Conception, Development and Validation of Classification Methods for Coding Support of Rare Diseases Using Artificial...

Area of Science:

• Medical Informatics
• Computational Linguistics
• Machine Learning

Background:

• Automated disease coding is crucial for hospital billing accuracy and efficiency.
• Accurate coding of rare diseases presents unique challenges due to limited data.

Purpose of the Study:

• To implement and evaluate machine learning classification methods for automated coding of two rare diseases.
• To assess the suitability and reliability of different classifiers for this task.

Main Methods:

• Utilized Natural Language Processing (NLP) and Machine Learning (ML) for automated disease entity recognition.
• Applied supervised learning with both off-the-shelf and custom-developed classification systems.
• Evaluated classifier performance on unstructured historical patient records and new billing cases.

Main Results:

• Achieved high correctness (F1 score approximately 0.8) in predicting new billing cases.
• Demonstrated effectiveness of automated coding methods even with small datasets (≤ 200 records).
• Compared performance of various classifiers, identifying suitable options for rare disease coding.

Conclusions:

• Automated disease coding using NLP and ML is a viable and effective approach for rare diseases.
• High accuracy is attainable even with limited data, supporting efficient hospital billing processes.
• The study validates the reliability of implemented classification methods for clinical applications.