Invited commentary

Abstract

The current study from Eslami et al has used the American College of Surgeons National Surgical Quality Improvement Project (ACS-NSQIP) to identify a risk prediction model for 30-day mortality after abdominal aortic aneurysm (AAA) repair by an open or endovascular approach. The model they developed showed that cardiac, pulmonary, and renal disease, in addition to age, functional dependence, American Society of Anesthesiologists class, gender, and modality of repair were all associated with higher 30-day mortality. The use of “big data” has permitted investigators to predict rare events and evaluate rare covariates with outcomes of interest. However, the use of secondary data to test a research hypothesis carries important considerations. How were the data collected, and for what purpose? Did the data set contain important covariates known to be associated with the outcome? Is the chosen outcome appropriate and clinically useful? In clinical practice, patients are counseled about their annual AAA rupture risk. A 30-day or in-hospital mortality risk does not capture later death, which is also an important consideration in the decision to proceed with aneurysm repair. Specific risk factors may assess 1-year mortality risk and be more helpful in contrasting annual rupture risk vs repair risk at 1 year, as opposed to 30-day outcomes alone.1Beck A.W. Goodney P.P. Nolan B.W. Likosky D.S. Eldrup-Jorgensen J. Cronenwett J.L. et al.Predicting 1-year mortality after elective abdominal aortic aneurysm repair.J Vasc Surg. 2009; 49 (discussion: 843-4): 838-843Abstract Full Text Full Text PDF PubMed Scopus (103) Google Scholar, 2Fitridge R.A. Boult M. de Loryn T. Cowled P. Barnes M. Predictors of 1-year survival after endovascular aneurysm repair.Eur J Vasc Endovasc Surg. 2016; 51: 528-534Abstract Full Text Full Text PDF PubMed Scopus (14) Google Scholar ACS-NSQIP is limited to only 30-day events. With the addition of this model, there are no fewer than five risk-prediction models for perioperative mortality after AAA repair.3Giles K.A. Schermerhorn M.L. O'Malley A.J. Cotterill P. Jhaveri A. Pomposelli F.B. et al.Risk prediction for perioperative mortality of endovascular vs open repair of abdominal aortic aneurysms using the Medicare population.J Vasc Surg. 2009; 50: 256-262Abstract Full Text Full Text PDF PubMed Scopus (120) Google Scholar, 4Grant S.W. Grayson A.D. Purkayastha D. Wilson S.D. McCollum C. participants in the Vascular Governance North West ProgrammeLogistic risk model for mortality following elective abdominal aortic aneurysm repair.Br J Surg. 2011; 98: 652-658Crossref PubMed Scopus (41) Google Scholar, 5Eslami M.H. Rybin D. Doros G. Kalish J.A. Farber A. Vascular Study Group of New EnglandComparison of a Vascular Study Group of New England risk prediction model with established risk prediction models of in-hospital mortality after elective abdominal aortic aneurysm repair.J Vasc Surg. 2015; 62: 1125-1133.e2Abstract Full Text Full Text PDF PubMed Scopus (36) Google Scholar, 6Grant S.W. Hickey G.L. Grayson A.D. Mitchell D.C. McCollum C.N. National risk prediction model for elective abdominal aortic aneurysm repair.Br J Surg. 2013; 100: 645-653Crossref PubMed Scopus (39) Google Scholar The ability of these models to discriminate risk, measured by the area under the receiver operating characteristic curve, ranges from 0.70 to 0.82. Although this is a fair degree of discrimination, a significant portion of the variation in death is left unexplained. Many risk-prediction models do not calibrate risk appropriately for patients at very low or very high risk. Although this study appears to calibrate well to a 6% mortality risk, whether that degree of calibration holds for higher predicted death risks is unclear. Overall, the current risk model appears to perform in a fashion similar to other available models. This may be due to the lack of important variables in ACS-NSQIP (ie, aneurysm size) or development of the model to be simplified (converting age to a dichotomous variable). Robust risk-prediction models with improved discrimination and calibration must identify those factors unaccounted for in our current data sets, potentially incorporating measures of frailty. In addition, models that evaluate late mortality are of equal importance to ensure optimal patient selection for AAA repair. Description of a risk predictive model of 30-day postoperative mortality after elective abdominal aortic aneurysm repairJournal of Vascular SurgeryVol. 65Issue 1PreviewDespite vast improvement in the field of vascular surgery, elective abdominal aortic aneurysm (AAA) repair still leads to perioperative death. Patients with asymptomatic AAAs, therefore, would benefit from an individual risk assessment to help with decisions regarding operative intervention. The purpose of this study was to describe such a 30-day postoperative (POD) risk prediction model using American College of Surgeons National Surgical Quality Improvement Project (NSQIP) data. Full-Text PDF Open Archive