To the Editor Gonzalez et al1 report that failure-to-rescue (FTR) rates in cardiovascular surgery were lower (better) in high-volume hospitals than in low-volume hospitals. This is an important finding consistent with a large body of literature on FTR rates that has already shown its association with numerous hospital and staffing characteristics.2- 5 However, unlike the traditional measure of FTR rates,2,4 Gonzalez et al1 did not count every death of a patient undergoing coronary artery bypass grafting (CABG). As reported, 75 391 of 119 434 patients (63%) in their study1 were admitted for CABG, and among these patients, those who experienced an acute myocardial infarction (AMI) (before or after surgery) and subsequently died were not counted as a death, and therefore excluded from their FTR analysis. Apparently, Gonzalez et al1 have chosen not to count deaths (and their associated failures) because they did not consider an AMI a reliable complication—they had difficulty using Medicare Provider Analysis and Review data to determine whether the AMI occurred before or after surgery. Nowhere in their article or eTable is the total number of CABG patients with AMI reported, nor is the number of deaths ultimately excluded from the FTR analysis reported. From the perspective of any clinical trial reporting survival, the exclusion of a death because of measurement problems related to a complication prior to that death would be highly suspect. This exclusion should also be considered suspect when evaluating the FTR rate. To confirm the stability of their findings, Gonzalez et al1 should have provided a full account of (1) all deaths (not just patients who died without an AMI complication); (2) all patients with complications (including those who had an AMI); and (3) the FTR rate based on all patients. Such an analysis would be consistent with the original FTR theory2- 4 that includes all deaths; a complete description of the construction of the FTR measure is available in the literature.4 It is difficult to imagine any theory, in any medical or surgical context, that would advocate the exclusion of known deaths based on a perceived difficulty in the measurement of complications. If the exclusion of deaths is due to the inadequacy of the available data set for a CABG analysis, then possibly a different data set should have been used (as was done in a previous analysis of death, complication, and FTR among CABG patients published in JAMA3).