In a seminal article, Wenger et al¹ describe the epidemiology of do-not-resuscitate (DNR) orders in a nationally representative sample of 14 008 Medicare patients hospitalized with congestive heart failure, myocardial infarction, pneumonia, stroke, or hip fracture during 1981 to 1982 or 1985 to 1986. Their sample was well designed and their analytic methods were appropriate. They found important differences in the use of DNR orders related to gender, race, insurance status, and hospital characteristics. However, they did not adequately adjust for several factors that appropriately influence the use of DNR orders.

Wenger and colleagues state that "the use of the DNR order should reflect patient preferences and the expected quality and quantity of a patient's life."¹ However, they were unable to estimate "expected quality" of life after discharge. Dementia, ambulation, and incontinence at admission are certainly important predictors, but they do not capture changes in functional status during