Development and performance of a diagnostic/prognostic scoring system for breakthrough pain

Area of Science:

• Pain Management
• Biostatistics
• Clinical Diagnostics

Background:

• Breakthrough pain (BTP) prevalence and treatment vary due to a lack of validated diagnostic tools.
• Accurate BTP assessment is crucial for effective pain management strategies.
• Existing tools lack prognostic capabilities for BTP.

Purpose of the Study:

• To develop and validate a novel scoring system for classifying breakthrough pain (BTP) likelihood.
• To utilize a naïve Bayes classifier (NBC) approach with a validated questionnaire for BTP assessment.
• To categorize BTP into three likelihood classes: High, Intermediate, and Low.

Main Methods:

• A training set of 120 mixed chronic pain patients was used to develop the NBC model.
• Predictors for BTP likelihood were identified using statistical tests (Pearson's χ², Fisher's exact test).
• A scoring algorithm was built, and thresholds were established using Patient's Global Score (PGS) medians for validation.

Main Results:

• The scoring system achieved an overall accuracy of 90.9%.
• Correct classification rates were 63.6% for 'High' and 100% for 'Intermediate' and 'Low' BTP likelihood classes.
• Key predictors identified included flair frequency, duration, and predictability.

Conclusions:

• The NBC-based BTP scoring system demonstrates satisfactory performance in classifying BTP likelihood.
• The system's reliability, flexibility, and simplicity offer potential relevance for BTP epidemiology and management.
• Further impact studies are recommended to generalize the findings.