Measuring Pain in Sickle Cell Disease using Clinical Text.

Area of Science:

• Hematology
• Medical Informatics
• Computational Biology

Background:

• Sickle Cell Disease (SCD) is an inherited red blood cell disorder.
• Pain, stroke, and organ failure are common SCD complications due to blocked blood vessels.
• Accurate pain assessment in SCD is challenging for medical practitioners (MPs) due to pain's subjective nature.

Purpose of the Study:

• To develop and evaluate machine learning (ML) models for classifying clinical notes of SCD patients based on pain.
• To predict the relevance of pain in clinical notes using binary classification.
• To predict the specific pain level in SCD patients using multiclass classification.

Main Methods:

• Four binary ML classifiers were trained to predict pain relevance in clinical notes.
• A multiclass ML classifier was developed to predict pain levels.
• Model performance was evaluated using metrics such as F-measure.

Main Results:

• Binary classifiers showed comparable performance in predicting pain relevance.
• The Decision Trees model achieved the highest F-measure of 0.70 for multiclass pain level prediction.
• Clinical text analysis and ML demonstrate potential for SCD pain management.

Conclusions:

• Machine learning models can effectively classify clinical notes for SCD pain assessment.
• Decision Trees offer a promising approach for multiclass pain level prediction in SCD.
• Automated analysis of clinical text can significantly improve pain management strategies for SCD patients.