Development and Validation of a Multivariable Prediction Model of Perioperative Mortality in Vascular Surgery: The New Zealand Vascular Surgical Risk Tool (Nzrisk-Vasc)

Abstract

Introduction: Risk calculators and prediction models are available to assist clinicians and patients in their peri-operative decision-making to optimise the outcome [1]. In a vascular surgical setting, the majority of these instruments are based on abdominal aortic aneurysm repair outcomes, and in general their clinical use is limited due to variability in their predictive ability at procedural and institutional levels [2]. The objective of this study was to develop and validate a simple and accurate vascular surgical risk-prediction model. Methods: Ethical approval was obtained for the study. National administrative database was accessed to collect information on all adult patients undergoing vascular surgery between 1st of July 2011 and 30th of June 2016 in New Zealand. The database was split randomly into a derivation (75%) and a validation (25%) dataset. The primary outcomes were mortality at 30-days, 1-year and 2-years. Previously established covariates including American Society of Anaesthesiologists physical status (ASA-PS), sex, surgical urgency, history of cancer and ethnicity were tested and other covariates such as smoking status, presence of renal failure, diabetes, anatomical site of operation and structure operated on (artery, vein, arterio-venous, amputations) and type of procedures (open and endovascular) were explored. LASSO regression was used to select variables for inclusion in the model. Results: A total of 21,597 cases formed our final risk-prediction model with co-variates including ASA-PS, gender, surgical urgency, cancer status, presence of renal failure, diabetes, anatomical site, structure and endovascular procedure. The area under the receiver operating curve (AUC) for the 30-day, 1-year and 2-year mortality using LASSO regression was 0.87, 0.82 and 0.80 respectively, demonstrating very good discrimination. Calibration was also sufficient with calibration plot slopes of 0.90, 0.95 and 1.02 respectively. The strongest predictors of mortality included higher ASA-PS, diabetes, anatomical site of operation and structure. Conclusion: We have developed and validated a simple and accurate multivariate risk calculator for vascular surgical patients in New Zealand with excellent discrimination and calibration for 30-day, 1-year and 2-year morality. This tool may assist with risk estimation and aid peri-operative decision-making for both clinicians and patients in a vascular surgical setting. Disclosure: Nothing to disclose