Academic journal article Demographic Research

Global Estimation of Neonatal Mortality Using a Bayesian Hierarchical Splines Regression Model

Academic journal article Demographic Research

Global Estimation of Neonatal Mortality Using a Bayesian Hierarchical Splines Regression Model

Article excerpt

(ProQuest: ... denotes formulae omitted.)


When evaluating a country's progress in reducing child mortality, it is important to obtain accurate estimates, be able to project mortality levels, and have some indication of the uncertainty in the estimates and projections. In practice, obtaining reliable mortality estimates is often most difficult in developing countries where mortality is relatively high, well-functioning vital registration systems are lacking, and the data that is available is often subject to large sampling errors and/or of poor quality. This situation calls for the use of statistical models to help estimate underlying mortality trends.

In recent years, much of the focus in monitoring child mortality has been on assessing changes in the under-5 mortality rate (U5MR), which refers to the number of deaths before the age of 5 per 1,000 live births. The focus was driven by Millennium Development Goal (MDG) 4, which called for a two-thirds reduction in under-5 mortality between 1990 and 2015. A report on MDG progress released in 2015 by the United Nations showed that, although this target was not met in most regions of the world, notable progress has been made (UN 2015). The global U5MR is less than half of its level in 1990, and despite population growth in developing regions, the number of deaths of children under 5 has declined. Reducing the U5MR continues to be a priority as part of the Sustainable Development Goals (SDG), which replaced the MDGs in 2015. Goal 3 of the SDG includes reducing the U5MR to at least as low as 25 deaths per 1,000 live births in all countries by 2030 (UN 2017).

As the U5MR decreases, the share of neonatal deaths, i.e., deaths occurring in the first month, tends to increase. Globally, the estimated share of under-5 deaths that were neonatal in 2015 was 45%, a 13% increase from 1990 (IGME 2015). Indeed, in most regions of the world, the majority of under-5 deaths are neonatal; for example, the share is 56% in developed regions, 51% in Latin America and the Caribbean, and 54% in Western Asia. The share is still less than 50%, however, where the U5MR is relatively high; for instance, in sub-Saharan Africa the share is only 34%.

The neonatal equivalent to the U5MR is the neonatal mortality rate (NMR), which is defined as the number of neonatal deaths per 1,000 live births. The increasing importance of neonatal deaths in child mortality has warranted increased efforts in monitoring NMR in addition to the U5MR (e.g., Bhutta et al. 2010; Lawn, Cousens, and Zupan 2004; Lozano et al. 2011). Goal 3 of the SDG explicitly includes a neonatal target, with the aim to reduce the NMR to at least as low as 12 deaths per 1,000 live births in all countries by 2030 (UN 2017).

The United Nations Inter-Agency Group for Child Mortality Estimation (IGME) publishes estimates of NMR for all 195 UN member countries (IGME 2015), and these estimates are used to monitor global levels and trends in NMR over time. Until 2014, IGME used a statistical model to obtain estimates for countries without high-quality vital registration data that uses the U5MR as a predictor (Oestergaard et al. 2011). While the method has worked well to capture the main trends in the NMR, it has some disadvantages. Most notably, trends in NMR within a country are driven by the U5MR trends, rather than being specifically influenced by the NMR data.

In this paper, we present a new model for estimating the NMR for countries worldwide, which overcomes some of the concerns with the previous IGME NMR model. We use a penalized splines regression model within a Bayesian hierarchical framework to estimate and project the NMR and to obtain uncertainty around these estimates and projections. In the model, the relationship between NMR and U5MR is used to inform estimates, and the spline regression model is used to capture country-specific trends. From the point of view of modeling mortality levels across countries, a Bayesian approach offers an intuitive way to share information across different countries and time periods, and a data model can incorporate different sources of error into the estimates. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed


An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.