Improving Pain Management in Patients with Sickle Cell Disease from Physiological Measures Using Machine Learning...

Area of Science:

• Biomedical Engineering
• Clinical Pain Research
• Machine Learning in Healthcare

Background:

• Accurate pain assessment is critical for effective pain management in Sickle Cell Disease (SCD).
• Subjective pain reporting presents challenges for objective clinical evaluation.
• Existing methods lack robust objective measures for pain in SCD.

Purpose of the Study:

• To develop and validate a machine learning system for mapping objective physiological measures to subjective pain scores in SCD patients.
• To explore the utility of machine learning in enhancing pain assessment within a clinical framework for SCD.
• To investigate the predictive accuracy of physiological data for pain levels in SCD.

Main Methods:

• Utilized Multinomial Logistic Regression, a machine learning technique.
• Collected physiological data and self-reported pain scores from 40 SCD patients.
• Trained and tested a model to predict pain scores on an 11-point and a condensed 4-point scale.

Main Results:

• Achieved an average intra-individual prediction accuracy of 0.578 on an 11-point pain scale.
• Obtained an inter-individual prediction accuracy of 0.429 on the 11-point scale.
• Improved inter-individual accuracy to 0.681 using a condensed 4-point pain scale.

Conclusions:

• Presented a preliminary machine learning model for predicting SCD pain scores with promising accuracy.
• Demonstrated the potential of integrating machine learning with physiological data for objective pain assessment in SCD.
• Highlighted the novelty of this approach within the SCD and pain research domains.