Objective:

To evaluate the comparative efficacy and safety of Oliceridine versus Sufentanil specifically in postoperative pain management for burn skin grafting using machine learning.

Key Findings:

• The XGBoost efficacy model (AUC = 0.843) identified total burn surface area and surgery duration as dominant predictors, highlighting the importance of these factors in pain management.
• Opioid choice had negligible SHAP impact, indicating a comparable analgesic profile between Oliceridine and Sufentanil.
• The XGBoost PONV model (AUC = 0.788) identified Sufentanil administration and female patients as key risk factors, suggesting targeted approaches for these groups.
• Substituting Sufentanil with Oliceridine correlated with a significantly lower predicted PONV risk in female patients.

Interpretation:

Machine learning evaluation suggests a distinction between opioid predicted efficacy and safety profiles, with Oliceridine showing similar analgesic outcomes to Sufentanil but a lower predicted risk of PONV, which may influence clinical decision-making.

Limitations:

• The study is retrospective and may not establish clinical validity, introducing potential biases.
• Further prospective trials are required to confirm findings and validate the predictive models.

Conclusion:

Exploratory insights provide frameworks for risk stratification in severe burn management, potentially guiding future clinical practices.