Protocol for Personalised Prediction of Persistent Postsurgical Pain.

Area of Science:

• Pain Medicine
• Surgical Outcomes Research
• Digital Health

Background:

• Persistent postsurgical pain (PPSP) affects up to 15% of patients, impacting quality of life and increasing long-term opioid use risk.
• Existing PPSP prediction models lack dynamic perioperative factors, limiting their accuracy.
• The Personalised Prediction of Persistent Postsurgical Pain (P5) study addresses this gap.

Purpose of the Study:

• To develop individualized, multimodal prediction models for PPSP.
• To integrate preoperative assessments (behavioral, psychophysical, neurocognitive) and high-frequency symptom monitoring.
• To enhance prediction accuracy beyond static or retrospective data.

Main Methods:

• Prospective, single-center cohort study of 2500 adult surgical patients.
• Collection of preoperative data, including surveys, cognitive, and quantitative sensory testing.
• Utilizing smartphone-based Ecological Momentary Assessments (EMAs) for dynamic, high-frequency postoperative data collection (pain, mood, medication use).
• Employing machine learning and structural equation modeling for predictive model development.
• Assessing model performance using AUC-ROC, AUC-PR, and SHAP values.

Main Results:

• Primary outcome is PPSP at 3 months post-surgery.
• Predictive models will be developed and validated.
• Model interpretability will be ensured using SHapley Additive exPlanations.

Conclusions:

• The P5 study aims to create advanced, personalized PPSP prediction models.
• Integrating dynamic perioperative data is key to improving prediction accuracy.
• Findings will inform strategies to mitigate PPSP and associated risks like long-term opioid use.