In recent years a significant increase in the incidence of Serratia marcescens infections was noted at the Chang Gung Memorial Hospital, Taoyuan, Taiwan. A review of laboratory (1991 to 2002) and infection control (1995 to 2002) records showed the possibility of an extended epidemic of nosocomial urinary tract infections (UTIs) caused by S. marcescens. Therefore, in 1998 and 1999, 87 isolates were collected from patients with such infections and examined and another 51 isolates were collected in 2001 and 2002. The patients were mostly elderly or the infections were associated with the use of several invasive devices. S. marcescens was usually the only pathogen found in urine cultures in our study. Neither prior infections nor disseminated infections with the organism were observed in these patients. Resistance to most antibiotics except imipenem was noted. Two genotyping methods, pulsed-field gel electrophoresis and infrequent-restriction-site PCR, were used to examine the isolates. A total of 12 genotypes were identified, and 2 predominant genotypes were found in 72 (82.8%) of the 87 isolates derived from all over the hospital. However, 63.9% of the isolates of the two genotypes were from neurology wards. A subsequent intervention by infection control personnel reduced the infection rate greatly. The number and proportion of the two predominant genotypes were significantly reduced among the 51 isolates collected in 2001 and 2002. Thus, a chronic and long-lasting epidemic of nosocomial UTIs caused by S. marcescens was identified and a successful intervention was carried out. Both a cautious review of laboratory and infection control data and an efficient genotyping system are necessary to identify such a cryptic epidemic and further contribute to the quality of patient care.