This paper examines the observed reversal in childhood mortality conditions in Kenya since the late 1980s, and the determinants of child mortality. Data from the 1998 Demographic and Health Survey are used to achieve these two objectives. Direct estimation techniques are used to examine regional levels and trends of mortality since the 1980s. It is clear from these analyses that regional mortality levels in Kenya are quite diversified. Nyanza has the highest mortality among children, while Central has the lowest. Although, mortality increased in the country in the early 1990s, most of the increases occurred in Nyanza province. Weibull hazard models, which also incorporate unobserved heterogeneity (frailty), are used to examine the correlates of childhood mortality. These hazard models clearly show that biodemographic factors are more important in explaining infant mortality, while the socioeconomic and socio-cultural factors (including hygienic factors) are more important in explaining child mortality. The results also show that the risk of death were higher for children born in the 1990s compared to those born before 1980. In spite of including all these factors into the model, the unobserved heterogeneity is still significant. The results suggest that mortality trends in the country are likely to be related to unmeasured factors especially HIV/AIDS prevalence.