Another potentially controversial finding in the study by Quill and colleagues4 is that black race varied from being associated with a fivefold decrease in the odds of limiting life-sustaining treatments in some ICUs to a 2.5-fold increase in the odds of limiting life-sustaining treatments in other ICUs. This contrasts the finding that other demographic factors, such as advanced age and female sex, are consistently associated with increased odds of deciding to limit life-sustaining treatments across ICUs. Importantly, the race finding may be influenced by sampling error because only one of the 15 ICUs with a point estimate for the odds of limiting life support among blacks > 1 was statistically significant. Ultimately, the variation between ICUs in their probability of limiting life-sustaining treatments is more likely to determine the treatment that black patients in those ICUs receive than their treatment compared with whites in the same ICU because of the unequal distribution of blacks across hospitals.9 Furthermore, differences in sociodemographics, such as education, insurance, and religion, may confound the relationship between race/ethnicity and intensity of care at the end of life. Importantly, if confounding were to explain the finding that blacks are more likely to be treated intensively than whites in some ICUs and less likely to be treated intensively than whites in other ICUs, then the third variable, such as religious beliefs predicting preferences not to limit life support, would have been more common among blacks in the former but more common in whites in the latter. Although we often invoke explanations like confounding to explain disturbing findings by race, the real explanations are undoubtedly more complicated because important differences in end-of-life care in the ICU by patient race/ethnicity are not explained by socioeconomic variables.10 Ultimately, to provide the most appropriate intensity of care at the end of life, we must be able to separate racial/ethnic differences in preferences for care that are based on the values of fully informed patients and families from differences due to disparities in the way end-of-life care is discussed or delivered. Differences based on the preferences of informed patients and families should be honored and incorporated; differences based on disparities in the way care is discussed or delivered should be eradicated.