Idea
A platform that standardizes and hierarchically organizes multi-source EHR codes for healthcare researchers and institutions.
Research Paper
Core Innovation
This paper introduces MASH, a novel automated framework that aligns heterogeneous medical codes across institutions using neural optimal transport and hyperbolic embeddings. It uniquely integrates multiple data sources including language models and co-occurrence patterns to build interpretable hierarchical graphs, addressing the challenge of unstructured and diverse EHR codes. This approach improves semantic and hierarchical understanding beyond prior manual or single-source methods.
Market Size (TAM)
$10–20B TAM, $2–5B SAM; assumption: large healthcare data management and analytics market driven by EHR standardization needs.
Potential Customers & Pain Points
- Hospitals needing standardized EHR data integration
- Healthcare researchers requiring interpretable clinical code hierarchies
- Health IT vendors seeking automated code alignment solutions
Business Model
SaaS platform licensing to healthcare providers and research institutions with tiered pricing based on data volume and features.
Competitive Landscape
- Epic Systems
- Cerner
- IBM Watson Health
Implementation Challenges
- Data privacy and compliance challenges
- Integration complexity with diverse EHR systems
- Adoption resistance due to workflow changes
Validation Strategy
- Pilot deployment with partner hospitals to demonstrate integration and accuracy
- Benchmark against existing EHR code mapping tools for performance
- Collect user feedback to refine hierarchy interpretability and usability
Research Paper Overview
Automated Hierarchical Graph Construction for Multi-source Electronic Health Records
Summary
Electronic Health Records contain diverse clinical data hindered by heterogeneous codes and lack of standardization. MASH is a fully automated framework aligning medical codes across institutions using neural optimal transport and hyperbolic embeddings to build hierarchical graphs. It integrates language models, co-occurrence, textual descriptions, and labels to capture semantic and hierarchical relationships, producing interpretable hierarchies for diagnosis, medication, and lab codes, including unstructured local lab codes.