Development of a Method for Automatic Matching of Unstructured Medical Data to ICD-10 Codes.

Area of Science:

• Medical Informatics
• Machine Learning
• Computational Biology

Background:

• Inconsistent disease coding standards hinder medical data exchange and analysis.
• Manual coding is time-consuming and prone to errors.
• Lack of standardized coding impacts research and clinical decision-making.

Purpose of the Study:

• To develop and evaluate a machine learning system for automated disease classification.
• To map unstructured clinical notes to standardized International Classification of Diseases, 10th Revision (ICD-10) codes.
• To improve the efficiency and accuracy of medical coding in healthcare informatics.

Main Methods:

• A novel machine learning architecture was designed, incorporating a training layer and a knowledge base.
• The system was trained to match unstructured medical text (e.g., doctor's notes) to ICD-10 codes.
• Logistic regression was identified as the optimal model through experimental evaluation on real-world data.

Main Results:

• The machine learning system demonstrated effectiveness in disease classification prediction.
• Logistic regression achieved a high accuracy of 81.07% during high-load testing.
• The system exhibited acceptable error rates, indicating reliability for practical application.

Conclusions:

• The proposed machine learning system offers a viable solution for overcoming coding standard incompatibilities.
• Automated prediction of ICD-10 codes from unstructured medical text can significantly enhance healthcare informatics.
• This approach promises to streamline data analysis and improve the quality of medical information systems.