Predictors of Mortality
Multivariable analysis (Table 4) identified the following
variables as independent predictors of mortality: 1) age (Odds Ratio
[OR] = 1.05, per 5-year increment), the need for urgent/emergent
surgery (OR = 5.46) and left ventricular ejection fraction (LVEF)
< 40% (OR = 2.14). Surgical procedure (repair vs.
replacement) was not an independent predictor of mortality. The 30-day
mortality predictors were internally validated using a 1000-resampling
bootstrap procedure that showed the stability of the results
(Table 5 ). Moreover, there were no features of
multicollinearity, as the standard errors of the affected coefficients
were relatively small (Table 5 ). The c-statistic for this
multivariable model was 0.80, with asymptotic 95% confidence of
0.75-0.86, indicating good discrimination of the model (Table 6
and Figure 1). The Hosmer-Lemeshow 2 statistic showed
that the model goodness of fit was good (x2 = 2.78, p
= 0.95), indicating good predictive performance (discrimination and
calibration) of the derived model.