AWS re:Invent 2025 - Accelerating Real-World Evidence Generation in Life Sciences (IND202)

Accelerating Real-World Evidence Generation in Life Sciences

The Importance of Real-World Data and Evidence

• Pharmaceutical companies analyze over 10 million patient data points for each breakthrough drug, yet 80% of healthcare data remains unstructured and difficult to access
• Real-world data and evidence are transforming how patient outcomes are understood and new treatments are developed
• Leveraging real-world evidence can reduce clinical trial times by up to 40% and cut costs by millions

The Challenges of Real-World Data Access and Analysis

• Exponential growth in healthcare data (projected 10 zettabytes by 2025) creates opportunities but also critical challenges
• Finding, accessing, and deriving value from massive, disparate data sets is a major hurdle
• 40% of data purchases are incomplete, and connecting patient data across multiple sources is extremely difficult

AWS and Partners' Comprehensive Solution

1. Streamlining Data Discovery and Access:

   • Collaboration with Datavant to build a privacy-preserving solution on AWS Clean Rooms
   • Datavant's tokenization technology allows linking patient-level data without revealing PII
   • Enables data producers to share data with consumers while maintaining control

2. Enabling Faster Insights through Scalable Data Harmonization:

   • Partnerships with Etone, Manifold, and Panalgo to provide technical and non-technical data analysis tools
   • Developed an AI agent for Eli Lilly that allows non-technical users to analyze complex healthcare data sets using natural language

Eli Lilly's Real-World Data Journey

• Transitioning from costly clinical trials to leveraging electronic health records and claims data
• Using real-world evidence across the product lifecycle - from early discovery to post-launch
• Challenges include data quality, completeness, and lack of clear standards

Expanding Access and Democratizing Real-World Data Analysis

• Empowering scientists to generate their own insights using low-code and no-code tools
• Developing an AI agent that allows users to get answers to specific questions in natural language
• Key learnings:
  • Providing business context is critical for AI agents to generate accurate results
  • Iterative dialogue and human validation are important for building trust in the outputs

Building Robust, Scalable Agent-Based Solutions

• Multi-agent architecture with specialized agents (SQL creator, medical coder, visualization, etc.)
• Importance of individual agent validation, system-level testing, and incorporating human feedback
• Developing observability, metrics, and auditable workflows to ensure safe and trustworthy AI systems
• Leveraging AWS services and accelerators to speed up development and time-to-production

Conclusion

• Real-world data and evidence are transforming life sciences, but accessing and analyzing the data remains a major challenge
• AWS and partners have developed a comprehensive solution to streamline data access and enable faster, more democratized insights
• Eli Lilly's experience showcases the power of agent-based systems to empower non-technical users and accelerate real-world evidence generation
• Building robust, scalable, and trustworthy agent-based solutions requires careful design, validation, and incorporation of human expertise