Analysis
Statistical analysis was performed with SAS 9.2 software (SAS Institute, Inc, Cary, NC). Categorical variables were expressed as percentages. Continuous variables were expressed as means ± standard deviations. We analyzed categorical variables with chi-square analysis, or if there were few outcomes, Fisher’s exact test was used. Continuous variables that had a normal distribution were analyzed using the Student’s t-test, and variables that had non-normal distribution were analyzed using the Wilcoxon rank-sum test. Statistical significance was based on two-sided p-values less than 0.05.
Variables with a univariate p-value of less than 0.25 and those of known biological importance were selected for inclusion in a multivariable logistic regression model to identify independent predictors of hospital mortality. Model discrimination and calibration were evaluated by the area under the receiver operating characteristic curve (C statistic), and the Hosmer–Lemeshow goodness-of-fit statistic, respectively. In the current study, a model’s discrimination with an area under the ROC curve greater than 0.7 was considered a good model, and Hosmer–Lemeshow test with P > 0.05 indicates a well-calibrated model. To validate the final predictive model, we did bootstrap replication with 1000 re-samples.