Preoperative Risk Score for Mortality Within 3 Years of Elective Aortobifemoral Artery Bypass.

BACKGROUND: The purpose of this study was to create a risk score for mortality within 3 years of elective aortobifemoral artery bypass for aortoiliac arterial occlusive disease based on variables at the time of preoperative clinical presentation utilizing 5,912 patients in the Vascular Quality Initiative (VQI). METHODS: Multivariable Cox regression time-dependent analysis was performed for the outcome of 3-year mortality utilizing variables which achieved a univariable P value of ≤ 0.05. Using this regression, it was determined which variables have a multivariable association for the outcomes as defined by a regression P value of 0.05 or less. A risk score was then created for the primary outcome. Variables with a multivariable P value of ≤0.05 from the above-mentioned regression were included in the risk score and weighted based on their respective regression beta coefficient in a point scale. Variables with a beta coefficient of less than 0.2 were assigned 1 point, and then a point was added for each rise in beta coefficient at 0.2 intervals. Mortality rate at each risk score bundle was then calculated for the testing and validation cohorts. Supplemental machine learning (ML) IBM SPSS modeler software analysis was conducted as well for the primary outcome. RESULTS: Characteristics with a multivariable (P < 0.05) association with 3-year mortality that ultimately included the risk score were home in a nonmetropolitan area (HR 1.44, P = 0.017); female sex (HR 1.25, P = 0.04); living in a neighborhood within the most 20% of disadvantaged on area deprivation index (HR = 1.31, P = 0.026); body mass index (BMI) < 20 kg/m2 (HR 1.80, P < 0.001); age <40 (HR 0.247, P = 0.036); non-insulin-dependent diabetes (HR 1.50, P = 0.005); congestive heart failure(HR 1.95, P = 0.036); end-stage renal disease on dialysis (HR 5.36, P < 0.001); renal insufficiency (HR 1.45, P = 0.023); anemia (HR 1.84, P < 0.001); indication of ischemic rest pain (HR 1.6, P < 0.001); and need for outflow endarterectomy at distal anastomosis sites (HR 1.35, P = 006). There was noted to be a steep escalation in 3-year mortality rates with each advancing risk score bundle. Patients with negative risk scores had just 2.6% 3-year mortality, while patients with scores over 16 experienced 43.8% 3-year mortality rate (odds ratio [OR] 28.6, P < 0.001). Hosmer-Lemeshow goodness of fit testing revealed 92.1% total model accuracy with a P = 1.0, where a P value of greater than 0.05 indicates a good model fit overall. There was no statistically significant difference in mortality rate between the testing and validation cohorts at any of the risk score bundles. SPSS modeler analysis revealed XG Boost methodology to be the most accurate ML method with an area under the curve of 0.843 and excellent variable importance agreement with the risk score weighting. CONCLUSION: A risk score for 3-year mortality following elective aortobifemoral artery bypass utilizing preoperative variables has been created which has both good accuracy and outstanding internal VQI validation. The data herein have the potential to guide physicians in choosing on which patients to perform direct aortic open surgical reconstruction versus endovascular therapy or extra-anatomic bypass.