Review of 2219 admissions to an intensive care unit at a large urban hospital from 1983 to 1985 revealed an epidemic of cardiac arrests during the evening shift from January 1984 to March 1985. Of the 88 evening-shift cardiac arrests during this time, one specific nurse (Nurse 14) was the care giver for 57 (65%). Eight of ten epidemic-period patients who experienced shift-specific circadian recurrences of cardiac arrests were patients of Nurse 14. No therapeutic intervention and no other health care provider was as strongly associated with evening-shift cardiac arrests as was Nurse 14. Multiple logistic regression analysis demonstrated that when risk of cardiac arrest was adjusted for age, sex, severity of illness, and postoperative status, patients of Nurse 14 were 47.5 times more likely to experience arrest than were other nurses' patients. An expert determined that, compared with other nurses, the cardiac arrests among Nurse 14's patients were more likely to be consistent with unexplained hyperkalemia, to be unexpected in timing, and to be inconsistent with the clinical course. The epidemic ceased when Nurse 14 left employment at the intensive care unit in March 1985. Epidemiologic surveillance of adverse outcomes in health care settings is recommended.