Objective: To describe local geospatial variation and geospatial risk factors for child mortality in rural western Kenya.
Methods: We calculated under-5 mortality rates (U5MR) in 217 villages in a Health and Demographic Surveillance System (HDSS) area in western Kenya from 1 May 2002 through 31 December 2005. U5MRs by village were mapped. Geographical positioning system coordinates of residences at the time of death and distances to nearby locations were calculated. Multivariable Poisson regression accounting for clustering at the compound level was used to evaluate the association of geospatial factors and mortality for infants and children aged 1-4 years.
Results: Among 54 057 children, the overall U5MR was 56.5 per 1000 person-years and varied by village from 21 to 177 per 1000 person-years. High mortality villages occurred in clusters by location and remained in the highest mortality quintile over several years. In multivariable analysis, controlling for maternal age and education as well as household crowding, higher infant mortality was associated with living closer to streams and further from public transport roads. For children 1-4 years, living at middle elevations (1280-1332 metres), living within lower population densities areas, and living in the northern section of the HDSS were associated with higher mortality.
Conclusions: Childhood mortality was significantly higher in some villages. Several geospatial factors were associated with mortality, which might indicate variability in access to health care or exposure and transmission of infectious diseases. These results are useful in prioritising areas for further study and implementing directed public health interventions.