Presentation (CTI)

OBJECTIVES: As Global Value Dossiers (GVDs) are critical in informing health technology assessments (HTA) and payer decisions, efficient and accurate content creation is paramount. The GVD Coauthoring Accelerator is a generative artificial intelligence (Gen AI) driven tool designed to streamline the traditionally labor-intensive GVD authoring process. This study aimed to evaluate the Accelerator’s capability to produce a high-quality draft for which tailored prompt repositories and accelerators have been developed. Specifically, we assessed the tool’s accuracy, time savings, and overall usability compared to conventional manual drafting methods.
METHODS: A pilot evaluation focused on breast cancer was conducted, with ~75 reference documents (e.g., peer-reviewed literature). We curated a template for the Disease Overview section, including predefined prompts capturing epidemiology, pathophysiology, disease burden, and unmet needs. A Retrieval-Augmented Generation (RAG) framework was used to apply the configured prompts to the reference documents. Outputs were generated and subsequently reviewed by three independent SMEs with over five years of medical writing experience. Key metrics included: (1) accuracy of extracted content; (2) completeness of critical disease-related information; (3) time to first draft completion; and (4) user satisfaction ratings regarding ease of use and clarity of generated text.
RESULTS: The Accelerator demonstrated a high accuracy rate (~95%) in extracting the key required evidence. Completeness scores approached 90%, as most crucial epidemiological data, pathophysiology, and risk factors were captured. Draft creation time was reduced by ~60% compared to manual methods, significantly accelerating the time to first GVD draft. Reviewers reported high satisfaction levels (~80%), noting improved efficiency and the ability to focus more on strategic content refinement rather than initial drafting.
CONCLUSIONS: The GVD Coauthoring Accelerator effectively aggregates the provided evidence in the preferred GVD format. These findings support its potential to enhance the scalability, consistency, and strategic quality of GVD development, thereby improving alignment with HTA and payer expectations.