A multiethnic cohort of adult members of the Kaiser Permanente Medical Care Program (55300 men and 65271 women) was followed for 15 years (1979-93) to assess the association between total cholesterol and risk of infections (other than respiratory and HIV) diagnosed in the in-patient setting. Using multivariate Cox regression, total cholesterol was inversely and significantly related to urinary tract, venereal, musculo-skeletal, and all infections among men; and to urinary tract, all genito-urinary, septicaemia or bacteraemia, miscellaneous viral site unspecified, and all infections among women. The reduction of risk of all infections associated with a 1 S.D. increase in total cholesterol was 8% in both men (95% CI, 4-12 %) and women (95% CI, 5-11%). For urinary tract infections among men, as for septicaemia or bacteraemia and nervous system infections among women, the risk relation was restricted to persons aged 55-89 years. Nervous system infections were positively related to total cholesterol among women aged 25-54. In both genders, the significant inverse association with all infections persisted after excluding the first 5 years of follow-up. Collectively, these data are suggestive of an inverse association, although not entirely consistent, between total cholesterol and incidence of infections either requiring hospitalization or acquired in the hospital. Further research is needed to elucidate whether these associations are biologically plausible or represent uncontrolled confounding by unmeasured risk factors.