Postoperative Complications in Major Surgery

Overview

The primary objective of this study is to develop and validate a multivariable risk prediction model for 30-day major postoperative complications and mortality in patients undergoing major surgery across participating international centers. Despite advancements in perioperative care, surgical complications remain a leading cause of global morbidity and preventable death, particularly in resource-limited or low- and middle-income country (LMIC) settings. This study utilizes a prospective, multicenter, international cohort design. Data will be collected on patient-level risk factors (e.g., age, frailty, comorbidities), hospital-level infrastructure (e.g., nurse-to-patient ratios, rescue capacity), and perioperative safety processes (e.g., adherence to the WHO Surgical Safety Checklist). Patients will be followed prospectively for up to 30 days post-surgery. The collected data will be used to construct robust predictive models to identify individual patient risk and uncover actionable system-level factors to optimize surgical safety globally.

Background and Rationale: Postoperative complications impose a substantial clinical and economic burden worldwide. While extensive research has focused on patient-derived clinical risks, fewer prospective international studies have mathematically integrated hospital structural capabilities and perioperative safety processes into multi-level predictive frameworks. This protocol describes an international, prospective, multicenter cohort study designed to build and validate a predictive model for adverse surgical outcomes. Study Design and Population: This is a multicenter, prospective cohort study. Participating centers will recruit consecutive adult or pediatric patients undergoing major elective or emergency non-cardiac surgery. Major surgery is defined internationally as any procedure requiring general or neuraxial anesthesia with an anticipated duration \> 90 minutes, estimated blood loss \> 500 mL, or requiring routine postoperative intensive care admission. Patients undergoing minor procedures or those unable to provide informed consent will be excluded. Data Collection and Covariates: Standardized electronic case report forms (eCRFs) will be used to collect data across three main tiers: 1. Patient-level Predictors: Demographics, insurance coverage, American Society of Anesthesiologists (ASA) physical status, Charlson Comorbidity Index, and frailty (measured via the Modified Frailty Index, mFI-5). 2. Procedure: code of surgery. 3. Type of surgery: Emergency/elective surgery, Ambulatory/hospitalized, severity, specialty. Outcomes and Follow-up: All participants will be systematically tracked during their inpatient stay, with mandatory clinical follow-up at postoperative day 30 (via medical record review). Primary Outcome: Composite incidence of major 30-day postoperative complications, graded according to the Clavien-Dindo classification (Grade \> III, indicating complications requiring surgical, endoscopic, or radiological intervention, life-threatening complications, or death). Secondary Outcomes: 30-day all-cause mortality, individual complication rates (surgical site infections, major bleeding, thromboembolic events, organ failure), hospital length of stay, and "Failure to Rescue" rates (proportion of patients who die after developing a major complication). Statistical Analysis and Predictive Modeling: Sample size calculations are based on the Events Per Variable (EPV) criterion, ensuring a minimum of 15-20 events per candidate predictor in the multivariable model to prevent overfitting. Multilevel multivariable logistic regression and mixed-effects Cox proportional hazards models will be constructed, treating the hospital/country of origin as a random effect to account for institutional clustering. Model performance will be rigorously evaluated. Discriminatory capacity will be assessed using the area under the receiver operating characteristic curve (AUROC). Calibration will be assessed via calibration curves (observed vs. predicted risk). Sensitivity analyses will compare traditional regression models. Reporting will adhere strictly to TRIPOD and STROBE guidelines.