Medical guidelines are constantly going out of date as new studies are published, but this group will build automated systems that keep the evidence continuously updated. The problem is that traditional systematic reviews—the gold standard for summarising research—are static snapshots that quickly become obsolete, especially on fast-moving topics like new drugs or treatments. The Bristol-UCL-King’s Living Evidence Synthesis Group will use machine learning tools, including large language models, to monitor new research as it appears, automatically flag relevant studies, and integrate findings into living reviews that stay current. They will also develop living network meta-analyses, which compare multiple treatment options at once and update recommendations as new interventions emerge. If successful, this could transform how health policymakers and clinicians get reliable evidence—instead of waiting years for an updated review, they would have a constantly refreshed, accessible summary of the best available science. The group will also evaluate their automation tools through embedded studies, ensuring the technology works reliably before recommending it to other evidence synthesis teams.
View original technical description
Living evidence syntheses allow accumulating evidence on important, fast-moving topics to be monitored continually, with findings rapidly synthesized and presented in accessible formats. A collaboration between the University of Bristol (UoB) and University College London (UCL) wishes to host the NIHR Living Evidence Synthesis Group (LESG), to develop and maintain a portfolio of living systematic evidence syntheses on priority research questions. With teams led by Prof Julian Higgins at UoB and Prof James Thomas at UCL, and in collaboration with Dr Iain Marshall (King’s College London) and Dr Andy Gibson (UWE, Bristol), the Bristol-UCL-King’s Living Evidence Synthesis (BUcKLES) Group will use state-of-the-art methods and technologies to deliver policy-relevant and timely living evidence products for NIHR’s clients. The Bristol and UCL teams both have extensive experience of conducting living evidence syntheses. Co-directors Higgins and Thomas co-authored the founding paper on the notion of living systematic reviews and have subsequently participated extensively in networks and methodological research. We have assembled an international advisory board of leaders in living evidence synthesis to guide our LESG’s work. Through our collaborators and strong networks, we will identify content expertise to support us in each synthesis. Both teams also have established and highly successful infrastructures for co-production of reviews with service users and other stakeholders, which will be bolstered by inclusion of Bristol’s regional patient and public involvement (PPI) lead, Andy Gibson, as PPI lead for the LESG. We will form a public advisory group for each living evidence synthesis we undertake, formed from public contributors with service use experience relevant to the topic of the synthesis. Our commitment to equality, diversity and inclusion informs our working practices and our research, and we will work hard to ensure representation from underserved communities on these advisory groups. We propose that living network meta-analyses are major candidates for living evidence syntheses, so that recommendations on which intervention options are most suitable can be continually updated as new interventions arise and new evidence emerges. The Bristol team has specialist expertise in network meta-analysis and related techniques such as component-based network meta-analyses for complex interventions, and we are keen to work with the NIHR Evidence Synthesis Programme to develop a programme of these. We also propose that efficient and user-friendly production and updating workflows are critical for the sustainability of living evidence synthesis. Living evidence synthesis may particularly benefit from the use of automation tools, facilitating the rapid integration of new evidence into syntheses, conclusions and subsequent guidelines. The UCL team has developed continual evidence surveillance systems using large language models and other state-of-the-art machine learning tools, and we will develop this work with Iain Marshall, co-lead of the RobotReviewer project. Our use of automation tools will be underpinned by an extensive programme of studies within reviews (SWARs) designed to evaluate their performance in our semi-automated workflows. We will deploy those with acceptable performance and encourage evidence synthesis teams in other Evidence Synthesis Groups (ESGs) and elsewhere to adopt them.
Plain English summaries and category classifications on this site are generated by AI and may not perfectly reflect the original research.
Is something wrong? Let us know