We also believe the scoring system underestimates the effects of myocardial infarction on left ventricular performance, particularly factors such as refractory congestive heart failure or cardiogenic shock, which may develop later in the hospital course, after APACHE II scores have been assigned.
Our study identified and then validated a multiple logistic regression model of five variables at the time of intubation (Table 4) that can be used to predict mortality. The most important variable was systolic blood pressure. To enhance the utility and simplicity of the model, we analyzed a receiver operator characteristic curve to find a discrete, discriminating value. A systolic pressure of 130 mm Hg provided the best balance of sensitivity and specificity, and was chosen for the model. This choice did not change the predictive accuracy of the logistic regression equation.
Plotnick et aln and Goldberger et al also found that initial systolic blood pressure predicted mortality. Goldberger et al reported a 12.5 percent mortality among patients with systolic pressure 160 mm Hg or greater and a 26.7 percent mortality for those with lower pressures, while Plotnick et al, using 150 mm Hg, reported a difference in mortality of 3.4 percent vs 30.8 percent.