Background: Do-not-resuscitate (DNR) orders are often active in patients with multiple comorbidities and a short natural life expectancy, but limited information exists as to how often these patients undergo high-risk operations and of the perioperative outcomes in this population. Methods: Using comprehensive inpatient administrative data from the Public Discharge Data file (years 2005 through 2010) of the California Office of Statewide Health Planning and Development, which includes a dedicated variable recording DNR status, we identified cohorts of DNR patients who underwent major cardiac or thoracic operations and compared them to age- and procedure-matched comparison cohorts. The primary study outcome was in-hospital mortality. Results: DNR status was not uncommon in cardiac (n=2,678, 1.1% of all admissions for cardiac surgery, age 71.6 ± 15.9 years) and thoracic (n=3,129, 3.7% of all admissions for thoracic surgery, age 73.8 ± 13.6 years) surgical patient populations. Relative to controls, patients who were DNR experienced significantly greater in-hospital mortality after cardiac (37.5% vs. 11.2%, p<0.0001) and thoracic (25.4% vs. 6.4%) operations. DNR status remained an independent predictor of in-hospital mortality on multivariate analysis after adjustment for baseline and comorbid conditions in both the cardiac (OR 4.78, 95% confidence interval 4.21-5.41, p<0.0001) and thoracic (OR 6.11, 95% confidence interval 5.37-6.94, p<0.0001) cohorts. Conclusions: DNR status is associated with worse outcomes of cardiothoracic surgery even when controlling for age, race, insurance status, and serious comorbid disease. DNR status appears to be a marker of substantial perioperative risk, and may warrant substantial consideration when framing discussions of surgical risk and benefit, resource utilization, and biomedical ethics surrounding end-of-life care.