A note on estimating the cause of death structure in high mortality populations.

Abstract

Strategies for reducing mortality should be based on an informed assessment of the underlying structure of causes of death. In the absence of an adequate vital registration scheme, cause of death information is generally derived from fragmentary data which are often limited to specific sectors of the population. Using the previous mortality experience of industrialized countries at higher levels of mortality and the cause of death returns of contemporary developing societies with reasonably accurate and complete mortality data, a linear model has been developed to estimate the cause-structure of mortality in a population from knowledge of the crude death rate alone. Through an appropriate weighting of age-cause-specific mortality rates, emphasis is placed on those causes of death likely to be of greatest concern in high mortality populations. Using the model parameters derived in this fashion, the technique is applied to estimate the cause-structure of mortality in Java. The resulting distribution of deaths differs markedly from that suggested by an older age-standard and is much more in accord with epidemiological opinion about likely causes of death based on rural health surveys, hospital records, and other health data. Regression parameters for estimating the cause of death structure during infancy and early childhood are also shown and may be used independently of the all-age regression estimates.