Med."},{"key":"2139_CR29","doi-asserted-by":"publisher","unstructured":"Ruffini, F., Ayllon, E. M., Shen, L., Soda, P. & Guarrasi, V. Benchmarking foundation models and parameter-efficient fine-tuning for prognosis prediction in medical imaging. Preprint at https:\/\/doi.org\/10.48550\/arXiv.2506.18434 (2025).","DOI":"10.48550\/arXiv.2506.18434"},{"key":"2139_CR30","doi-asserted-by":"publisher","unstructured":"Rahman, U., Imam, R., Yaqub, M. & Mahapatra, D. Decoupling clinical and class-agnostic features for reliable few-shot adaptation under shift. In Uncertainty for Safe Utilization of Machine Learning in Medical Imaging (eds Sudre, C. H. et al.) 123\u2013133 (Springer Nature Switzerland, Cham, 2026). https:\/\/doi.org\/10.1007\/978-3-032-06593-3_12.","DOI":"10.1007\/978-3-032-06593-3_12"},{"key":"2139_CR31","doi-asserted-by":"publisher","first-page":"88","DOI":"10.1097\/EJA.0000000000000118","volume":"32","author":"I Jammer","year":"2015","unstructured":"Jammer, I. et al. Standards for definitions and use of outcome measures for clinical effectiveness research in perioperative medicine: European Perioperative Clinical Outcome (EPCO) definitions: a statement from the ESA-ESICM joint taskforce on perioperative outcome measures. Eur. J. Anaesthesiol. 32, 88\u2013105 (2015).","journal-title":"Eur. J. Anaesthesiol."},{"key":"2139_CR32","doi-asserted-by":"crossref","unstructured":"Johnson, A. E. W. et al. MIMIC-IV, a freely accessible electronic health record dataset. Sci. Data 10 (2023).","DOI":"10.1038\/s41597-022-01899-x"},{"key":"2139_CR33","doi-asserted-by":"publisher","first-page":"205","DOI":"10.1097\/01.sla.0000133083.54934.ae","volume":"240","author":"D Dindo","year":"2004","unstructured":"Dindo, D., Demartines, N. & Clavien, P.-A. Classification of surgical complications: a new proposal with evaluation in a cohort of 6336 patients and results of a survey. Ann. Surg. 240, 205\u2013213 (2004).","journal-title":"Ann. Surg."},{"key":"2139_CR34","doi-asserted-by":"publisher","first-page":"633","DOI":"10.1038\/s41586-025-09422-z","volume":"645","author":"D Guo","year":"2025","unstructured":"Guo, D. et al. DeepSeek-R1 incentivizes reasoning in LLMs through reinforcement learning. Nature 645, 633\u2013638 (2025).","journal-title":"Nature"},{"key":"2139_CR35","unstructured":"DeepSeek-AI. DeepSeek-V3 Technical Report. 2412.19437 https:\/\/arxiv.org\/abs\/2412.19437 (2024)."},{"key":"2139_CR36","unstructured":"Qwen Team. QwQ-32B: Embracing the Power of Reinforcement Learning. https:\/\/qwenlm.github.io\/blog\/qwq-32b\/ (2025)."},{"key":"2139_CR37","unstructured":"Yang, A. et al. Qwen3 Technical Report. arXiv preprint arXiv:2505.09388 (2025)."},{"key":"2139_CR38","doi-asserted-by":"publisher","first-page":"29733","DOI":"10.1609\/aaai.v39i28.35383","volume":"39","author":"Y Zhao","year":"2025","unstructured":"Zhao, Y. et al. SWIFT: A Scalable Lightweight Infrastructure for Fine-Tuning. AAAI 39, 29733\u201329735 (2025).","journal-title":"AAAI"},{"key":"2139_CR39","doi-asserted-by":"publisher","unstructured":"Kwon, W. et al. Efficient memory management for large language model serving with pagedattention. In Proc. 29th Symposium on Operating Systems Principles 611\u2013626 (Association for Computing Machinery, New York, NY, USA, 2023). https:\/\/doi.org\/10.1145\/3600006.3613165.","DOI":"10.1145\/3600006.3613165"},{"key":"2139_CR40","unstructured":"Willard, B. T. & Louf, R. Efficient Guided Generation for LLMs. arXiv preprint arXiv:2307.09702 (2023)."},{"key":"2139_CR41","doi-asserted-by":"publisher","unstructured":"Renze, M. & Guven, E. The effect of sampling temperature on problem solving in large language models. In Findings of the Association for Computational Linguistics: EMNLP 2024, 7346\u20137356 (2024). https:\/\/doi.org\/10.18653\/v1\/2024.findings-emnlp.432.","DOI":"10.18653\/v1\/2024.findings-emnlp.432"},{"key":"2139_CR42","doi-asserted-by":"publisher","first-page":"159","DOI":"10.2307\/2529310","volume":"33","author":"JR Landis","year":"1977","unstructured":"Landis, J. R. & Koch, G. G. The measurement of observer agreement for categorical data. Biometrics 33, 159\u2013174 (1977).","journal-title":"Biometrics"},{"key":"2139_CR43","doi-asserted-by":"publisher","first-page":"164","DOI":"10.4338\/ACI-2011-11-RA-0070","volume":"3","author":"PD Stetson","year":"2012","unstructured":"Stetson, P. D., Bakken, S., Wrenn, J. O. & Siegler, E. L. Assessing electronic note quality using the physician documentation quality instrument (PDQI-9). Appl. Clin. Inf. 3, 164\u2013174 (2012).","journal-title":"Appl. Clin. Inf."}],"container-title":["npj Digital Medicine"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/www.nature.com\/articles\/s41746-025-02139-3.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/www.nature.com\/articles\/s41746-025-02139-3","content-type":"text\/html","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/www.nature.com\/articles\/s41746-025-02139-3.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,12,19]],"date-time":"2025-12-19T11:57:15Z","timestamp":1766145435000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.nature.com\/articles\/s41746-025-02139-3"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,12,13]]},"references-count":43,"journal-issue":{"issue":"1","published-online":{"date-parts":[[2025,12]]}},"alternative-id":["2139"],"URL":"https:\/\/doi.org\/10.1038\/s41746-025-02139-3","relation":{},"ISSN":["2398-6352"],"issn-type":[{"value":"2398-6352","type":"electronic"}],"subject":[],"published":{"date-parts":[[2025,12,13]]},"assertion":[{"value":"20 June 2025","order":1,"name":"received","label":"Received","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"31 October 2025","order":2,"name":"accepted","label":"Accepted","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"13 December 2025","order":3,"name":"first_online","label":"First Online","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"The authors declare no competing interests.","order":1,"name":"Ethics","group":{"name":"EthicsHeading","label":"Competing interests"}}],"article-number":"773"}}