This presentation discusses the implementation of an innovative AI-driven solution at Labcorp designed to streamline and accelerate the test matching process. As a global leader in laboratory services, Labcorp processes thousands of test matching inquiries monthly, requiring significant subject matter expert (SME) bandwidth. The solution leverages advanced AWS services, including Amazon Bedrock, OpenSearch Serverless, and Lambda, in a Retrieval-Augmented Generation (RAG) workflow to automate and enhance the matching process.

The system demonstrates robust performance with a 76% top-5 accuracy rate across diverse test matching scenarios. The architecture incorporates daily updates to the test compendium and weekly reference lab scrapes to maintain current information. A key feature is the LLM’s ability to provide detailed explanations for match suggestions, which has received positive feedback from SMEs.

The implementation aims to reduce test matching time by one-third while maintaining accuracy. This solution particularly excels in handling complex queries with variable information content, from simple test names to detailed specifications including methodology, specimen types, and CPT codes. The project showcases successful collaboration between AI teams and subject matter experts, demonstrating how artificial intelligence can enhance healthcare operations while preserving expert oversight.

Keywords: Healthcare AI, Test Matching, AWS, Machine Learning, Laboratory Operations, RAG Architecture.