In order to determine the strongest baseline and follow-up hemodynamic predictors of 30-day mortality, logistic regression analyses were performed; the change in hemodynamic variables between baseline and follow-up was not entered in a multivariate model since the number of paired measurements was somewhat limited. All hemodynamic parameters were assessed univariately. Baseline and follow-up heart rate as well as follow-up mean right atrial pressure, cardiac index, and pulmonary artery diastolic BP required categorization by tertiles, and follow-up cardiac power index and left ventricular work index were log-transformed to account for nonlinearity with respect to 30-day mortality. Due to its established power as a clinically relevant independent predictor of mortality, age was entered into the models as a permanent predictor of mortality and adjustment factor. Indexes of the hemodynamic parameters (ie, cardiac index, cardiac power index, systemic vascular resistance index, SVI, SWI, and left ventricular work index), but not the unindexed measures, were entered one at a time into multivariate models paired with age. For baseline variables, a stepwise backward logistic regression analysis was performed with the following variables included: heart rate tertiles, PCWP, cardiac index, cardiac power index, SVI, SWI, and left ventricular work index. For follow-up variables, a stepwise backward logistic regression analysis was performed with the following variables included: heart rate tertiles by treatment group interaction, systolic BP, mean arterial pressure, PCWP, cardiac index tertiles, log-transformed cardiac power index, SWI, SVI, and log-transformed left ventricular work index. SVI and SWI were then compared in models to all other hemodynamic data to determine the strength of these predictors assessed by p value, odds ratio, and the c-statistic. For baseline variables, SVI plus all other single baseline hemodynamic variables with p < 0.20 in univariate analysis were analyzed in individual models adjusted for age. For follow-up variables, SWI and SVI plus all other single follow-up hemodynamic variables with p < 0.05 in univariate analysis were analyzed separately in individual models adjusted for age. In addition, logistic regression analysis was used to examine the relationship between in-hospital mortality and hemodynamic parameters, and to display predicted mortality as a function of SVI both at baseline and follow-up and of SWI at baseline. All p values were two-tailed, and a p value of < 0.05 was considered to statistically significant. Analyses were conducted using statistical software packages (SAS, version 8.2; SAS Institute; Cary, NC; and S-Plus 6; Insightful Corporation; Seattle, WA).