TRUSTGUIDES:评估卫生指南推荐意见可信度的工具开发
卫生指南推荐意见的可信度可能因方法学不严谨、缺乏透明度、利益冲突及标准应用不一致而受损。本研究旨在通过多步骤混合方法开发并验证TRUSTGUIDES工具,以评估单条指南推荐意见的可信度,涵盖透明度、严谨性、独立性、适用性及人工智能的适当使用等维度。
要点
- TRUSTGUIDES评估单条推荐意见的可信度,而非仅整体指南质量。
- 可信度涵盖严谨方法、透明度、独立性和适用性,区别于方法学质量。
- 工具开发将整合范围综述、德尔菲调查及心理测量学验证。
核心结论
- TRUSTGUIDES旨在全面评估单条指南推荐意见的可信度,弥补现有工具仅关注部分质量维度的不足。
研究设计与样本
- 采用多步骤混合方法开发工具,包括范围综述、专家咨询、焦点小组和国际多学科工作组参与的迭代德尔菲调查。
- 通过评估内部一致性、评定者间信度、内容效度和结构效度进行验证,并与AGREE II、PANELVIEW等工具比较。
- 心理测量学特性将用因子分析及必要时的项目反应理论模型检验。
关键数值与推荐
- 可信度涵盖严谨方法、透明度、独立性和适用性,决定推荐意见是否值得用户信任。
- 评估领域包括证据确定性、推荐强度、利益冲突、适用性、适应性、时效性、认证及人工智能的适当使用。
临床怎么用
- 临床医生和指南使用者可依据TRUSTGUIDES评估单条推荐意见的可信度,辅助临床决策。
- 工具支持对活指南、GRADE、改编及人工智能应用等新方法学进展的评估。
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
原文
BACKGROUND AND OBJECTIVES: Health guidelines play a central role in informing clinical practice, public health measures and health policy. But their trustworthiness may be undermined by factors such as insufficient methodological rigor, lack of transparency, conflicts of interest, and inconsistent application of established standards. Existing appraisal tools address selected aspects of guideline quality but do not comprehensively assess the trustworthiness of individual recommendations, nor do they adequately reflect recent advances in guideline methodology, including living guidelines, Grading of Recommendations, Assessment, Development, and Evaluation, adaptation, and the use of artificial intelligence (AI). This study aims to develop and validate Transparent, Rigorous, Useable, Standardized, and Trustworthy Guide (TRUSTGUIDES), a globally applicable, flexible set of tools to assess the trustworthiness of health guideline recommendations. We define trustworthiness as distinct from methodological quality: it encompasses not only rigorous methods but also transparency, independence, and applicability, which together determine whether a recommendation merits user confidence. METHODS: TRUSTGUIDES will be developed through a multistep, mixed-methods process. First, a scoping review and expert consultation will identify existing guideline appraisal tools and inform domains and items generation. Using deductive and inductive approaches, domains and items will be generated and may be refined through focus groups and selected through iterative Delphi surveys involving an international, multidisciplinary working group. TRUSTGUIDES will be validated by assessing internal consistency, inter-rater reliability, content validity, and construct validity, including comparisons with established instruments such as the Grading of Recommendations, Assessment, Development, and Evaluation certainty domains, AGREE II, and PANELVIEW. Psychometric properties will be examined using factor analysis and, as necessary, item response theory models. AI will be integrated both as an object of assessment and as methodological support for tool application, with large language models evaluated against a human reference standard. CONCLUSION: TRUSTGUIDES will be designed to evaluate the trustworthiness of individual guideline recommendations across key factors, including transparency and credibility, and to address relevant domains such as the certainty of evidence, strength of recommendations, conflicts of interest, applicability, adaptability, currency, certification, and the appropriate use of AI. TRUSTGUIDES addresses critical gaps in current guideline appraisal by offering a comprehensive, recommendation-level assessment of trustworthiness aligned with the World Health Organization guideline standard methodology. By integrating AI, our tools will support efficient, transparent, and future-ready guideline evaluation within an evolving health evidence ecosystem.
引用
- Arienti C, Neumann I, Akl EA, et al. Assessing the trustworthiness of health guidelines recommendations: the transparent, rigorous, useable, standardized, and trustworthy guide (TRUSTGUIDES) tools development. J Clin Epidemiol. 2026. PMID: 42203175. doi:10.1016/j.jclinepi.2026.112348
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。