Objective: To examine differences in end-of-life decision-making in patients dying at home, in a hospital or in a care home. Design: A death certificate study: certifying physicians from representative samples of death certificates, taken between June 2001 and February 2002, were sent questionnaires on the end-of-life decision-making preceding the patient’s death. Setting: Four European countries: Belgium (Flanders), Denmark, Sweden, and Switzerland (German-speaking part). Main outcome measures: The incidence of and communication in different end-of-life decisions: physician-assisted death, alleviation of pain/symptoms with a possible life-shortening effect, and non-treatment decisions. Results: Response rates ranged from 59% in Belgium to 69% in Switzerland. The total number of deaths studied was 12 492. Among all non-sudden deaths the incidence of several end-of-life decisions varied by place of death. Physician-assisted death occurred relatively more often at home (0.3–5.1%); non-treatment decisions generally occurred more often in hospitals (22.4–41.3%), although they were also frequently taken in care homes in Belgium (26.0%) and Switzerland (43.1%). Continuous deep sedation, in particular without the administration of food and fluids, was more likely to occur in hospitals. At home, end-of-life decisions were usually more often discussed with patients. The incidence of discussion with other caregivers was generally relatively low at home compared with in hospitals or care homes. Conclusion: The results suggest the possibility that end-of-life decision-making is related to the care setting where people die. The study results seem to call for the development of good end-of-life care options and end-of-life communication guidelines in all settings.