Medical records for hospitalized patients between 30 days and 30 years of age at two hospitals, one in each of two Colorado counties, were reviewed for possible cases of toxic shock syndrome conforming to one or both of two clinical case definitions. Patients with toxic shock syndrome were detected in each year from 1970 through 1982. Fifty patients with possible toxic shock syndrome were identified; 14 were males, five were premenarchal females, and the remaining 31 were females of menstrual age of whom 21 (42% overall) were menstruating at the time of illness. The combined annual toxic shock syndrome incidence rate ranged between 0.8/100,000 population less than or equal to 30 years of age (1974) and 9.1/100,000 (1980) with a maximum of 15.8/100,000 (1980) in females between the ages of 10 and 30. Incidence curves for males, females less than or equal to 30 years of age, and females of menstrual age (10-30 years) fluctuated in a statistically (p less than 0.02) nonrandom fashion, each peaking in 1979-1980 and declining in 1981. In a separate statewide voluntary case reporting system, 103 cases of toxic shock syndrome of all ages were reported by health care providers with onset from 1970-1982. A minimum estimate of 26 (26%) of these were not menstrually-associated, occurring in five females prior to menarche, three following menopause, three postpartum, and 15 males. Total cases reached an annual peak in 1980, declined somewhat in 1981, and increased during 1982 to levels comparable to 1980. Less than half (46%) of the strictly defined cases actively ascertained in the retrospective hospitalized population study had been reported voluntarily to the statewide system. Jurisdictions depending primarily on passive toxic shock syndrome case ascertainment techniques, even with periodic active solicitations, may not be reliably detecting the majority of cases, and those cases which are so reported may be limited by physician perception of syndrome spectrum.