{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2025,7,4]],"date-time":"2025-07-04T04:05:19Z","timestamp":1751601919219,"version":"3.41.0"},"reference-count":22,"publisher":"Springer Science and Business Media LLC","issue":"1","license":[{"start":{"date-parts":[[2025,7,3]],"date-time":"2025-07-03T00:00:00Z","timestamp":1751500800000},"content-version":"tdm","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0"},{"start":{"date-parts":[[2025,7,3]],"date-time":"2025-07-03T00:00:00Z","timestamp":1751500800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0"}],"content-domain":{"domain":["link.springer.com"],"crossmark-restriction":false},"short-container-title":["BMC Med Inform Decis Mak"],"abstract":"<jats:title>Abstract<\/jats:title>\n          <jats:sec>\n            <jats:title>Background<\/jats:title>\n            <jats:p>Prompt-based learning involves the additions of prompts (i.e., templates) to the input of pre-trained large language models (PLMs) to adapt them to specific tasks with minimal training. This technique is particularly advantageous in clinical scenarios where the amount of annotated data is limited. This study aims to investigate the impact of template position on model performance and training efficiency in clinical note classification tasks using prompt-based learning, especially in zero- and few-shot settings.<\/jats:p>\n          <\/jats:sec>\n          <jats:sec>\n            <jats:title>Methods<\/jats:title>\n            <jats:p>We developed a keyword-optimized template insertion method (KOTI) to enhance model performance by strategically placing prompt templates near relevant clinical information within the notes. The method involves defining task-specific keywords, identifying sentences containing these keywords, and inserting the prompt template in their vicinity. We compared KOTI with standard template insertion (STI) methods in which the template is directly appended at the end of the input text. Specifically, we compared STI with na\u00efve tail-truncation (STI-s) and STI with keyword-optimized input truncation (STI-k). Experiments were conducted using two pre-trained encoder models, GatorTron and ClinicalBERT, and two decoder models, BioGPT and ClinicalT5, across five classification tasks, including dysmenorrhea, peripheral vascular disease, depression, osteoarthritis, and smoking status classification.<\/jats:p>\n          <\/jats:sec>\n          <jats:sec>\n            <jats:title>Results<\/jats:title>\n            <jats:p>Our experiments revealed that the KOTI approach consistently outperformed both STI-s and STI-k in zero-shot and few-shot scenarios for encoder models, with KOTI yielding a significant 24% F1 improvement over STI-k for GatorTron and 8% for Clinical BERT. Additionally, training with balanced examples further enhanced performance, particularly under few-shot conditions. In contrast, decoder-based models exhibited inconsistent results, with KOTI showing significant improvement in F1 score over STI-k for BioGPT (+19%), but a significant drop for ClinicalT5 (\u221218%), suggesting that KOTI is not beneficial across all transformer model architectures.<\/jats:p>\n          <\/jats:sec>\n          <jats:sec>\n            <jats:title>Conclusion<\/jats:title>\n            <jats:p>Our findings underscore the significance of template position in prompt-based fine-tuning of encoder models and highlights KOTI\u2019s potential to optimize real-world clinical note classification tasks with few training examples.<\/jats:p>\n          <\/jats:sec>","DOI":"10.1186\/s12911-025-03071-y","type":"journal-article","created":{"date-parts":[[2025,7,3]],"date-time":"2025-07-03T09:00:45Z","timestamp":1751533245000},"update-policy":"https:\/\/doi.org\/10.1007\/springer_crossmark_policy","source":"Crossref","is-referenced-by-count":0,"title":["Keyword-optimized template insertion for clinical note classification via prompt-based learning"],"prefix":"10.1186","volume":"25","author":[{"given":"Eugenia","family":"Alleva","sequence":"first","affiliation":[]},{"given":"Isotta","family":"Landi","sequence":"additional","affiliation":[]},{"given":"Leslee J.","family":"Shaw","sequence":"additional","affiliation":[]},{"given":"Erwin","family":"B\u00f6ttinger","sequence":"additional","affiliation":[]},{"given":"Ipek","family":"Ensari","sequence":"additional","affiliation":[]},{"given":"Thomas J.","family":"Fuchs","sequence":"additional","affiliation":[]}],"member":"297","published-online":{"date-parts":[[2025,7,3]]},"reference":[{"key":"3071_CR1","unstructured":"Vaswani A, Shazeer N, Parmar N, Uszkoreit J, Jones L, Gomez AN, et al. Attention is all you need [Internet]. arXiv; 2023 [2024 Jul 25]. Available from: http:\/\/arxiv.org\/abs\/1706.03762"},{"key":"3071_CR2","doi-asserted-by":"crossref","unstructured":"Gao T, Fisch A, Chen D Making pre-trained language models better few-shot learners. In: Zong C, Xia F, Li W, Navigli R, editors. Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers) [Internet]. Online: Association for Computational Linguistics; 2021 [2024 Jul 25]. p. 3816\u201330. Available from: https:\/\/aclanthology.org\/2021.acl-long.295","DOI":"10.18653\/v1\/2021.acl-long.295"},{"key":"3071_CR3","unstructured":"Xia F, Yetisgen-Yildiz M Clinical corpus annotation: challenges and strategies."},{"key":"3071_CR4","unstructured":"LiuPengfei Y, FuJinlan J, HayashiHiroaki N. Pre-train, prompt, and predict: a systematic survey of prompting methods in natural language processing. ACM Comput Surv [Internet]. 2023 Jan 13; [2024 Jul 25]. Available from: https:\/\/dl.acm.org\/doi\/10.1145\/3560815."},{"key":"3071_CR5","unstructured":"Brown T, Mann B, Ryder N, Subbiah M, Kaplan JD, Dhariwal P, et al. Language models are few-shot learners. In: Advances in Neural Information Processing Systems [Internet]. Curran Associates, Inc.; 2020 [2024 Jul 25]. p. 1877\u2013901. Available from: https:\/\/papers.nips.cc\/paper\/2020\/hash\/1457c0d6bfcb4967418bfb8ac142f64a-Abstract.html"},{"key":"3071_CR6","doi-asserted-by":"crossref","unstructured":"Schick T, Sch\u00fctze H. Exploiting cloze questions for few shot text classification and natural language inference [Internet]. arXiv; 2021; [2024 Jul 25]. Available from: http:\/\/arxiv.org\/abs\/2001.07676.","DOI":"10.18653\/v1\/2021.eacl-main.20"},{"key":"3071_CR7","unstructured":"Landi I, Alleva E, Valentine AA, Lepow LA, Charney AW Clinical text deduplication practices for efficient pretraining and improved clinical tasks [Internet]. arXiv; 2023 [2024 Jul 25]. Available from: http:\/\/arxiv.org\/abs\/2312.09469"},{"key":"3071_CR8","doi-asserted-by":"crossref","unstructured":"Ramachandran GK, Fu Y, Han B, Lybarger K, Dobbins N, Uzuner O, et al. Prompt-based extraction of social determinants of health using few-shot learning. In: Naumann T, Ben Abacha A, Bethard S, Roberts K, Rumshisky A, editors. Proceedings of the 5th Clinical Natural Language Processing Workshop [Internet]. Toronto, Canada: Association for Computational Linguistics; 2023 [2025 Mar 1]. p. 385\u201393. Available from: https:\/\/aclanthology.org\/2023.clinicalnlp-1.41\/","DOI":"10.18653\/v1\/2023.clinicalnlp-1.41"},{"key":"3071_CR9","first-page":"972","volume":"2022","author":"S Sivarajkumar","year":"2022","unstructured":"Sivarajkumar S, Wang Y. Healthprompt: a zero-shot learning paradigm for clinical natural language processing. AMIA Annu Symp Proc. 2022;2022:972\u201381.","journal-title":"AMIA Annu Symp Proc"},{"key":"3071_CR10","doi-asserted-by":"crossref","unstructured":"Taylor N, Zhang Y, Joyce DW, Gao Z, Kormilitzin A, Nevado-Holgado A. Clinical Prompt Learning With Frozen Language Models. IEEE Trans Neural Netw Learn Syst. 202;35(11):16453\u201363.","DOI":"10.1109\/TNNLS.2023.3294633"},{"key":"3071_CR11","unstructured":"Wang Y, Deng J, Wang T, Zheng B, Hu S, Liu X, et al. Exploiting prompt learning with pre-trained language models for alzheimer\u2019s disease detection [Internet]. arXiv; 2023 [2024 Jul 25]. Available from: http:\/\/arxiv.org\/abs\/2210.16539"},{"key":"3071_CR12","doi-asserted-by":"crossref","unstructured":"Yang Z, Wang S, Rawat BPS, Mitra A, Yu H. Knowledge injected prompt based fine-tuning for multi-label few-shot ICD coding. Proc Conf Empir Methods Nat Lang Process. Dec 2022:1767\u201381.","DOI":"10.18653\/v1\/2022.findings-emnlp.127"},{"key":"3071_CR13","doi-asserted-by":"crossref","unstructured":"Wu Z, Wang S, Gu J, Hou R, Dong Y, Vydiswaran VGV, et al. IDPG: an instance-dependent prompt generation method. In: Carpuat M, de Marneffe MC, Ruiz IV M, editors. Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies [Internet]. Seattle, United States: Association for Computational Linguistics; 2022 [2024 Jul 25]. p. 5507\u201321. Available from: https:\/\/aclanthology.org\/2022.naacl-main.403","DOI":"10.18653\/v1\/2022.naacl-main.403"},{"key":"3071_CR14","doi-asserted-by":"crossref","unstructured":"Mao J, Middleton SE, Niranjan M Do prompt positions really matter? [Internet]. arXiv; 2024 [2024 Jul 25]. Available from: http:\/\/arxiv.org\/abs\/2305.14493","DOI":"10.18653\/v1\/2024.findings-naacl.258"},{"key":"3071_CR15","unstructured":"Yang X, Cheng W, Zhao X, Yu W, Petzold L, Chen H Dynamic prompting: a unified framework for prompt tuning [Internet]. arXiv; 2023 [2024 Jul 25]. Available from: http:\/\/arxiv.org\/abs\/2303.02909"},{"issue":"1","key":"3071_CR16","doi-asserted-by":"publisher","first-page":"14","DOI":"10.1197\/jamia.M2408","volume":"15","author":"\u00d6 Uzuner","year":"2008","unstructured":"Uzuner \u00d6, Goldstein I, Luo Y, Kohane I. Identifying patient smoking status from medical discharge records. J Am Med Inform Assoc. 2008;15(1):14\u201324.","journal-title":"J Am Med Inform Assoc"},{"issue":"4","key":"3071_CR17","doi-asserted-by":"publisher","first-page":"561","DOI":"10.1197\/jamia.M3115","volume":"16","author":"O Uzuner","year":"2009","unstructured":"Uzuner O. Recognizing obesity and comorbidities in sparse data. J Am Med Inform Assoc. 2009;16(4):561\u201370.","journal-title":"J Am Med Inform Assoc"},{"key":"3071_CR18","doi-asserted-by":"crossref","unstructured":"Ding N, Hu S, Zhao W, Chen Y, Liu Z, Zheng H, et al. OpenPrompt: an open-source framework for prompt-learning. In: Basile V, Kozareva Z, Stajner S, editors. Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics: System Demonstrations [Internet]. Dublin, Ireland: Association for Computational Linguistics; 2022 [2024 Jul 25]. p. 105\u201313. Available from: https:\/\/aclanthology.org\/2022.acl-demo.10","DOI":"10.18653\/v1\/2022.acl-demo.10"},{"issue":"1","key":"3071_CR19","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1038\/s41746-022-00742-2","volume":"5","author":"X Yang","year":"2022","unstructured":"Yang X, Chen A, PourNejatian N, Shin HC, Smith KE, Parisien C, et al. A large language model for electronic health records. Npj Digit Med. 2022 Dec 26;5(1):1\u20139.","journal-title":"Npj Digit Med"},{"key":"3071_CR20","doi-asserted-by":"crossref","unstructured":"Alsentzer E, Murphy J, Boag W, Weng WH, Jindi D, Naumann T, et al. Publicly available clinical bert embeddings. In: Rumshisky A, Roberts K, Bethard S, Naumann T, editors. Proceedings of the 2nd Clinical Natural Language Processing Workshop [Internet]. Minneapolis, Minnesota, USA: Association for Computational Linguistics; 2019 [2024 Jul 25]. p. 72\u201378. Available from: https:\/\/aclanthology.org\/W19-1909","DOI":"10.18653\/v1\/W19-1909"},{"key":"3071_CR21","doi-asserted-by":"crossref","unstructured":"Luo R, Sun L, Xia Y, Qin T, Zhang S, Poon H, et al. BioGPT: generative pre-trained transformer for biomedical text generation and mining. Brief Bioinform. 2022 Nov 1;23(6):bbac409.","DOI":"10.1093\/bib\/bbac409"},{"key":"3071_CR22","doi-asserted-by":"publisher","first-page":"5436","DOI":"10.18653\/v1\/2022.findings-emnlp.398","volume-title":"Findings of the Association for Computational Linguistics: EMNLP 2022 [Internet]","author":"Q Lu","year":"2022","unstructured":"Lu Q, Dou D, Nguyen T. ClinicalT5: a generative language model for clinical text. In: Goldberg Y, Kozareva Z, Zhang Y, editors. Findings of the Association for Computational Linguistics: EMNLP 2022 [Internet]. Abu Dhabi, United Arab Emirates: Association for Computational Linguistics; 2022 [Mar 1]. p. 5436\u201343. Available from: https:\/\/aclanthology.org\/2022.findings-emnlp.398\/."}],"container-title":["BMC Medical Informatics and Decision Making"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/link.springer.com\/content\/pdf\/10.1186\/s12911-025-03071-y.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/link.springer.com\/article\/10.1186\/s12911-025-03071-y\/fulltext.html","content-type":"text\/html","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/link.springer.com\/content\/pdf\/10.1186\/s12911-025-03071-y.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,7,3]],"date-time":"2025-07-03T09:00:51Z","timestamp":1751533251000},"score":1,"resource":{"primary":{"URL":"https:\/\/bmcmedinformdecismak.biomedcentral.com\/articles\/10.1186\/s12911-025-03071-y"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,7,3]]},"references-count":22,"journal-issue":{"issue":"1","published-online":{"date-parts":[[2025,12]]}},"alternative-id":["3071"],"URL":"https:\/\/doi.org\/10.1186\/s12911-025-03071-y","relation":{},"ISSN":["1472-6947"],"issn-type":[{"value":"1472-6947","type":"electronic"}],"subject":[],"published":{"date-parts":[[2025,7,3]]},"assertion":[{"value":"6 August 2024","order":1,"name":"received","label":"Received","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"13 June 2025","order":2,"name":"accepted","label":"Accepted","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"3 July 2025","order":3,"name":"first_online","label":"First Online","group":{"name":"ArticleHistory","label":"Article History"}},{"order":1,"name":"Ethics","group":{"name":"EthicsHeading","label":"Declarations"}},{"value":"This work uses deidentified documents from the publicly available N2C2 NLP Research Data Sets. The use of the notes for the dysmenorrhea task dataset was approved by the institutional review board at the Icahn School of Medicine at Mount Sinai (IRB-HRP-503).","order":2,"name":"Ethics","group":{"name":"EthicsHeading","label":"Ethics approvals and consent to participate"}},{"value":"Not applicable","order":3,"name":"Ethics","group":{"name":"EthicsHeading","label":"Consent for publication"}},{"value":"The authors declare no competing interests.","order":4,"name":"Ethics","group":{"name":"EthicsHeading","label":"Competing interests"}}],"article-number":"247"}}