Background
Child mortality is an important indicator of a country’s developmental status. Neonatal mortality and stillbirth shared a higher proportion of child deaths. However, in developing countries where there is no civil registration and most deliveries occur at home, it is difficult to measure the magnitude of neonatal mortality. Data from continuous demographic surveillance systems could provide reliable information. To this effect, the outputs in this analysis are based on a 22 year dataset from Butajira demographic surveillance site.

Methods
The Butajira Rural Health Programme was launched in 1987 with an objective of developing and evaluating a system for a continuous registration of vital events. The surveillance system operates in an open cohort. An event history analysis was carried out to calculate the yearly neonatal mortality and its association with selected covariates. Poisson regression model was used to elicit neonatal mortality risk factors.

Results
The trends of neonatal mortality equaled out at a higher level over the study period (P-value?=?0.099). There was a high burden of early neonatal mortality (incidence rate ratio 4.8 [4.5, 5.2]) with the highest risk of death on the first day of life 18 [16.6, 19.4]. In multivariate analysis, males 1.6(1.4-1.9), neonates born from poor mothers who had no oxen 1.2(1.0-1.3) lived in thatched houses 2.9(2.4-3.5) and a distance to a health facility 1.5(1.1-2.0) conferred the highest risk of neonatal deaths.

Conclusion
Despite an urgent need in reducing neonatal mortality which contributes to more than 40% to child mortality, no significant change was observed in Butajira. Death was significantly associated with sex of the child, socio-economic variables and physical access to hospital. Prevention strategies directed at reducing neonatal death should address policy and household and level factors, which significantly influence neonatal mortality in Butajira.