LLM Concept Mapping for Accurate EHR Code Standardization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The variability in electronic health record (EHR) data encoding due to institution-specific formats and coding systems leads to ambiguity and challenges in multi-site clinical information exchange, exacerbated by factors like misspellings, omissions, and synonyms, hindering seamless data interoperability.
Innovation Solution
Fine-tuning pre-trained vector embedding models, such as SAPBERT and BIOBERT, to map proprietary codes to standardized codes like LOINC, using techniques like vector embeddings, similarity measures, and hard triplet mining to enhance data interoperability by improving the accuracy of code recommendations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If institution-specific coding systems and formats are used in EHRs, then local data encoding flexibility is improved, but data interoperability and ambiguity across multiple sites deteriorate
Solution Approach 1:
The patent introduces a code mapping system that acts as an intermediary between institution-specific coding systems and standardized codes. The mapping engine translates proprietary codes from different healthcare institutions into standardized codes, enabling interoperability while preserving local encoding flexibility. This mediator layer resolves the contradiction by allowing each institution to maintain its preferred coding system internally while ensuring consistent external communication through standardized code translation.
2Ease of operation
If manual data entry is performed, then data input flexibility is improved, but errors such as misspellings and omissions increase
Solution Approach 1:
The patent implements feedback mechanisms where the code mapping system provides suggestions and corrections to users during data entry. The system analyzes input data, identifies potential errors such as misspellings or omissions, and offers corrective feedback to improve data accuracy. This feedback loop maintains ease of operation by allowing flexible input while systematically reducing errors through automated validation and suggestion systems.
3Reliability
If proprietary codes are mapped to standardized codes, then data interoperability is improved, but mapping accuracy deteriorates due to ambiguity
Solution Approach 1:
The patent employs preliminary action by pre-establishing comprehensive mapping relationships between proprietary codes and standardized codes before actual data mapping occurs. The system maintains extensive mapping tables and performs preliminary analysis of code relationships to resolve ambiguities in advance. This preliminary preparation improves mapping accuracy by having pre-validated mapping rules ready, reducing errors during the actual code translation process while maintaining high interoperability.
Data Source
AI summary
Techniques for fine-tuning a pre-trained vector embedding model for recommending standard codes for mapping with proprietary codes are disclosed. Proprietary codes, as referred to herein, include reference codes particular to organizations or vendors. Standard codes, as referred to herein, are industry or standardized codes. The system access a candidate set of standard codes that have been mapped to one or more proprietary codes. The system determines a number of times a standard code is mapped to a proprietary code. Standard codes that have been mapped to proprietary codes a number of times that meet a threshold are selected to be included in a training set for fine-tuning the pre-trained vector embedding model. Standard codes with a number of mappings that does not meet the threshold are not included in the training set. The system may also use vector embedding models for generating aggregated datasets for datasets of the training set.