Scholarlink

<p>Large language models (LLMs) show promise for supporting clinical decision‑making in perioperative and intensive care settings. The recent study by Xu et al. on pre‑trained language models for preoperative anesthesia triage demonstrates that such models can effectively integrate structured and unstructured clinical data to support triage decisions. However, the translation of these tools from research prototypes to routine clinical use requires more than technical validation; it demands explicit, operationalised “instructions for use” analogous to those required for pharmaceuticals and medical devices. We argue that responsible deployment of LLMs in ICU and perioperative workflows must clarify: (1) intended clinical scope and non‑indications; (2) role in the decision‑making hierarchy and when clinicians should override model recommendations; and (3) mechanisms for transparency, governance, and staff training. Drawing on Xu et al.‘s methodological rigor and Bignami et al.‘s AI policy checklist framework, we outline a concise, practice‑oriented approach to embedding LLMs safely in critical care. We emphasise that without explicit instructions for use, clear governance structures, and comprehensive training, there is a risk of introducing inscrutable systems into the heart of critical care. The time to define these safeguards is now, before ad hoc, ungoverned adoption becomes the norm.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

From promising prototypes to “instructions for use”: embedding LLMs safely in perioperative and intensive care

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年2月4日

查看全文
  • Elena Giovanna Bignami,
  • Michele Russo

摘要

Large language models (LLMs) show promise for supporting clinical decision‑making in perioperative and intensive care settings. The recent study by Xu et al. on pre‑trained language models for preoperative anesthesia triage demonstrates that such models can effectively integrate structured and unstructured clinical data to support triage decisions. However, the translation of these tools from research prototypes to routine clinical use requires more than technical validation; it demands explicit, operationalised “instructions for use” analogous to those required for pharmaceuticals and medical devices. We argue that responsible deployment of LLMs in ICU and perioperative workflows must clarify: (1) intended clinical scope and non‑indications; (2) role in the decision‑making hierarchy and when clinicians should override model recommendations; and (3) mechanisms for transparency, governance, and staff training. Drawing on Xu et al.‘s methodological rigor and Bignami et al.‘s AI policy checklist framework, we outline a concise, practice‑oriented approach to embedding LLMs safely in critical care. We emphasise that without explicit instructions for use, clear governance structures, and comprehensive training, there is a risk of introducing inscrutable systems into the heart of critical care. The time to define these safeguards is now, before ad hoc, ungoverned adoption becomes the norm.