We examined the charts of 911 nursing home patients in Hennepin County, Minnesota, to determine the prevalence of written do-not-resuscitate (DNR) orders. Information regarding demographic characteristics, and whether a surrogate decisionmaker was available and participated in the decision, was also collected. Twenty-seven percent of patients had DNR orders. Ninety percent of all patients had potentially available surrogate decisionmakers. However, for 31% of patients with DNR orders, there was no documentation of patient or surrogate participation in the DNR decision. Univariate analysis identified female sex; increased age, level of care (skilled versus intermediate), presence of a potential surrogate decisionmaker, and increasing length of time since nursing home admission as factors associated with presence of DNR orders. When a logistic regression model was used, increased age, increased length of time since nursing home admission, skilled versus intermediate level of care, and presence of a surrogate decisionmaker were independently associated with presence of DNR status. Several variables are independently associated with written DNR orders; their relationship to the factors physicians use in decision making requires further study.