A backward stepwise selection procedure with a cutoff level of p = 0.05 was used.
Multivariate (backward stepwise selection method with probability for the removal of 0.10) logistic regression analyses were used to determine the association of variables with 1-year mortality.
Multivariable linear regression models were run to adjust for confounders using a backward stepwise selection procedure with a significance level for removal from the model of 0.1.
Multivariate survival analysis was based on Cox's proportional hazards model using a backward stepwise selection procedure with entry and removal criteria of P=0.05 and P=0.10, respectively.
Univariate and multivariate (backward stepwise selection method with probability for the removal of 0.10) logistic regression analyses were used to determine the association of variables with 28-day mortality.
The relationship between EQ-5D score (response variable) and the candidate explanatory variables was assessed using a backward stepwise selection process with α = 0.05 as criteria for model inclusion in SAS statistical software (proc glmselect).
Cox proportional hazards regression model was used to estimate the risk of death by multivariate analysis (backward stepwise selection method with probability for the removal of 0.10) for the whole population.
A multivariate Cox regression analysis was then performed using a backward stepwise selection method, with a P value less than 0.05 as the entry criterion and a P value 0.10 or higher as the removal criterion.
Kaplan-Meier logrank and univariate and multivariate (backward stepwise selection method with probability for removal of 0.10) Cox proportional hazards regression models were used to identify the strongest predictors of overall time-tagged mortality using time to death as a continuous variable.
