Event

Systematic literature reviews (SLRs) supply the evidence base for HTA submissions but can be slow and resource intensive to undertake. Efforts to accelerate SLRs have been underway for years, with AI offering fresh opportunities for efficiency. Recently, machine learning has proven adept at title/abstract screening, a tractable classification task. However, other SLR activities, such as constructing and validating a search strategy or extracting structured data from long, inconsistently reported documents, present much greater complexity, yet AI can be applied to these tasks, too.

In this webinar, we will consider how AI might be applied across major SLR phases to streamline the process while maintaining methodological rigor, such that HTA acceptance is assured. First, we will address AI-based search and its performance versus manual search. Next, we will turn to screening, data extraction, and critical appraisal, where published research and associated claims regarding AI performance have been most extensive. We will look to the literature to compare AI’s actual performance against the hype, and we will assess the resource impact created by the need to configure, train, and validate a tool before use.

The webinar will close with a discussion of the emerging expectations of HTA bodies, initiatives such as RAISE that support the considered uptake of AI in evidence synthesis, and the role of human oversight in building a robust SLR that stands up to payer and regulatory scrutiny. A case study will be presented detailing how AstraZeneca employed AI to prepare SLRs for a recent NICE submission, what documentation was required, and how NICE responded to the approach.