Despite the compelling reasons for advance directives and their endorsement by the public and medical professions, little is known about their actual use and impact on site of death. This study was conducted to examine the role of advance directives and other "drivers" of hospitalization of the long-term care end-of-life patient. The medical records of 100 deceased consecutive nursing home residents, stratified by site of death (skilled nursing facility or acute care hospital), were reviewed by a team of geriatric researchers to obtain patient information in the following domains: sociodemographic, advance directives, transfer and death information, patient diagnoses at admission, discharge, and other time intervals; medication usage and signs and symptoms precipitating death. Severity of illness was assessed using the Cumulative Illness Rating Scale-G (CIRS-G). In testing for differences between patients by site of death, sociodemographic variables (gender, age, race, payer at discharge, cognitive capacity) did not significantly differ between the two groups of patients. Strong similarities between the groups were also found in terms of severity of illness and medication usage. Significantly higher proportions of patients dying in the nursing home had specific advance directives (do not resuscitate, do not intubate, do not artificially feed, do not hydrate, and do not hospitalize), as opposed to those dying in the hospital. The findings of this study demonstrate the impact of the explicit advance directive on the decision to transfer the patient to the acute care setting at the end of life.