STANDARDIZING AI MODEL INTERACTION WITH HEALTHCARE SYSTEMS USING THE MODEL CONTEXT PROTOCOL (MCP)
Author(s)
Rajdeep Kaur, PhD1, Barinder Singh, RPh1, Mrinal Mayank, B.Tech1, Ruhi j, B.Tech1, Nicola Waddell, HNC2, Shubhram Pandey, MSc1;
1Pharmacoevidence Pvt. Ltd., Mohali, India, 2Pharmacoevidence Pvt. Ltd., London, United Kingdom
1Pharmacoevidence Pvt. Ltd., Mohali, India, 2Pharmacoevidence Pvt. Ltd., London, United Kingdom
OBJECTIVES: Clinical data is distributed across heterogeneous systems, including clinical databases, research repositories, electronic health records (EHRs), external data sources, and trial management platforms. This proof-of-concept aimed to validate the MCP as a standardized, secure, and reproducible interface for enabling AI access to distributed HEOR and life sciences data sources.
METHODS: A proof-of-concept AI framework was developed to evaluate MCP as a unified interface for AI interaction within complex healthcare environments. The system integrated with SharePoint to retrieve contextually relevant unstructured evidence (including dossiers, reports, PDF files for a specific disease area) via metadata-driven search. In addition, the system was connected to Databricks using MCP, enabling the secure execution of SQL and Python analyses directly within the platform to retrieve structured data. A virtual assistant was integrated to synthesize information from the appropriate data sources and generate HEOR-relevant insights. Generated insights were reviewed and validated by Subject Matter Experts (SMEs) to ensure accuracy, relevance, clarity, and alignment with established HEOR evidence standards
RESULTS: SMEs evaluated the system using more than 50 prompts to assess integration performance. MCP-based integration reduced the effort required to identify relevant evidence across systems and minimized the cross-platform navigation compared to traditional manual workflows. This resulted into faster insight generation. SMEs confirmed that the generated insights were relevant, interpretable, and offered transparent traceability to underlying data sources, demonstrating suitability for healthcare evidence generation and decision support
CONCLUSIONS: This study demonstrated that MCP enables standardized and governed integration of structured and unstructured HEOR data sources, supporting efficient, consistent, and traceable evidence generation. Furthermore, the MCP-enabled architecture is extensible, allowing for the incorporation of additional data sources and platforms while maintaining necessary governance controls and expert oversight
METHODS: A proof-of-concept AI framework was developed to evaluate MCP as a unified interface for AI interaction within complex healthcare environments. The system integrated with SharePoint to retrieve contextually relevant unstructured evidence (including dossiers, reports, PDF files for a specific disease area) via metadata-driven search. In addition, the system was connected to Databricks using MCP, enabling the secure execution of SQL and Python analyses directly within the platform to retrieve structured data. A virtual assistant was integrated to synthesize information from the appropriate data sources and generate HEOR-relevant insights. Generated insights were reviewed and validated by Subject Matter Experts (SMEs) to ensure accuracy, relevance, clarity, and alignment with established HEOR evidence standards
RESULTS: SMEs evaluated the system using more than 50 prompts to assess integration performance. MCP-based integration reduced the effort required to identify relevant evidence across systems and minimized the cross-platform navigation compared to traditional manual workflows. This resulted into faster insight generation. SMEs confirmed that the generated insights were relevant, interpretable, and offered transparent traceability to underlying data sources, demonstrating suitability for healthcare evidence generation and decision support
CONCLUSIONS: This study demonstrated that MCP enables standardized and governed integration of structured and unstructured HEOR data sources, supporting efficient, consistent, and traceable evidence generation. Furthermore, the MCP-enabled architecture is extensible, allowing for the incorporation of additional data sources and platforms while maintaining necessary governance controls and expert oversight
Conference/Value in Health Info
2026-05, ISPOR 2026, Philadelphia, PA, USA
Value in Health, Volume 29, Issue S6
Code
MSR241
Topic
Methodological & Statistical Research
Topic Subcategory
Artificial Intelligence, Machine Learning, Predictive Analytics
Disease
No Additional Disease & Conditions/Specialized Treatment Areas