OBJECTIVE: To investigate the association, if any, between child mortality and distance to the nearest hospital. METHODS: The study was based on data from a 1-year study of the cause of illness in febrile paediatric admissions to a district hospital in north-east Tanzania. All villages in the catchment population were geolocated, and travel times were estimated from availability of local transport. Using bands of travel time to hospital, we compared admission rates, inpatient case fatality rates and child mortality rates in the catchment population using inpatient deaths as the numerator. RESULTS: Three thousand hundred and eleven children under the age of 5 years were included of whom 4.6% died; 2307 were admitted from <3 h away of whom 3.4% died and 804 were admitted from ≥ 3 h away of whom 8.0% died. The admission rate declined from 125/1000 catchment population at <3 h away to 25/1000 at ≥ 3 h away, and the corresponding hospital deaths/catchment population were 4.3/1000 and 2.0/1000, respectively. Children admitted from more than 3 h away were more likely to be male, had a longer pre-admission duration of illness and a shorter time between admission and death. Assuming uniform mortality in the catchment population, the predicted number of deaths not benefiting from hospital admission prior to death increased by 21.4% per hour of travel time to hospital. If the same admission and death rates that were found at <3 h from the hospital applied to the whole catchment population and if hospital care conferred a 30% survival benefit compared to home care, then 10.3% of childhood deaths due to febrile illness in the catchment population would have been averted. CONCLUSIONS: The mortality impact of poor access to hospital care in areas of high paediatric mortality is likely to be substantial although uncertainty over the mortality benefit of inpatient care is the largest constraint in making an accurate estimate.