Indirect methods of estimating infant and child mortality in sub-Saharan Africa

Application of the Brass child survival method to data from sub-Saharan Africa gives an upwardly biased estimate of 1q0 and 2q0. Comparison between directly and indirectly estimated rates derived from the reports of the same women for comparable time periods for these indices also show wide discrepancy. Misspecification of the model age patterns of mortality used in the derivation of the multiplying factors, homogeneous mortality assumption by age of mother and birth order, selectivity by age at first marriage and for the 2q0, the high proportion of the children being borne when the 20-24 years old women were teenagers is assumed to account for the observed discrepancy. This thesis attempts to examine the various sources of discrepancy using the 1988 DHS for Zimbabwe and Ghana as a case study.Given the limitations of the data, it is evident that the existing empirical model life tables and the parametric models don't adequately describe the age pattern of mortality in the region. Moreover, the evidence for the generally accepted assumption of regional homogeneity of age pattern of mortality is weak, although the assumption that the mortality of a country tends to decrease according to a constant pattern appears, to some extent, to hold. However, the age pattern of childhood mortality is better modelled by the proportional hazards model.A hazards model analysis also does not support the frequently accepted assumption that the upward bias in the indirectly estimated 2q0 is because a high proportion of the children ever born to women 20-24 were borne when they were 15-19 years of age and because they are selected by age at first marriage. However, a short birth interval among younger mothers appears to explain the discrepancy. The analysis also shows that the effect of those variables that cannot be parameterised are more important than the commonly stated causes, which implies the urgent need for an alternative data collection system.