Danzhen You

Statistics and Monitoring Specialist,

Statistics and Monitoring Section

Division of Policy and Planning

UNICEF

Postal Address:Three United Nations Plaza, New York, NY 10017

Telephone: 1 212-326-7235

E-mail: dyou@unicef.org

Tessa Wardlaw

Chief, Statistics and Monitoring Section and Associate Director, Policy and Practice

United Nations Children’s Fund

Postal Address:Three United Nations Plaza, New York, NY 10017

Telephone:212 326 7243

E-mail: twardlaw@unicef.org

The infant mortality rate is the probability (expressed as a rate per 1,000 live births) of a child born in a specified year dying before reaching the age of one if subject to current age-specific mortality rates.

A live birth is the complete expulsion or extraction from its mother of a product of conception, irrespective of the duration of the pregnancy, which, after such separation, breathes or shows any other evidence of life—such as beating of the heart, pulsation of the umbilical cord, or definite movement of voluntary muscles—whether or not the umbilical cord has been cut or the placenta is attached. Each product of such a birth is considered a live birth.

The methods used to calculate the infant mortality rate (IMR) depend on the type of data available. In practice, data can be obtained from vital registration systems, sample registration systems, national population censuses, and/or household surveys. When data collected via vital registration systems is of good quality, the IMR can be directly estimated by observing the survival status of different cohorts along time and to specific ages since the moment of birth. IMR can be derived from household survey data using direct or indirect methods.

The direct method uses data collected on birth histories of women of childbearing age and produces the probability of dying before age one for children born alive, among women of childbearing age, during five year periods before the survey (0-4, 5-9, etc.). Direct methods require each child’s date of birth, survival status, and date or age at death. This information is typically found in vital registration systems and in household surveys that collect complete birth histories from women of childbearing age. Birth histories include a series of detailed questions on each child a woman has given birth to during her lifetime, including the date the child was born, whether or not the child is still alive, and if the child has died, the age at death.

The indirect method uses the Brass method, named after its original developer, William Brass, which converts the proportion dead of children ever born reported by women in age groups 15-19, 20-24,… 45-49 into estimates of probability of dying before attaining certain exact childhood ages. Brass’s method assumes that the age of the mother can serve as a proxy for the age of her children and thus for how long they have been exposed to the risk of dying.

Indirect methods require less detailed information that is available in censuses and general surveys, including the total number of children a woman has ever borne, the number who survive and the woman’s age (or the number of years since she first gave birth). However, indirect methods require model life tables to adjust the data for the age pattern of mortality in the general population. Finding an appropriate model life table can be challenging, since the Coale and Demeny model life tables are derived largely from the European experience.

Vital registration systems are the preferred source of data on infant mortality because they collect information prospectively and cover the entire population. However, many developing countries lack fully functioning vital registration systems that accurately record all births and deaths. Thus, household surveys, such as Demographic and Health Surveys (DHS) and Multiple Indicator Cluster Surveys (MICS), have become the primary source of data on infant and child mortality in developing countries; but there are some limits to their quality.

Estimates obtained from household surveys have attached confidence intervals that need to be considered when comparing values along time or across countries. Similarly, these estimates are often affected by no-sampling errors that may affect equally recent levels and trends of IMR.

Like census data, survey data on infant mortality may omit births and deaths, include stillbirths along with live births, and suffer from survivor selection bias and age truncation. Direct estimates of infant mortality based on survey data may also suffer from mothers misreporting their children’s birth dates, current age or age at death—perhaps more so if the child has died. The heaping of deaths at age 12 months is especially common. Age heaping may transfer deaths across the one-year boundary and lead to underestimates of infant mortality rates.

Global figures produced by the inter-agency group for infant mortality estimation (IGME) may differ from the ones produced at the country level for different reasons. The global estimates use all available data obtained from different sources (vital registration, census, and household surveys) to produce estimates that represent trends and levels of infant mortality in the country. This is done by applying a regression model, described in the section below. Country estimates, on the other hand, can be obtained from just one specific source (more often from household surveys such as DHS), a combination of data sources, or from using different estimation methods.

The first step in the process is to proactively seek out all possible sources of data, including vital registration systems, national censuses, household surveys conducted by global programmes, and multi-purpose surveys conducted without international sponsorship.

To seek out national data sources that might be overlooked, UNICEF conducts an annual exercise called the Country Reports on Indicators for the Goals (CRING)1. CRING gathers recent information for all indicators regularly reported on by UNICEF, including the infant and under-five mortality rates.

Different data sources and calculation methods often yield widely differing estimates of infant mortality for a given time and place. In order to reconcile these differences, UNICEF developed, in coordination with WHO, the WB and UNPD,2 an estimation methodology that minimizes the errors embodied on each estimate and harmonize trends along time.3 Since the estimates are not necessarily the exact values used as inputs for the model, they are often not recognized as the official IMR estimates used at the country level. However, as mentioned before, these estimates minimize errors and maximize the consistency of trends along time. Applying a consistent methodology also allows comparisons between countries, despite the varied number and types of data sources.

After plotting all available values for infant and under-five mortality, analysts use weighted least squares to fit a multi-spine regression line to the data points and extrapolate the trend to the present. The use of weights allows analysts to make a judgment about the relative quality of each data set and how representative it is likely to be of the population. The last step is to decide which set of estimates (for infant mortality or under-five mortality) is more consistent and to use a model life table to derive the other set of estimates from it.

Additional details of the methodology 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.

http://mdgs.un.org/unsd/mdg/Resources/Attach/Capacity/Ind%204-1.pdf

The most recent IMR estimates produced by the IAGME are based on the extrapolation of observed trends obtained from observed IMR values derived from different data sources and methods. This process is described in the “Process of obtaining data” section, above.

Current estimates of IMR are generally based on empirical data from several or even many years before. Vital registration data are available on a yearly basis but often are published at the country level with a lag of 2 or more years. Unfortunately, vital statistics are unreliable in most developing countries. Population censuses are regularly conducted every ten years and results are published within 1-3 years after the population count. Household surveys, such as DHS and MICS, are in general implemented every 3-5 years with results published within a year of field data collection. On average, the most recent IMR estimates from household surveys refer to 2.5 years before the time of the survey or 3.5 years at the moment of publication of findings.

The lag time between the collection and publication of data is generally two years for vital registration, one year for surveys, and one to three years for censuses (Child Mortality Coordination Group, 2006). IMR estimates from the IGME refer to the year before the year the respective agency publishes (in the case of UNICEF for example, the 2008 State of the World’s Children published in November 2007 includes IMR estimates corresponding to the year 2006). Recent estimates are consistent with the trend observed during the last 30 years.

IGME estimates are updated annually.

IMR estimates are produced and presented by region and globally only if data are available for at least 50% of the region or the total population of the countries considered. Estimates are not included if this rule is not accomplished.

Yearly estimates of infant mortality rates are published in annual reports, at the end of the calendar year, by UNICEF in The State of the World’s Children and by The World Bank, in World Development Indicators. They are also included in the World Health Organization’s digest of World Health Statistics.