Outcomes and statistical analysis
We analyzed pre-, intra-, and early postoperative outcomes. Operative mortality is expressed as procedural mortality, hospital death and 30-day mortality. Neurological complications were divided into permanent and temporary, depending on their presence at discharge, and included focal (stroke), non-focal (coma) and spinal neurologic deficit (paraplegia and paraparesis). The exact definitions of these complications have been provided previously [10]. Myocardial infarction was defined as ischemia diagnosed based on clinical symptoms, new typical ECG changes, decreased local regional wall contractility and elevated biomarkers, and requiring pharmacological, percutaneous or open surgical intervention. Low cardiac output syndrome was defined as patients requiring significant pharmacological support, extracorporeal membrane oxygenation, or ventricular assist device. Pulmonary complications were classified as acute respiratory distress syndrome, pneumonia, pulmonary edema and severe atelectasis, as well as prolonged mechanical ventilation. Reexploration for bleeding included only cases with acute postoperative bleeding (diffuse or with a defined source of bleeding), requiring urgent or emergent intervention. Elective open revisions for pericardial effusion were not included in this definition. Gastrointestinal complications including prolonged ileus (>72 h) or gastric paresis, or new hepatobiliary disfunction (with metabolic acidosis or increase in lactate) were recorded [11].
Data were analyzed using GraphPad Prism Version 7.00 (GraphPad Software, La Jolla, Ca, USA). The distribution of continuous variables was evaluated using the Kolmogorov-Smirnov test and Q-Q plots. Continuous variables are expressed as mean ± standard deviation (when normally distributed) or median and range (non-normal distribution). Categorical data are reported as frequencies and percentages. Multivariable logistic regression was used to determine the independent predictors of in-hospital mortality. Clinically relevant preoperative risk factors for hospital mortality were selected using univariate analysis where p value of less than 0.05 was considered statistically significant. The variance inflation factor (VIF) was evaluated in order to exclude multicollinearity (VIF of 4.0 or more indicated intercorrelation between the analyzed variables). Multivariate logistic regression model was controlled by means of Hosmer-Lemeshow goodness-of-fit test and the area under the receiver operating characteristic curve (ROC).