BACKGROUND: Estimates of the burden of invasive bacterial disease in sub-Saharan Africa have previously relied on selected groups of patients, such as inpatients; they are, therefore, probably underestimated, potentially hampering vaccine implementation. Our aim was to assess the incidence of bacteraemia in all children presenting to a hospital in Kenya, irrespective of clinical presentation or decision to admit. METHODS: We did a community-based observational study for which we cultured blood from 1093 children who visited a Kenyan hospital outpatient department. We estimated bacteraemia incidence with a Demographic Surveillance System, and investigated the clinical significance of bacteraemia and the capacity of clinical signs to identify cases. RESULTS: The yearly incidence of bacteraemia per 100,000 children aged younger than 2 years and younger than 5 years was 2440 (95% CI 1307-3573) and 1192 (692-1693), respectively. Incidence of pneumococcal bacteraemia was 597 (416-778) per 100,000 person-years of observation in children younger than age 5 years. Three-quarters of episodes had a clinical focus or required admission, or both; one in six was fatal. After exclusion of children with occult bacteraemia, the incidence of clinically significant bacteraemia per 100,000 children younger than age 2 years or 5 years fell to 1741 (790-2692) and 909 (475-1343), respectively, and the yearly incidence of clinically significant pneumococcal bacteraemia was 436 (132-739) per 100,000 children younger than 5 years old. Clinical signs identified bacteraemia poorly. INTERPRETATION: Clinically significant bacteraemia in children in Kilifi is twice as common, and pneumococcal bacteraemia four times as common, as previously estimated. Our data support the introduction of pneumococcal vaccine in sub-Saharan Africa.