After Independence in the early 1960s, child mortality in Kenya fell rapidly. Until around 1980, the under 5 mortality rate (U5MR), the probability of dying by age 5, fell at an annual rate of about 4 percent per annum. This rate of decline slowed in the early 1980s, to about 2 per cent per annum. Recent data from the 1998 Kenya
Demographic and Health Survey showed that, far from declining, the U5MR increased by as much as 25 percent from the late 1980s to the mid 1990s. This adverse trend coincided with a number of other adverse trends: stagnation in growth of per capita income, declining levels of immunization, falling school enrolment, and the emergence of an HIV/AIDS epidemic. On a more positive note, fertility fell by about 30 percent from the mid 1980s to the mid 1990s. Controversy surrounds the factors responsible for the increase in child mortality in the 1990s, and the objective of this paper is to clarify the situation. Data from the 1993 and 1998 DHSs have been merged into a single data set, and multivariate analysis used to examine the factors associated with mortality risks in childhood. Dummy variables were used to represent different three-year time periods, from 1984-86 to
1996-98. Socioeconomic controls, including mother¹s education, an indicator of household wealth, urban/rural residence, and indicators of health service utilization, plus controls for reproductive dynamics such as age of mother at the birth, birth order, sex and preceding birth interval, were developed. In addition, an indicator of the HIV epidemic, the prevalence of HIV in the district of birth at the time of each child¹s birth, was developed. With no controls, the models confirmed an increase in mortality of about 25 percent. Including socioeconomic and biodemographic controls tended to strengthen the upward trend in mortality; in other words, had there been no
changes in these factors, child mortality would have been expected to decline. Introducing controls for health variables immunization, pregnancy and delivery care, prevalence of childhood diseases and maternal and child malnutrition ­ also did not alter the underlying trends substantially. Thus rising child mortality could not be
explained by socioeconomic, biodemographic or health status factors. Including the prevalence of HIV in the models, however, changed the underlying trends fundamentally, from sharp increase to monotonic decline. Although models of this sort cannot demonstrate causation, only association, the HIV epidemic appears to be the most probable cause of the recent increases in child mortality in Kenya. Of the health variables, the only one found to be significantly protective was immunization coverage.