A descriptive epidemiology on an outbreak of tuberculosis (TB) in a long-term care unit of a mental hospital was conducted. A female inpatient of 60 years with pulmonary TB was reported to the local health department (LHD) from the hospital in July 1999. Her sputum was negative both by smear and culture. From then to Mar 2001, a total of ten TB cases were reported. All of them were inpatients or workers of the hospital. Among them, four cases turned out to be culture positive and for three out of them a Restriction Fragment Length Polymorphism (RFLP) analysis was performed. All three turned out to be the identical strain suggesting that the outbreak was derived from one index case. After November 1999, the active case findings were conducted by the LHD, however no case of possible source of the outbreak was found. On the other hand, the retrospective investigation revealed that a female inpatient (case Z) of 70 years must have been pulmonary TB. She had had respiratory symptoms such as severe cough and sputum for two years and consequently died of pneumonia in February 1999, five months before the onset of the TB outbreak. She had a thoracic CT scan test and a sputum PCR test just before her death in another outpatient clinic and turned out to have a cavity in a lung and to be PCR positive for Mycobacterium tuberculosis complex. However the result was never reported to the hospital nor to the LHD, because she died before the PCR test was completed. She had had close contact with all of the TB cases except one for over two years. Considering all these epidemiological results, case Z was suggested to be the source of this outbreak. To prevent this kind of TB outbreak, institutions like mental and/or long-term care units should carefully prepare a proper precaution plan against the nosocomial infection of TB. In addition, if two or more TB cases are reported from the same unit or institution, LHDs should pay special attention and investigate the possibility of nosocomial infection.