The reduced incidence of mortality observed in patients receiving albumin was confirmed in a multivariable analysis. Following adjustment for the effect of 16 covariates (gender, age, COPD, gastric ulcer, obesity, PVD, number of CABGs, mammary graft, valve procedure, ED admission, hospital teaching status, hospital volume, diabetes, hypertension, length of stay, payer status), multivariable logistic regression analysis showed that albumin use was associated with 25% lower odds of mortality compared to nonprotein colloid use (odds ratio [OR], 0.80; 95% confidence interval [CI], 0.67 to 0.96) [Table 2
. Female gender, age, low-volume hospital, ED admission, simultaneous valve procedure, and PVD independently increased the risk of mortality, whereas mammary grafting reduced mortality. Diabetes, COPD, gastric ulcer, number of bypass grafts, Medicare status, and teaching status of the hospital were also included in the model; however, they were not independent predictors of mortality (their parameter estimates were not significant at 5% level). The parameter estimates of hypertension and teaching status of the hospital despite being significant were not robust at different stages of modeling.