Child mortality estimation methods
Topics include: an overview of the child mortality estimation methodology developed by the UN IGME; methods used to adjust for bias due to AIDS; estimation of sex differences in child mortality; and more. The Collection was produced with support from UNICEF and the independent technical advisory group of the UN IGME.
Definition of indicators
Under-five mortality rate: Probability of dying between birth and exactly five years of age, expressed per 1,000 live births.
Infant mortality rate: Probability of dying between birth and exactly one year of age, expressed per 1,000 live births.
Neonatal mortality rate: Probability of dying during the first month of life, expressed per 1,000 live births.
Data sources and methodology
Generating accurate estimates of child mortality is a considerable challenge because of the limited availability of high-quality data for many countries. Vital registration systems are the preferred source of data on child mortality because they collect information as events occur and they cover the entire population. If registration coverage is complete and the systems function efficiently, the resulting child mortality estimates will be accurate and timely. However, many countries lack fully functioning vital registration systems that accurately record all births and deaths. Therefore, household surveys, such as the UNICEF-supported Multiple Indicator Cluster Surveys and the US Agency for International Development–supported Demographic and Health Surveys, have become the primary sources of data on child mortality in countries without functioning vital registration systems. These surveys ask women about the survival of their children, and they provide the basis of child mortality estimates for most of these countries.
The UN IGME seeks to compile all available national-level data on child mortality, including data from vital registration systems, population censuses, household surveys and sample registration systems. To estimate the under-five mortality trend series for each country, a statistical model is fitted to data points that meet quality standards established by the UN IGME and then used to predict a trend line that is extrapolated to a common reference year, set at 2012 for the estimates in this report. Infant mortality rates are generated by either applying a statistical model or transforming under-five mortality rates based on model life tables. Neonatal mortality rates are produced using a statistical model that uses under-five mortality rates as an input. These methods provide a transparent and objective way of fitting a smoothed trend to a set of observations and of extrapolating the trend from 1960 to the present.
Changes in the estimation process
The UN IGME continually seeks to improve its methods and may introduce changes from one year to the next. This year, a new estimation method is used for estimating and extrapolating the under-five mortality rate, referred to as the Bayesian B-splines bias-adjusted model or the B3 model. Compared with the Loess estimation approach that the UN IGME used in previous years, the B3 model better accounts for data errors, including biases and sampling and nonsampling errors in the data; it can better capture short-term fluctuations in the under-five mortality rate and its annual rate of reduction. Thus, it is better able to account for evidence of acceleration in the decline of under-five mortality from new surveys. Validation exercises show that the B3 model also performs better in projections. In 2012 the UN IGME produced estimates of the under-five mortality rate for males and females separately for the first time. In many countries fewer sources have provided data by sex than have provided data for both sexes combined. For this reason the UN IGME uses the available data by sex to estimate a time trend in the sex ratio (male : female) of child mortality, rather than estimate child mortality trends by sex directly from reported sex-specific mortality rates. This year, new Bayesian methods have been developed by the UN IGME for estimating sex ratios of child mortality, with a focus on identifying countries with outlying levels or trends.
More details on the data used in deriving estimates are available in CME Info (www.childmortality.org).
The full details of the methodology used in the estimation of infant and under-five mortality rates for 2006 are available in the following working paper: UNICEF, WHO, The World Bank and UN Population Division, ‘Levels and Trends of Child Mortality in 2006: Estimates developed by the Inter-agency Group for Child Mortality Estimation’, New York, 2007. Working Paper [PDF]
Other references include:
UNICEF, WHO, The World Bank, United Nations Population Division, Levels and Trends in Child Mortality: Report 2012.
UNICEF, WHO, The World Bank, United Nations Population Division, Levels and Trends in Child Mortality: Report 2011.
UNICEF, WHO, The World Bank, the United Nations Population Division, Levels and Trends in Child Mortality: Report 2010.
Hill, Kenneth, Rohini Pande, Mary Mahy (John Hopkins University) and Gareth Jones (UNICEF) (KH-98.1), Trends in Child mortality in the Developing World: 1960-1996 - Full publication [zip].