In the two-week period November 13-27, 1984, 12 patients died in a 54-bed nursing home in Florida; based on previous mortality patterns, 2.5 deaths would have been expected for the whole month. There was no similar increase in deaths in November 1984 and no comparable monthly death rate for any of 69 nursing homes in the same county from 1976-84. Comparison of the 12 deaths in November with 28 deaths that occurred during the previous 10 months and with 31 surviving patients who were continuously present in the nursing home between November 12-28, 1984 revealed that the patients who died in November were more likely to have had onset of the terminal event during the night shift, had a recent visitor, and had an admitting diagnosis of organic brain syndrome. The abrupt increase in the death rate for November 1984 was not associated with a measurable change in population characteristics, an outbreak of infectious disease, or changes in procedures or the environment. Reviews of employee schedules revealed a consistent and strong association between the duty times of two nurses and the onsets of the terminal episode and the times of patient deaths. Continuing epidemiologic surveillance of adverse outcomes in nursing homes is recommended.