Automating Clinical Study Protocol Synopsis Generation for Real-World Evidence

Client Need: The Real-World Evidence (RWE) team faced a significant challenge in generating concise and standardized study synopses from comprehensive protocol documents. These documents, often exceeding 40 pages, required careful review to extract essential information for summaries limited to 2–3 pages. The process was not only time-intensive but also prone to inconsistencies, impacting the team's efficiency. The RWE team sought a solution leveraging Generative AI (GenAI) to automate this task, ensuring accuracy, clarity, and significant time savings.

service

TrieDatum’s Solution: TrieDatum designed and delivered an innovative solution tailored to the RWE team’s requirements. The implementation included three core pillars:

1. Streamlined Protocol Intake:

  • A wizard-driven application was developed to simplify the process of onboarding protocol files.
  • Automated metadata extraction and vectorization capabilities were integrated to enable the efficient organization and analysis of content.
  • The system prepared protocols for GenAI processing, ensuring seamless analysis and summary generation.

2. User-Friendly Interface:

  • TrieDatum prioritized the creation of an intuitive interface, enabling users to interact with the application effortlessly.
  • Users could provide study-specific prompts, initiate summary generation, and download output in a standardized Word format with minimal training.
  • TrieDatum also introduced a “history feature” which enabled the user to quickly go back a previously generated synopsis.
  • The application’s design focused on ensuring ease of adoption and consistent user experience across teams.

3. GenAI Integration:

  • TrieDatum implemented foundational large language models (LLMs) combined with a Retrieval-Augmented Generation (RAG) framework, which utilized metadata filtering to further fine-tune the RAG process.
  • Preloading protocol documents allowed the system to perform similarity analysis, ensuring accurate and contextually relevant outputs.
  • The GenAI-powered content preserved the integrity of essential sections, such as Inclusion/Exclusion Criteria and Schedule of Events.

Rapid Delivery: TrieDatum delivered a fully functional, end-to-end solution within five weeks, demonstrating their agility and deep understanding of client needs. The rapid implementation enabled the RWE team to quickly evaluate and adopt the new technology, transforming their workflow and significantly reducing manual effort.

Outcome:

  • Efficiency Gains: Reduced the time required to create study synopses from days to minutes.
  • Consistency: Ensured uniformity and accuracy in generated summaries across multiple protocols.
  • Enhanced Productivity: Allowed the RWE team to focus on higher-value tasks by automating a manual and repetitive process.

Conclusion: TrieDatum’s innovative use of Generative AI technologies empowered the RWE team to streamline their synopsis creation process, saving time while maintaining precision. By delivering a rapid, user-friendly, and scalable solution, TrieDatum showcased their expertise in leveraging AI to address complex business challenges, cementing their role as a trusted technology partner.