Abstract A series of studies were performed to investigate the ability of rapid antigen-detection kits in diagnosis of influenza. Clinical specimens were collected from a total of more than 3000 cases of influenza-like illness who visited out-patient clinics in Sendai and Fukuoka, Japan, during the 2001–2002 and 2002–2003 influenza seasons. Each clinical specimen was examined with a kit and virus isolation, and results were analyzed retrospectively by several points of interest: sensitivity of the kits based on the virus isolation and the sensitivity by type and subtype of influenza, kinds of specimens, the time course of influenza, and variation of sources of the specimens. Thus, the sensitivities of a kit were 83%, 52% and 58% in general for influenza A(H3N2), A(H1N1), and B, respectively, and those of another kit were 76%, 71%, and 62%, respectively, in 2001–2002 influenza season. The sensitivity using nasal swabs was better than that using throat swab, though only 5% up. Specimens tended to be less sensitive when collected shortly after the onset of illness, and the sensitivity seemed quite varied (17–92%) by source of specimens: staff who collect the specimens, and/or by population of the patients.