{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,8,10]],"date-time":"2026-08-10T09:30:29Z","timestamp":1786354229417,"version":"build-2736575974"},"reference-count":31,"publisher":"European Respiratory Society (ERS)","issue":"4","license":[{"start":{"date-parts":[[2026,5,28]],"date-time":"2026-05-28T00:00:00Z","timestamp":1779926400000},"content-version":"vor","delay-in-days":0,"URL":"http:\/\/creativecommons.org\/licenses\/by-nc\/4.0\/"}],"content-domain":{"domain":["publications.ersnet.org"],"crossmark-restriction":true},"short-container-title":["ERJ Open Res"],"accepted":{"date-parts":[[2026,2,7]]},"published-print":{"date-parts":[[2026,7]]},"abstract":"<jats:sec>\n                    <jats:title>Background<\/jats:title>\n                    <jats:p>Large language models (LLMs) are rapidly entering respiratory medicine workflows. Their clinical role remains unclear. A central concern is whether they function as autonomous decision-makers or as tools that depend on clinician input.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Methods<\/jats:title>\n                    <jats:p>We conducted a structured narrative review across three domains aligned with respiratory practice, including bedside clinical decision support and triage, pulmonary function test (PFT)\/spirometry interpretation, and chest radiograph (CXR) reporting. We included quantitative studies comparing LLMs or specialised artificial intelligence (AI) systems with clinician performance and extracted accuracy, agreement and safety-critical failure modes.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Results<\/jats:title>\n                    <jats:p>At bedside, LLMs approached expert diagnosis in straightforward cases, but performance fell in highly complex or multi-step management tasks. The dominant error was omission of required actions, unsafe under-triage or incomplete plans. In PFTs, guideline-anchored and spirogram-trained LLMs achieved high agreement and reduced inter-observer variability on structured inputs. In contrast, generalist LLMs showed only moderate reliability and inconsistent handling of borderline or mixed patterns. In CXR reporting, chest-specific multimodal systems outperformed generalist LLMs, with fewer hallucinations and better clinical acceptability; yet clinically relevant discrepancies persisted without radiologist review.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Conclusions<\/jats:title>\n                    <jats:p>Across domains, AI does not replace respiratory expertise. When inputs are complete, structured and guideline-concordant, outputs stabilise and add value. When inputs are ambiguous or incomplete, errors expand, often through dangerous omissions. Safe deployment therefore requires specialised training, strict input structure and continuous clinician oversight.<\/jats:p>\n                  <\/jats:sec>","DOI":"10.1183\/23120541.01699-2025","type":"journal-article","created":{"date-parts":[[2026,5,28]],"date-time":"2026-05-28T15:15:11Z","timestamp":1779981311000},"page":"01699-2025","update-policy":"https:\/\/doi.org\/10.1183\/ers-crossmark-policy","source":"Crossref","is-referenced-by-count":0,"title":["Large language models in respiratory care: safety and clinical integration"],"prefix":"10.1183","volume":"12","author":[{"given":"Alessandro","family":"Porcella","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0009-0006-6070-428X","authenticated-orcid":false,"given":"Antonio","family":"Fabozzi","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Matteo","family":"Bonini","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Paolo","family":"Palange","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"81","published-online":{"date-parts":[[2026,5,28]]},"reference":[{"key":"2026081001451670000_12.4.01699-2025.1","doi-asserted-by":"crossref","first-page":"110351","DOI":"10.1016\/j.compbiomed.2025.110351","article-title":"Evaluating large language and large reasoning models as decision support tools in emergency internal medicine","volume":"192","author":"Vrdoljak","year":"2025","journal-title":"Comput Biol Med"},{"key":"2026081001451670000_12.4.01699-2025.2","doi-asserted-by":"crossref","first-page":"7866","DOI":"10.1038\/s41467-025-64769-1","article-title":"Quantifying the reasoning abilities of large language models on real-world clinical cases (MedR-Bench)","volume":"16","author":"Qiu","year":"2025","journal-title":"Nat Commun"},{"key":"2026081001451670000_12.4.01699-2025.3","doi-asserted-by":"crossref","first-page":"22106","DOI":"10.1038\/s41598-024-73229-7","article-title":"Emergency department triaging using ChatGPT based on emergency severity index principles: a cross-sectional study","volume":"14","author":"Colakca","year":"2024","journal-title":"Sci Rep"},{"key":"2026081001451670000_12.4.01699-2025.4","doi-asserted-by":"crossref","first-page":"165","DOI":"10.1016\/j.ajem.2025.03.028","article-title":"Chat-GPT in triage: still far from surpassing human expertise \u2013 an observational study","volume":"92","author":"Zaboli","year":"2025","journal-title":"Am J Emerg Med"},{"key":"2026081001451670000_12.4.01699-2025.5","doi-asserted-by":"publisher","DOI":"10.1183\/13993003.01660-2018"},{"key":"2026081001451670000_12.4.01699-2025.6","doi-asserted-by":"crossref","first-page":"e0297519","DOI":"10.1371\/journal.pone.0297519","article-title":"Development and evaluation of a computerized algorithm for the interpretation of pulmonary function tests","volume":"19","author":"Huang","year":"2024","journal-title":"PLoS One"},{"key":"2026081001451670000_12.4.01699-2025.7","doi-asserted-by":"crossref","first-page":"7057","DOI":"10.3390\/jcm13237057","article-title":"Comparative analysis of M4CXR, an LLM-based chest X-ray report generation model, and ChatGPT in radiological interpretation","volume":"13","author":"Lee","year":"2024","journal-title":"J Clin Med"},{"key":"2026081001451670000_12.4.01699-2025.8","doi-asserted-by":"publisher","DOI":"10.1038\/s41591-024-03302-1"},{"key":"2026081001451670000_12.4.01699-2025.9","doi-asserted-by":"crossref","first-page":"225","DOI":"10.1161\/HYPERTENSIONAHA.125.25492","article-title":"Performance of large language models in analyzing common hypertension scenarios","volume":"83","author":"Zand","year":"2026","journal-title":"Hypertension"},{"key":"2026081001451670000_12.4.01699-2025.10","doi-asserted-by":"publisher","DOI":"10.1038\/s41386-024-01841-2"},{"key":"2026081001451670000_12.4.01699-2025.11","doi-asserted-by":"crossref","first-page":"2043","DOI":"10.1007\/s00296-024-05675-5","article-title":"Vignette-based comparative analysis of ChatGPT and specialist treatment decisions for rheumatic patients: results of the Rheum2Guide study","volume":"44","author":"Labinsky","year":"2024","journal-title":"Rheumatol Int"},{"key":"2026081001451670000_12.4.01699-2025.12","first-page":"100580","article-title":"Evaluation of GPT-4 concordance with North American Spine Society guidelines for lumbar fusion surgery","volume":"21","author":"Khoylyan","year":"2025","journal-title":"North Am Spine Soc J"},{"key":"2026081001451670000_12.4.01699-2025.13","doi-asserted-by":"crossref","first-page":"15439","DOI":"10.1038\/s41598-025-99774-3","article-title":"GPT-4's performance in supporting physician decision-making in nephrology multiple-choice questions","volume":"15","author":"Noda","year":"2025","journal-title":"Sci Rep"},{"key":"2026081001451670000_12.4.01699-2025.14","doi-asserted-by":"crossref","first-page":"e59267","DOI":"10.2196\/59267","article-title":"Evaluating ChatGPT-4's accuracy in identifying final diagnoses within differential diagnoses compared with those of physicians","volume":"8","author":"Hirosawa","year":"2024","journal-title":"JMIR Form Res"},{"key":"2026081001451670000_12.4.01699-2025.15","doi-asserted-by":"publisher","DOI":"10.1183\/23120541.00116-2025"},{"key":"2026081001451670000_12.4.01699-2025.16","first-page":"AIoa2400804","article-title":"AI-assisted spirometry interpretation in primary care: a randomized controlled trial","volume":"2","author":"Doe","year":"2025","journal-title":"NEJM AI"},{"key":"2026081001451670000_12.4.01699-2025.17","first-page":"e82056","article-title":"Application of artificial intelligence in the interpretation of pulmonary function tests","volume":"17","author":"Saad","year":"2025","journal-title":"Cureus"},{"key":"2026081001451670000_12.4.01699-2025.18","first-page":"535","article-title":"Enhancing pulmonary function test reporting with artificial intelligence: a retrospective observational study","volume":"32","author":"Pandiarajan","year":"2025","journal-title":"TPM"},{"key":"2026081001451670000_12.4.01699-2025.19","first-page":"813","article-title":"Accuracy of spirometric interpretation by artificial intelligence-based software in comparison with a qualified respiratory physician in a tertiary care center in Chengalpattu District","volume":"14","author":"Rohita","year":"2025","journal-title":"J Neonatal Surg"},{"key":"2026081001451670000_12.4.01699-2025.20","doi-asserted-by":"crossref","unstructured":"Park JH , Cheong C , Kang S , et al. Evaluating advanced large language models for pulmonary disease diagnosis using portable spirometer data: a comparative analysis of Gemini 1.5 Pro, GPT-4o, and Claude 3.5 Sonnet. In: Proceedings of the IEEE-EMBS International Conference on Biomedical and Health Informatics (BHI 2024). Houston, IEEE, 2024. doi:10.1109\/BHI62660.2024.10913702","DOI":"10.1109\/BHI62660.2024.10913702"},{"key":"2026081001451670000_12.4.01699-2025.21","doi-asserted-by":"crossref","first-page":"e0001300","DOI":"10.1371\/journal.pdig.0001300","article-title":"SpiroLLM: finetuning pretrained large language models to understand spirogram time series with clinical validation in COPD reporting","volume":"5","author":"Mei","year":"2026","journal-title":"PLOS Digit Health"},{"key":"2026081001451670000_12.4.01699-2025.22","first-page":"784","article-title":"Artificial intelligence: a promising tool in diagnosis of respiratory diseases","volume":"2","author":"Yadav","year":"2024","journal-title":"Intell Pharm"},{"key":"2026081001451670000_12.4.01699-2025.23","doi-asserted-by":"crossref","first-page":"4374","DOI":"10.1007\/s00330-024-11339-6","article-title":"CXR-LLaVA: a multimodal large language model for interpreting chest X-ray images","volume":"35","author":"Lee","year":"2025","journal-title":"Eur Radiol"},{"key":"2026081001451670000_12.4.01699-2025.24","doi-asserted-by":"crossref","first-page":"1435","DOI":"10.1186\/s12880-024-01435-w","article-title":"An open-source fine-tuned large language model for radiology report impression summarization: multi-institutional evaluation and reader study","volume":"24","author":"Serapio","year":"2024","journal-title":"BMC Med Imaging"},{"key":"2026081001451670000_12.4.01699-2025.25","doi-asserted-by":"crossref","first-page":"e2336100","DOI":"10.1001\/jamanetworkopen.2023.36100","article-title":"Generative artificial intelligence for chest radiograph interpretation in the emergency department","volume":"6","author":"Huang","year":"2023","journal-title":"JAMA Netw Open"},{"key":"2026081001451670000_12.4.01699-2025.26","first-page":"e240091","article-title":"Large language models in radiology reporting: systematic review of performance, safety, and clinical integration","volume":"7","author":"Artsi","year":"2025","journal-title":"Radiol Artif Intell"},{"key":"2026081001451670000_12.4.01699-2025.27","unstructured":"Ovadia Y , Fertig E , Ren J , et al. Can you trust your model's uncertainty? Evaluating predictive uncertainty under dataset shift. In: Wallach H , Larochelle H , Beygelzimer A , et al. , eds. Advances in Neural Information Processing Systems 32 (NeurIPS 2019). New York, Curran Associates, Inc., 2019; pp. 13991\u201314002.10.5555\/3454287.34555411"},{"key":"2026081001451670000_12.4.01699-2025.28","doi-asserted-by":"crossref","first-page":"4","DOI":"10.1038\/s41746-020-00367-3","article-title":"Second opinion needed: communicating uncertainty in medical machine learning","volume":"4","author":"Kompa","year":"2021","journal-title":"NPJ Digit Med"},{"key":"2026081001451670000_12.4.01699-2025.29","doi-asserted-by":"crossref","first-page":"139","DOI":"10.1093\/jamia\/ocae254","article-title":"Large language model uncertainty proxies: discrimination and calibration for medical diagnosis and treatment","volume":"32","author":"Savage","year":"2025","journal-title":"J Am Med Inform Assoc"},{"key":"2026081001451670000_12.4.01699-2025.30","doi-asserted-by":"crossref","first-page":"5635","DOI":"10.1007\/s11263-024-02117-4","article-title":"Generalized out-of-distribution detection: a survey","volume":"132","author":"Yang","year":"2024","journal-title":"Int J Comput Vis"},{"key":"2026081001451670000_12.4.01699-2025.31","unstructured":"Kaur R , Jha S , Roy A , et al. Detecting OODs as datapoints with high uncertainty. arXiv 2021; preprint [https:\/\/doi.org\/10.48550\/arXiv.2108.06380]"}],"container-title":["ERJ Open Research"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/syndication.highwire.org\/content\/doi\/10.1183\/23120541.01699-2025","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,8,10]],"date-time":"2026-08-10T08:45:25Z","timestamp":1786351525000},"score":1,"resource":{"primary":{"URL":"https:\/\/publications.ersnet.org\/lookup\/doi\/10.1183\/23120541.01699-2025"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,5,28]]},"references-count":31,"journal-issue":{"issue":"4","published-online":{"date-parts":[[2026,8,10]]},"published-print":{"date-parts":[[2026,7]]}},"alternative-id":["10.1183\/23120541.01699-2025"],"URL":"https:\/\/doi.org\/10.1183\/23120541.01699-2025","relation":{},"ISSN":["2312-0541"],"issn-type":[{"value":"2312-0541","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,5,28]]}}}