Machine learning research methods to predict postoperative pain and opioid use: a narrative review

Area of Science:

• * Medical Informatics
• * Computational Biology
• * Health Services Research

Background:

• * The opioid crisis and chronic postsurgical pain highlight the need for predictive tools.
• * Machine learning (ML) offers potential for predicting postoperative pain and opioid use.
• * Advancements in data availability and computational power facilitate ML applications in healthcare.

Purpose of the Study:

• * To review and characterize ML studies predicting pain or opioid use post-surgery.
• * To identify methodological trends and variations in existing ML research.
• * To provide recommendations for future ML studies in this domain.

Main Methods:

• * A narrative review of PubMed-indexed articles was conducted.
• * Two independent reviewers screened 280 titles and abstracts.
• * Data were extracted from 61 studies meeting inclusion criteria.

Main Results:

• * A significant increase in relevant publications over time was observed.
• * Most studies focused on predicting chronic postsurgical pain or opioid use, often using orthopedic surgery data.
• * Methodological variability exists in sample size, predictors, and outcome definitions.
• * Patient-reported predictors were identified as highly informative for ML models.

Conclusions:

• * Machine learning shows promise for predicting postoperative pain and opioid use.
• * Future ML models should prioritize the inclusion of patient-reported data.
• * Enhancing the performance and clinical utility of ML algorithms is crucial for patient care.