Does Anesthesia Contribute to Operative Mortality?-Reply

In Reply. — Kissin et al raise some interesting questions about the interpretation of variables estimated in multiple logistic regression models. The overall fit of a model is not altered by the choice of reference category. We chose to "dummy code" categories of anesthetic techniques using inhalation alone as the reference category. The Table shows that the relative odds of mortality can be seen to vary considerably from 0.36 to 5.00 depending on the technique chosen as the reference. We used "inhalation alone" as the reference technique because patients who had this technique were found to have the lowest mortality in the bivariate analysis and because it was the most commonly used technique. Our assessment of the importance of a variable, or in this case a group of like variables, was not by the magnitude of the relative odds but by how much they contributed to the overall fit of