Sounderajah Viknesh, Guni Ahmad, Liu Xiaoxuan, Collins Gary S, Karthikesalingam Alan, Markar Sheraz R, Golub Robert M, Denniston Alastair K, Shetty Shravya, Moher David, Bossuyt Patrick M, Darzi Ara, Ashrafian Hutan
Institute of Global Health Innovation, Imperial College London, London, UK.
Department of Surgery and Cancer, Imperial College London, London, UK.
Nat Med. 2025 Sep 15. doi: 10.1038/s41591-025-03953-8.
The Standards for Reporting Diagnostic Accuracy (STARD) 2015 statement facilitates transparent and complete reporting of diagnostic test accuracy studies. However, there are unique considerations associated with artificial intelligence (AI)-centered diagnostic test studies. The STARD-AI statement, which was developed through a multistage, multistakeholder process, provides a minimum set of criteria that allows for comprehensive reporting of AI-centered diagnostic test accuracy studies. The process involved a literature review, a scoping survey of international experts, and a patient and public involvement and engagement initiative, culminating in a modified Delphi consensus process involving over 240 international stakeholders and a consensus meeting. The checklist was subsequently finalized by the Steering Committee and includes 18 new or modified items in addition to the STARD 2015 checklist items. Authors are encouraged to provide descriptions of dataset practices, the AI index test and how it was evaluated, as well as considerations of algorithmic bias and fairness. The STARD-AI statement supports comprehensive and transparent reporting in all AI-centered diagnostic accuracy studies, and it can help key stakeholders to evaluate the biases, applicability and generalizability of study findings.
《2015年诊断准确性报告标准》(STARD)声明有助于对诊断试验准确性研究进行透明且完整的报告。然而,以人工智能(AI)为中心的诊断试验研究存在一些独特的考量因素。通过多阶段、多利益相关方流程制定的STARD-AI声明提供了一套最低标准,可实现对以AI为中心的诊断试验准确性研究的全面报告。该流程包括文献综述、对国际专家的范围界定调查、患者及公众参与倡议,最终形成了一个涉及240多名国际利益相关方的改良德尔菲共识流程以及一次共识会议。该清单随后由指导委员会最终确定,除了《2015年STARD清单》项目外,还包括18项新的或修改后的项目。鼓励作者提供数据集实践、AI索引测试及其评估方式的描述,以及算法偏差和公平性的考量因素。STARD-AI声明支持对所有以AI为中心的诊断准确性研究进行全面且透明的报告,并且有助于关键利益相关方评估研究结果的偏差、适用性和普遍性。