Global estimation of child mortality using a Bayesian B-spline bias-reduction model

摘要

The importance of estimating under five child mortality rate is discussed as being the barometer of well-being of children of any country. It is observed that the a well functioning of vital registration system is a key to develop a valid estimation model. However, due to the issues with data quantity and quality in most of the developing countries, the estimation of child mortality has become a challenging problem. Notwithstanding, though there are methods available for estimation of child mortality, it is discussed that these models have their own limitations and lack calibration. Therefore, in this paper it is proposed to develop an estimation procedure using Bayesian B-spline bias-reduction model. In this regard the role played by Bayesian penalized B-spline regression model in evaluating the levels and trends in the under five child mortality rate for all countries in the world is described. This model estimates biases in data series for all non-vital registration source types using a multilevel model to improve upon the limitations of current methods. For this purpose, the improved spline extrapolations are obtained via logarithmic pooling of the posterior predictive distribution of country-specific changes in spline coefficients with observed changes on the global level. It is highlighted that the proposed model can flexibly capture changes in under five child mortality rate over time, provides point estimates and credible intervals that take into consideration potential biases in data series and gives better model validation results compared to other estimation approaches. As a result, the proposed approach has been accepted by the United Nations Inter-agency Group for Child Mortality Estimation to generate estimates for all member countries. (30 refs.)