Objectives: To establish a chest pain observation unit, monitor its performance in terms of appropriate discharge after assessment, and estimate the cost per patient. Methods: Prospective, observational, cohort study of patients attending a large, city, teaching hospital accident and emergency department between 1 March 1999 and 29 February 2000 with acute undifferentiated chest pain. Patients were managed on a chest pain observation unit, entailing two to six hours of observation, serial electrocardiograph recording, cardiac enzyme measurement, and, where appropriate, exercise stress test. Patients were discharged home if all tests were negative and admitted to hospital if tests were positive or equivocal. The following outcomes were measured—proportion of participants discharged after assessment; clinical status three days after discharge; cardiac events and procedures during the following six months; and cost of assessment and admission. Results: Twenty three participants (4.3%) had a final diagnosis of myocardial infarction. All were detected and admitted to hospital. A total of 461 patients (86.3%) were discharged after assessment, 357 (66.9%) avoided hospital admission entirely. At review three days later these patients had no new ECG changes and only one raised troponin T measurement. In the six months after assessment, three cardiac deaths, two myocardial infarctions, and two revascularisation procedures were recorded among those discharged. The mean cost of assessment and hospital admission was £221 per patient, or £323 if subsequent interventional cardiology costs were included. Conclusions: The chest pain observation unit is a practical alternative to routine care for acute chest pain in the United Kingdom. Negative assessment effectively rules out immediate, serious morbidity, but not longer term morbidity and mortality. Costs seem to be similar to routine care.