Evaluation of five risk prediction models for elective abdominal aortic aneurysm repair using the UK National Vascular Database

Abstract

Abstract Background There is no consensus on the best risk prediction model for mortality following elective abdominal aortic aneurysm (AAA) repair. The objective was to evaluate the performance of five risk prediction models using the UK National Vascular Database (NVD). Methods Data on elective AAA repairs from the NVD between January 2008 and December 2010 were analysed. The models assessed were: Glasgow Aneurysm Score (GAS), Vascular Biochemical and Haematological Outcome Model (VBHOM), physiological component of the Vascular Physiological and Operative Severity Score for enUmeration of Mortality (V-POSSUM), Medicare and Vascular Governance North West (VGNW). Overall model discrimination and calibration in equally sized risk-group quintiles were assessed. Results The study cohort included 10 891 patients undergoing elective AAA repair (median age 74 years, 87·3 per cent men). The in-hospital mortality rates following endovascular and open repair were 1·3 and 4·7 per cent respectively (2·9 per cent overall). The Medicare and VGNW models both showed good discrimination (area under receiver operating characteristic (ROC) curve 0·71), whereas the GAS, VBHOM and V-POSSUM models showed poor discrimination (area under ROC curve 0·60, 0·61 and 0·62 respectively). The VGNW model was the only one to predict the overall mortality rate in the cohort (3·3 per cent predicted versus 2·9 per cent observed; P = 0·066). The VGNW model demonstrated good calibration, predicting risk accurately in four risk-group quintiles. The Medicare, V-POSSUM and VBHOM models accurately predicted risk in three, two and no risk-group quintiles respectively. Conclusion The Medicare and VGNW models contain similar risk factors and showed good discrimination when applied to the NVD. Both models would be suitable for risk prediction after elective AAA repair in the UK.