Generalizability of Readability Models for Medical Terms

Area of Science:

• Natural Language Processing
• Medical Informatics
• Machine Learning

Background:

• Accurate comprehension of medical texts is vital for patient safety and effective healthcare.
• Identifying complex medical terminology is a significant challenge in clinical documentation and patient education.

Purpose of the Study:

• To develop and evaluate machine learning models for detecting difficult-to-understand words in French medical texts.
• To assess the generalizability of these models across multiple human annotators.

Main Methods:

• Utilized supervised machine learning algorithms combined with word embeddings capturing contextual information.
• Employed manually cross-annotated French medical data from seven annotators.
• Implemented cross-validation techniques to test model generalization.

Main Results:

• The proposed models effectively detected difficult medical words.
• Demonstrated significant generalizability of the models across different annotators, indicating robustness.

Conclusions:

• Combining machine learning with word embeddings offers a promising approach for identifying challenging medical vocabulary.
• The developed models can reliably generalize, suggesting their utility in real-world medical text analysis.