Medical Coding Knowledge Graph Sub-Code Precision
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current medical coding technologies, particularly relying on natural language processing (NLP), face challenges in accurately determining detailed sub-codes from handwritten medical records due to lack of data and contextual limitations, often failing to capture hospital-specific abbreviations and procedures.
Innovation Solution
A computer-implemented method and system that converts medical records into machine-readable text, utilizes a knowledge graph and coding catalog to query for higher-order medical codes, and employs NLP for context analysis, providing a two-step process to determine both main and sub-codes with enhanced precision, including a confidence score and human confirmation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If NLP technology is used to identify medical codes from handwritten records, then automation is improved, but measurement precision deteriorates due to lack of data and contextual limitations
Solution Approach 1:
The medical coding process is segmented into two distinct stages: first identifying main codes using NLP, then determining sub-codes through human expert review. This segmentation allows each stage to be optimized independently - automation for efficiency in the first stage, and human precision for accuracy in the second stage where contextual nuance is critical.
Solution Approach 2:
A structured interface and protocol are introduced as intermediaries between the NLP system and human coders. The system presents NLP-generated main codes to human experts with specific guidance on what sub-code information to seek, mediating the transition from automated output to refined final codes while maximizing both automation benefits and human expertise.
2Speed
If NLP engine analyzes only sentence and paragraph level text, then processing speed is improved, but measurement precision deteriorates due to lack of broader context
Solution Approach 1:
The NLP engine performs preliminary analysis at the sentence and paragraph level to quickly identify potential main codes before human review. This preliminary action captures the bulk of coding opportunities through fast automated processing, while the limited scope maintains processing speed. Human experts then review only the specific cases where sub-code precision is critical.
3Measurement precision
If detailed sub-codes are determined manually, then measurement precision is improved, but productivity deteriorates due to time-consuming review process
Solution Approach 1:
Instead of requiring manual review of all medical records, the system applies partial human action only to cases where NLP-generated main codes need sub-code refinement. This selective approach achieves high precision for sub-codes while maintaining overall productivity by avoiding excessive manual intervention in cases where automation suffices.
4Ease of operation
If hospital-specific abbreviations are not accounted for, then ease of operation is improved, but measurement precision deteriorates due to misinterpretation of treatments
Solution Approach 1:
The system allows human expert coders to self-service by contributing their knowledge of hospital-specific abbreviations and local practices during the review process. This user-generated knowledge is fed back into the system, enabling it to adapt and improve its interpretation of institution-specific terminology while maintaining overall system simplicity for general use.
Data Source
AI summary
The exemplary embodiments disclose a system and method, a computer program product, and a computer system for assigning medical codes. The exemplary embodiments may include receiving a medical record in machine-readable text-form, wherein the medical record comprises at least one treatment, converting a portion of the medical record into a determined first medical code of a first length, querying a knowledge graph comprising medical records and a coding catalog for a second medical code of higher order than the first medical code, wherein the second medical code relates to the first medical code, and searching evidence in the medical record for the second medical code by comparing at least a portion of clear text relating to the second medical code with the medical record.