Risk Factors for Chronic Post-Surgical Pain in the Elderly: A Single-Center Retrospective Study.

Purpose: Chronic Post-Surgical Pain (CPSP) is a common surgical complication, but the association between perioperative complications, patients' intrinsic mental status, and 3-month CPSP remains unclear in elderly surgical populations. This study thus aims to identify perioperative risk factors for 3-month CPSP in elderly patients after non-cardiac surgery, with CPSP here defined as pain intensity ≥3 on the Numerical Rating Scale at 3-month follow-up. Patients and Methods: This retrospective study included 367 elderly patients. We first analyzed variables with descriptive statistics, then conducted all subsequent analyses separately for each of the three surgical subgroups, allowing for potential nuances in the contributory patterns of key factors across groups. To predict 3-month CPSP, we used 10 machine learning algorithms. Model performance was assessed via repeated 5-fold cross-validation, and top-performing models were interpreted using SHapley Additive exPlanations (SHAP) to clarify how key factors contribute. Results: Of 367 patients, the overall prevalence of 3-month CPSP was 25.07%, with significant variation across surgical subgroups: 48.05% in orthopedic surgery, 10.34% in urinary tumor surgery, and 7.14% in abdominal tumor surgery. The Random Forest model showed strong, consistent predictive ability across the three subgroups. Frailty was a key shared risk factor for CPSP across all surgical types, and further analyses identified surgery-specific risk factors. Conclusion: These findings demonstrate that data-driven models can reliably predict CPSP across studied surgical types, with frailty state as a universal risk factor and distinct surgery-specific profiles supporting tailored perioperative risk assessment and prevention strategies.