Prognostic Factors for Postoperative Complications. An Aggregate Protocol for 10 Observational Studies From the Danish
TRIPLE
‐A Cohort of 1.2 Million Surgeries
Anders Peder Højer Karlsen, Asta Prescott, Caroline Folkersen, Jens Laigaard, Atena Saito, Caroline Lando Klammer, Mik Wetterslev, Emil Ipsen Ørskov, Mathilde Fahrendorff, Rebecca Kiær, Helene Holm Laigaard, Nichlas Hovmand, Lars Hyldborg Lundstrøm, Sebastian Buhl Rasmussen, Anders Kehlet Nørskov, Markus Harboe Olsen ABSTRACT
Background
Postoperative complications substantially increase morbidity, mortality and healthcare costs. Understanding prognostic factors is essential for risk stratification, targeted prevention strategies, and development of prediction models.
Objectives
To identify prognostic factors for 18 major postoperative complications in 10 studied domains using the Danish TRIPLE‐A database: (1) acute postoperative pain, nausea and vomiting, and need for rescue opioids; (2) persistent opioid use; (3) acute kidney injury; (4) venous thromboembolism; (5) staphylococcal surgical site infection and infections requiring antibiotic treatment; (6) three airway management complications; (7) transfusion requirements; (8) delirium; (9) new‐onset postoperative atrial fibrillation; and (10) readmission, reoperation and unplanned ICU admission.
Methods
We will conduct 10 retrospective cohort studies using electronic health record data from 1.2 million surgical procedures performed in the Capital and Zealand Regions of Denmark from 2017 to 2025. Each individual complication will be identified via validated combinations of ICD‐10 codes, medication administration, laboratory values, radiology examinations, and free‐text mining of clinical notes. Candidate prognostic factors will be selected a priori based on literature review and clinical expertise, and will encompass patient demographics, comorbidities, surgical characteristics, anaesthetic management, vital parameters and biomarkers. We will estimate adjusted associations between candidate prognostic factors and complications using multivariable logistic regression with LASSO penalisation to reduce overfitting. Effect estimates will be reported as odds ratios with 95% confidence intervals, and the relative contribution of predictors will be assessed by changes in model discrimination (Δ‐AUC). The approach for surgical procedure, calendar year, and hospital site will be determined for each study (adjustment, stratification, or inclusion as candidate predictors).
Conclusion
These studies aim to validate established prognostic factors and identify novel ones for major postoperative complications, providing a foundation for the development of individualised perioperative risk stratification models.