{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,7,27]],"date-time":"2026-07-27T01:21:01Z","timestamp":1785115261533,"version":"3.55.0"},"reference-count":30,"publisher":"Oxford University Press (OUP)","license":[{"start":{"date-parts":[[2026,7,26]],"date-time":"2026-07-26T00:00:00Z","timestamp":1785024000000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/academic.oup.com\/pages\/standard-publication-reuse-rights"}],"funder":[{"DOI":"10.13039\/100000002","name":"National Institutes of Health","doi-asserted-by":"publisher","award":["#1R21HL184302-01"],"award-info":[{"award-number":["#1R21HL184302-01"]}],"id":[{"id":"10.13039\/100000002","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"abstract":"<jats:title>Abstract<\/jats:title>\n                  <jats:sec>\n                    <jats:title>Introduction<\/jats:title>\n                    <jats:p>In future large-scale combat operations, military medics will likely face staggering numbers of casualties in the face of resource constrained evacuation and forward medical care. Artificial intelligence (AI) may enable triage at scale and improve operational efficiencies. Although many assume that sophisticated large language models (LLMs) hold the most promise, simpler machine learning models (MLMs) may offer advantages in edge implementation because of lower computational burden. Here we search for computational efficiency\u2014achieving equivalent clinical performance with computationally simpler AI\u2014in the domain of on-scene triage.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Materials and Methods<\/jats:title>\n                    <jats:p>We conducted a retrospective study via a civilian critical care air transport service from 2012 to 2021 (approved by the University of Pittsburgh IRB). Free-text impressions recorded by treating clinicians and records of patient lifesaving intervention (LSI; e.g., airway management; blood transfusion) were abstracted from the electronic health record. Two modeling approaches were used to predict LSI receipt from medic impression text. MLMs involved using an ensemble classifier model fed with word frequencies. LLMs involved using 2 transformer models (DistilBERT and BioBERT) fed with tokenized word embeddings.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Results<\/jats:title>\n                    <jats:p>A total of 12,913 patients were included (mean age\u2009=\u200952.3\u2009years, 63% men). MLMs and LLMs produced broadly equivalent performance in predicting LSI (MLM: AUROC = .793 [.776, .810]; AP = .670 [.643, .695] vs. BioBERT: AUROC = .803 [.796, .816]; AP = .674 [.649, .698]) vs. DistilBERT: AUROC = .803 [.785, .811]; AP = .673 [.651, .688]). Broad predictive equipoise held across 7 independent LSI categories. Yet LLMs required considerably greater computational resources: Model training took 9.42 seconds for MLMs, 3 minutes 57 seconds for DistilBERT, and 7 minutes 14 seconds for BioBERT, whereas model prediction took .048 seconds for MLMs, 7.37 seconds for DistilBERT, and 13.36 seconds for BioBERT.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Conclusions<\/jats:title>\n                    <jats:p>Our findings counter the \u201cbigger is better\u201d paradigm in medical AI: MLMs performed equivalently to LLMs in task mirroring on-scene triage while requiring a fraction of the computational resources. Nimble MLMs offer several advantages for military AI, including seamless deployment on edge computing devices and explainable decision support for combat medics. As AI technologies expand across warfighting domains, considerations of computational efficiency are critical to adopt and field new medical capabilities in the forward environment.<\/jats:p>\n                  <\/jats:sec>","DOI":"10.1093\/milmed\/usag325","type":"journal-article","created":{"date-parts":[[2026,7,26]],"date-time":"2026-07-26T13:51:48Z","timestamp":1785073908000},"source":"Crossref","is-referenced-by-count":1,"title":["Bigger Is Not Always Better: Computational Efficiency in Lexical Prehospital Triage Modeling"],"prefix":"10.1093","author":[{"ORCID":"https:\/\/orcid.org\/0000-0002-9395-4051","authenticated-orcid":false,"given":"Aaron C","family":"Weidman","sequence":"first","affiliation":[{"name":"Department of Emergency Medicine, School of Medicine, University of Pittsburgh , Pittsburgh, PA 15213,","place":["United States"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Chase","family":"Zikmund","sequence":"additional","affiliation":[{"name":"Department of Emergency Medicine, School of Medicine, University of Pittsburgh , Pittsburgh, PA 15213,","place":["United States"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Leonard S","family":"Weiss","sequence":"additional","affiliation":[{"name":"Department of Emergency Medicine, School of Medicine, University of Pittsburgh , Pittsburgh, PA 15213,","place":["United States"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Francis X","family":"Guyette","sequence":"additional","affiliation":[{"name":"Department of Emergency Medicine, School of Medicine, University of Pittsburgh , Pittsburgh, PA 15213,","place":["United States"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Ronald K","family":"Poropatich","sequence":"additional","affiliation":[{"name":"Center for Military Medicine Research, School of Medicine, University of Pittsburgh , Pittsburgh, PA 15213,","place":["United States"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"David D","family":"Salcido","sequence":"additional","affiliation":[{"name":"Department of Emergency Medicine, School of Medicine, University of Pittsburgh , Pittsburgh, PA 15213,","place":["United States"]}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"286","published-online":{"date-parts":[[2026,7,26]]},"reference":[{"key":"2026072609514442400_usag325-B1","doi-asserted-by":"crossref","first-page":"199","DOI":"10.1093\/milmed\/usae428","article-title":"Artificial intelligence in the US military health system: forging a new frontier for clinical care and efficiency","volume":"190","author":"Adirim","year":"2025","journal-title":"Mil Med"},{"key":"2026072609514442400_usag325-B2","doi-asserted-by":"crossref","first-page":"81","DOI":"10.1093\/milmed\/usae377","article-title":"Fully autonomous casualty care on the future battlefield","volume":"190","author":"Pamplin","year":"2025","journal-title":"Mil Med"},{"key":"2026072609514442400_usag325-B3","article-title":"Automating the survival chain and revolutionizing combat casualty care","author":"Talley","year":"2024","journal-title":"Mil Rev"},{"key":"2026072609514442400_usag325-B4","first-page":"usaf165","article-title":"AI on the front lines: a primer for the military health professional","volume":"5","author":"Riggenbach","year":"2025","journal-title":"Mil Med"},{"key":"2026072609514442400_usag325-B5","doi-asserted-by":"crossref","first-page":"257","DOI":"10.1093\/milmed\/usaf010","article-title":"Artificial intelligence decision support systems in resource-limited environments to save lives and reduce moral injury","volume":"190","author":"Umlauf","year":"2025","journal-title":"Mil Med"},{"key":"2026072609514442400_usag325-B6","doi-asserted-by":"crossref","first-page":"e2339\u201349","DOI":"10.1093\/milmed\/usaf189","article-title":"The medical implications of emerging unmanned aircraft systems in military and combat environments: a narrative review","volume":"190","author":"Nowadly","year":"2025","journal-title":"Mil Med"},{"key":"2026072609514442400_usag325-B7","first-page":"usae569","article-title":"The expanding role of the US army paramedic","volume":"11","author":"Greener","year":"2025","journal-title":"Mil Med"},{"key":"2026072609514442400_usag325-B8","doi-asserted-by":"crossref","first-page":"S180","DOI":"10.1097\/TA.0000000000004063","article-title":"Casualty care implications of large-scale combat operations","volume":"95","author":"Remondelli","year":"2023","journal-title":"J Trauma Acute Care Surg"},{"key":"2026072609514442400_usag325-B9","doi-asserted-by":"crossref","first-page":"1052","DOI":"10.1093\/milmed\/usaf079","article-title":"An observational study of combat medics\u2019 tactical combat casualty care medication administration through a single-center simulated trauma","volume":"190","author":"Reynhout","year":"2025","journal-title":"Mil Med"},{"key":"2026072609514442400_usag325-B10","doi-asserted-by":"crossref","first-page":"e37\u201341","DOI":"10.1093\/milmed\/usab525","article-title":"Assessing baseline proficiency of national guard combat medics due for recertification: a pilot study","volume":"188","author":"Nichols","year":"2023","journal-title":"Mil Med"},{"key":"2026072609514442400_usag325-B11","doi-asserted-by":"crossref","first-page":"242","DOI":"10.1001\/jama.2023.25057","article-title":"Three epochs of artificial intelligence in health care","volume":"331","author":"Howell","year":"2024","journal-title":"JAMA"},{"key":"2026072609514442400_usag325-B12","doi-asserted-by":"crossref","first-page":"AIp2500154","DOI":"10.1056\/AIp2500154","article-title":"Energy considerations for scaling artificial intelligence adoption in medicine: first do no harm","volume":"2","author":"Malhotra","year":"2025","journal-title":"NEJM AI"},{"key":"2026072609514442400_usag325-B13","first-page":"61","author":"Varoquaux"},{"key":"2026072609514442400_usag325-B14","doi-asserted-by":"crossref","first-page":"6","DOI":"10.3109\/10903127.2015.1025157","article-title":"Prehospital trauma triage decision-making: a model of what happens between the 9-1-1 call and the hospital","volume":"20","author":"Jones","year":"2016","journal-title":"Prehosp Emerg Care"},{"key":"2026072609514442400_usag325-B15","doi-asserted-by":"crossref","first-page":"38","DOI":"10.1080\/10903127.2022.2043963","article-title":"Under-triage and over-triage using the field triage guidelines for injured patients: a systematic review","volume":"27","author":"Lupton","year":"2023","journal-title":"Prehosp Emerg Care"},{"key":"2026072609514442400_usag325-B16","doi-asserted-by":"crossref","first-page":"516","DOI":"10.1111\/acem.15067","article-title":"Words to live by: using medic impressions to identify the need for prehospital lifesaving interventions","volume":"32","author":"Weidman","year":"2025","journal-title":"Acad Emerg Med"},{"key":"2026072609514442400_usag325-B17","doi-asserted-by":"crossref","first-page":"1234","DOI":"10.1093\/bioinformatics\/btz682","article-title":"BioBERT: a pre-trained biomedical language representation model for biomedical text mining","volume":"36","author":"Lee","year":"2020","journal-title":"Bioinformatics"},{"key":"2026072609514442400_usag325-B18","author":"Sanh","year":"2019"},{"key":"2026072609514442400_usag325-B19","doi-asserted-by":"crossref","first-page":"e048008","DOI":"10.1136\/bmjopen-2020-048008","article-title":"Protocol for development of a reporting guideline (TRIPOD-AI) and risk of bias tool (PROBAST-AI) for diagnostic and prognostic prediction model studies based on artificial intelligence","volume":"11","author":"Collins","year":"2021","journal-title":"BMJ Open"},{"key":"2026072609514442400_usag325-B20","volume-title":"Natural Language Processing with Python: Analyzing Text with the Natural Language Toolkit","author":"Bird","year":"2009"},{"key":"2026072609514442400_usag325-B21","doi-asserted-by":"crossref","first-page":"7","DOI":"10.7205\/MILMED.170.1.7","article-title":"Prehospital physiologic data and lifesaving interventions in trauma patients","volume":"170","author":"Holcomb","year":"2005","journal-title":"Mil Med"},{"key":"2026072609514442400_usag325-B22","first-page":"2825","article-title":"Scikit-learn: machine learning in Python","volume":"12","author":"Pedregosa","year":"2011","journal-title":"J Mach Learn Res"},{"key":"2026072609514442400_usag325-B23","first-page":"111","article-title":"Evidence based emergency medicine; part 5 receiver operating curve and area under the curve","volume":"4","author":"Safari","year":"2016","journal-title":"Emergency"},{"key":"2026072609514442400_usag325-B24","doi-asserted-by":"crossref","first-page":"131","DOI":"10.1177\/1536867X20909693","article-title":"When to consult precision-recall curves","volume":"20","author":"Cook","year":"2020","journal-title":"Stata J"},{"key":"2026072609514442400_usag325-B25","doi-asserted-by":"crossref","DOI":"10.1007\/978-3-319-24277-4","volume-title":"ggplot2: Elegant Graphics for Data Analysis","author":"Wickham","year":"2016"},{"key":"2026072609514442400_usag325-B26","doi-asserted-by":"crossref","first-page":"307","DOI":"10.1016\/S0196-0644(97)70166-5","article-title":"What is confidence? Part 1: the use and interpretation of confidence intervals","volume":"30","author":"Young","year":"1997","journal-title":"Ann Emerg Med"},{"key":"2026072609514442400_usag325-B27","doi-asserted-by":"crossref","first-page":"386","DOI":"10.1111\/acem.15099","article-title":"Artificial intelligence\u2013based clinical decision support in the emergency department: a scoping review","volume":"32","author":"Kareemi","year":"2025","journal-title":"Acad Emerg Med"},{"key":"2026072609514442400_usag325-B28","doi-asserted-by":"crossref","first-page":"ooaf147","DOI":"10.1093\/jamiaopen\/ooaf147","article-title":"Benchmarking speech-to-text robust\u00adness in noisy emergency medical dialogues: an evaluation of models under realistic acoustic conditions","volume":"8","author":"Moser","year":"2025","journal-title":"JAMIA Open"},{"key":"2026072609514442400_usag325-B29","doi-asserted-by":"crossref","first-page":"321","DOI":"10.1613\/jair.953","article-title":"SMOTE: synthetic minority over-sampling technique","volume":"16","author":"Chawla","year":"2002","journal-title":"J Artif Intell Res"},{"key":"2026072609514442400_usag325-B30","doi-asserted-by":"crossref","first-page":"e259639","DOI":"10.1001\/jamanetworkopen.2025.9639","article-title":"A machine learning trauma triage model for critical care transport","volume":"8","author":"Weidman","year":"2025","journal-title":"JAMA Netw Open"}],"container-title":["Military Medicine"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/academic.oup.com\/milmed\/advance-article-pdf\/doi\/10.1093\/milmed\/usag325\/70101143\/usag325.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"syndication"},{"URL":"https:\/\/academic.oup.com\/milmed\/advance-article-pdf\/doi\/10.1093\/milmed\/usag325\/70101143\/usag325.pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,7,26]],"date-time":"2026-07-26T13:51:53Z","timestamp":1785073913000},"score":1,"resource":{"primary":{"URL":"https:\/\/academic.oup.com\/milmed\/advance-article\/doi\/10.1093\/milmed\/usag325\/8742086"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,7,26]]},"references-count":30,"URL":"https:\/\/doi.org\/10.1093\/milmed\/usag325","relation":{},"ISSN":["0026-4075","1930-613X"],"issn-type":[{"value":"0026-4075","type":"print"},{"value":"1930-613X","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,7,26]]},"article-number":"usag325"}}