A study was carried out to determine factors affecting place of death (home, hospital, nursing home or "other places") among all 426,115 resident deaths in Washington State during 1968-1981, using death certificate information. Sixteen percent of deaths occurred at home, 74% in institutions (51% in hospitals, 23% in nursing homes) and 9% at "other places." Age, marital status and cause of death all strongly affect place of death. Further, the effect of each factor was strongly dependent on the others. Sex had no effect on place of death after controlling for other factors. Elderly people died relatively more frequently in nursing homes, infants and middle aged people in hospitals and young adults in "other places." The frequency of deaths at home was quite constant by age. Hospitals were the most common place of death following both vascular disease (including heart attack) and neoplasms, and nursing homes were the most common place of death following cerebrovascular disease (including "stroke"). Race, socioeconomic status and urban or rural residents affected the place of death only slightly or not at all. The place of death pattern changed little during the time period 1968-1981, except for a slight increase in frequency of home deaths and a corresponding decrease in the frequency of deaths in other places. Among cancer patients, the likelihood of death at home was positively associated with longer periods of survival after diagnosis. Cancer patients of hospitals serving targeted populations, such as veterans, were relatively more likely to die in a hospital and less likely to die in a nursing home compared to other cancer patients, suggesting that the "targeted" hospitals are sometimes serving a nursing home function. There was a marked difference in the terminal cancer caseload by hospital. The number of cancer deaths per cancer diagnosis varied widely across hospitals (0.1 to 1.6) and was unrelated to size of the hospital or level of services offered. Intervention aimed at affecting place of death, such as increasing the number of deaths at home, will need to take account of the joint effect of age, marital status and disease.