Patient Trial Matching using C2Q

The challenge

A healthcare and clinical research organization aimed to improve patient recruitment efficiency by implementing an AI-powered patient trial matching solution using the Criteria2Query (C2Q) framework and OMOP standardized vocabularies. The objective was to automate the conversion of clinical trial eligibility criteria into structured database queries, enabling faster and more accurate identification of eligible patients. The solution was designed to streamline clinical trial matching, reduce manual screening efforts, and improve semantic understanding of medical eligibility criteria using Large Language Models (LLMs).

Key challenges

• Difficulty converting unstructured clinical trial eligibility criteria into executable database queries
• Managing complex medical terminologies and vocabulary standardization
• Ensuring accurate mapping between patient records and trial requirements
• Reducing manual effort involved in patient screening and recruitment
• Maintaining scalability for large healthcare datasets and clinical environments

Technology used

• Programming Language: Python
• Healthcare Standards: OMOP Common Data Model (CDM v5) for standardized healthcare data representation
• Framework: Criteria2Query (C2Q) for converting eligibility criteria into database queries
• OHDSI Tools: Athena, Usagi, and WebAPI for vocabulary mapping and integration
• LLMs Used: OpenAI and Claude via AWS Bedrock for semantic reasoning and concept understanding
• Cloud Integration: AWS-based GenAI services including Bedrock, Comprehend Medical, and Textract

What we built

• Criteria2Query Exploration: Evaluates and utilizes C2Q framework versions for clinical eligibility query generation
• Vocabulary Indexing: Uses Usagi and OMOP vocabularies for concept standardization and mapping
• Concept Extraction & Reasoning: Applies LLMs for semantic understanding of clinical criteria
• SQL Query Generation: Converts eligibility conditions into executable SQL queries for patient matching
• OMOP Integration: Connects with OMOP CDM and WebAPI tables for clinical data querying
• Automated Trial Matching: Identifies eligible patients based on structured eligibility criteria

Objectives

• Enable automated patient trial matching using OMOP standardized vocabularies
• Improve accuracy in converting clinical eligibility criteria into database queries
• Accelerate patient recruitment for clinical trials through intelligent matching
• Enhance semantic understanding of medical concepts using LLM-based reasoning
• Enable scalable and automated patient eligibility screening workflows
• Support integration with modern healthcare and clinical data systems

Outcomes

• Automated mapping of clinical eligibility criteria into structured database queries
• Faster and more scalable patient recruitment for clinical trials
• Improved semantic understanding of medical concepts using LLM-based reasoning
• Reduced manual effort in patient screening and trial eligibility assessment
• Better accuracy in patient-to-trial matching through standardized vocabularies
• Strong foundation for scalable healthcare GenAI solutions using AWS services