Bulletin of the World Health Organization

Infant and under-five mortality in Afghanistan: current estimates and limitations

Kavitha Viswanathan a, Stan Becker b, Peter M Hansen c, Dhirendra Kumar d, Binay Kumar d, Haseebullah Niayesh e, David H Peters a & Gilbert Burnham a

a. Department of International Health, Johns Hopkins Bloomberg School of Public Health, 615 N Wolfe Street, Baltimore, MD, 21205, United States of America (USA).
b. Department of Population, Family and Reproductive Health, Johns Hopkins Bloomberg School of Public Health, Baltimore, USA.
c. GAVI Alliance, Geneva, Switzerland.
d. Indian Institute of Health Management Research, Rajasthan, India.
e. Department of Monitoring and Evaluation, Ministry of Public Health, Kabul, Afghanistan.

Correspondence to Kavitha Viswanathan (e-mail: kviswana@gmail.com).

(Submitted: 20 June 2009 – Revised version received: 22 November 2009 – Accepted: 24 November 2009 – Published online: 07 April 2010.)

Bulletin of the World Health Organization 2010;88:576-583. doi: 10.2471/BLT.09.068957

Introduction

According to the State of the world’s children, Afghanistan’s under-five mortality rate of 257 deaths per 1000 live births is the third highest in the world, surpassed only by the rates for Angola and Sierra Leone1 Cited infant mortality rates for 2005 are also very high, at 165 deaths per 1000 live births.1 However, these are model-based projections that have not been updated since at least 1993,2 despite widespread changes in Afghanistan over the intervening period, such as the implementation of the Basic Package of Health Services in 2003 and 2004. Also, these projections and other estimates are derived largely from a seminal demographic study conducted between 1972 and 1974 – the National Demographic and Family Guidance Survey (NDFGS). Other demographic and health surveys have been conducted in Afghanistan in the 30 years since the NDFGS but are of limited usefulness in explaining demographic patterns in the country because of methodological constraints and geographical coverage. More recently, the Afghanistan Health Survey (AHS) 2006, with a multistage cluster design, was conducted to gather information on maternal and child health, child survival, family planning, health care use and related expenditures in rural Afghanistan. The aim was to determine how much progress had been made in delivering the Basic Package of Health Services implemented in 2003 and 2004.

Efforts to estimate mortality in Afghanistan, as in many other developing countries, face numerous constraints. Women seldom know their exact ages or birth dates, and there is extensive age-heaping (i.e. rounding of actual ages to end in 5 or 0). Underreporting of deaths is also a common problem; in any retrospective survey, the longer the time since a child’s death, the greater the likelihood the death is not reported.3 Many areas are geographically difficult to reach and unsafe. Literate female interviewers who can travel to remote areas are few and, in the absence of female surveyors, male interviewers are often unable to interview female respondents, especially in conservative Pashtun areas.

This study had three objectives: to examine historical estimates of infant and under-five mortality in Afghanistan; to describe the mortality results derived from the AHS, which are the first estimates based on current population data; and to discuss common methodological challenges in Afghanistan that constrain data quality and hinder retrospective estimation of mortality.

Methods

Historical estimates

To examine historical estimates of infant and under-five mortality in Afghanistan, we compared the results (overall and rural) of surveys conducted in Afghanistan between 1972 and 2003, including the NDFGS, the partially completed 1979 Afghanistan Census, and the Multiple Indicator Cluster Survey of 1997 and 2003.

Current estimates

The sampling frame used to select villages for the survey was obtained from the Central Statistics Office in July 2006. The sampling frame, a precensus household listing conducted between 2003 and 2005, included over 45 000 enumeration units in rural and urban areas. The AHS included a total of 425 sampled clusters (enumeration units), of which 397 were completed; the remaining 28 were not completed due to security-related reasons. The final sample included 8278 households. Data collection started in mid-September 2006 and was completed by late December 2006.

The AHS covered 29 of the 34 provinces of Afghanistan. The provinces of Helmand, Kandahar, Nuristan, Uruzgan and Zabul, as well as some districts in other provinces, were excluded before sampling because they were unsafe for survey teams and monitors. The six major cities of Herat, Jalalabad, Kabul, Kandahar, Kunduz and Mazar were also excluded because the primary interest of the Ministry of Public Health was to gather information on priority indicators for rural areas. Thus, the results from the AHS are representative of 72% of the rural population of Afghanistan.

Two-stage cluster sampling was used to select households. In the first stage, clusters were selected from a list of all enumeration units with probability proportional to size. In the second stage, a compact segment method was used to sample households within a cluster.4 The enumeration unit was segmented into groups, each containing an equal and predetermined number of compounds. One segment was randomly chosen. A separate household listing exercise was not conducted due to time and financial constraints, and the number of households within each compound was not known; therefore, all compounds in a segment were listed and a fixed number of compounds were selected based on simple random sampling. If a compound contained multiple households, all of them were interviewed to ensure that all households in a segment had the same probability of being selected.

The survey interviewed: (i) all women who had ever been married and who were between the ages of 10 and 49 years, and (ii) unmarried primary caretakers at least 18 years of age in charge of children between the ages of 0 and 59 months whose mothers were either dead or no longer living in the household.

Ethical approval was obtained from the institutional review boards at the Ministry of Public Health and the Johns Hopkins Bloomberg School of Public Health, United States of America (USA).

The Brass indirect method of estimation, specifically the Trussell variant for West model life tables, was used to calculate the infant and under-five mortality rates.3,5 The Brass technique uses age patterns of fertility (parity) and the proportions of children dead among children ever born to women of different ages, and converts them into the probabilities of a child dying before reaching certain ages. These probabilities of dying are, in turn, translated into mortality estimates by using model life tables. The infant and under-five mortality rates were estimated using the QFive programme (United Nations Population Division, New York, USA).6

All women between the ages of 15 and 49 years were included in calculations of mortality. Each eligible respondent was asked questions on the total number of children that she had delivered and the total number of children that were dead.

Given evidence of underreporting of deaths of children in women aged 25 years and above, the mortality analysis was confined to women aged 15–19 and 20–24 years. To detect unusual data patterns that could be masked by larger age groupings, further data-quality checks were conducted in these age groups using mothers’ exact age. The data were examined on overall proportion dead, proportion dead by sex, sex ratio at birth, and ratio of proportion of dead boys to proportion of dead girls by age of mother. Sex-specific differences in underreporting of births and deaths were also examined by parity of mother. The natural sex ratio quotient of 0.512 (i.e. a total of 105 boys born for every 100 girls born) was used to calculate the expected numbers of girls and boys by parity, and these values were then compared to the observed numbers. Differential reporting of births was also examined by parity for women who had only girls, only boys, or mixed births. Observed numbers lower than those expected could reflect underreporting of births, which could in turn be associated with underreporting of deaths if respondents are more likely to report the births of surviving than of dead children. Therefore, sensitivity analyses were conducted to assess the effect of different proportions of dead children among underreported births on the estimated infant and under-five mortality rates in this study population.

Results

Historical estimates

Table 1 shows the infant and under-five mortality rates estimated from surveys conducted in Afghanistan between 1972 and 2003.

Overall mortality estimates

Table 2 shows, for each 5-year age group of women, the percentage who were married, the number of children ever born and the percentage of children still living.

Table 3 gives estimates of infant and under-five mortality rates by age of mother, with their reference dates as produced by the QFive programme. The low mortality rates estimated for early reference years compared to more recent years suggests underreporting of child deaths by older women. The final mortality rates presented are weighted estimates from two age groups (15–19 and 20–24 years) based on the proportion of ever-married women in each group. The reference date for the estimates is also a weighted date. The estimated infant and under-five mortality rates were lower for girls than for boys (Table 4).

Sex-specific differences

Table 5 shows a breakdown of the sex ratio at birth and the ratio of the proportion of dead boys to proportion of dead girls by parity of the mother for women aged 15–24 years. Column 3 of the table shows the sex ratio at birth, which was highly skewed towards boys, especially for lower parities (the typical range of the sex ratio at birth is 104–107 boys for every 100 girls).18 Column 4 of Table 5 shows a very high ratio of boy deaths to girl deaths; model life tables show infant and child mortality rates that are only 10–20% higher for boys than for girls, at most. Also, the proportion of dead boys among women who gave birth to only boys was much higher than the proportion of dead girls among women who gave birth to only girls (column 5, Table 5). The ratio of the proportion of dead boys to the proportion of dead girls was lower for women who had given birth to at least three children, only one of them a boy (column 7, Table 5).

Fig. 1 presents the estimated infant and under-five mortality rates obtained by assuming that deaths were underreported and applying the sensitivity analysis with varying proportions of girls among the deaths presumed to be missing. The weighted infant mortality rate estimates ranged from 129 deaths per 1000 live births (no adjustments) to 156 deaths per 1000 live births (after adjustment for all presumably unreported dead girls). The under-five mortality rate ranged from 191 deaths per 1000 live births (no adjustments) to 234 deaths per 1000 live births (after adjustment for all presumably unreported dead girls). Studies conducted in Bangladesh suggest that liveborn children not reported are more often dead than those who are reported.19,20 Becker & Mahmud found that 25% of missed liveborns were children who had died;19 Espeut found the percentage to be 85.20 If Becker & Mahmud’s percentage is applied, the infant and under-five mortality rates obtained are 132 and 196 per 1000 live births, respectively; if Espeut’s percentage is applied, the rates are 151 and 231 per 1000 live births. If the imbalance between boys and girls is assumed to be entirely attributable to unreported births in girls and 50% of the unreported liveborn girls are dead (i.e. an average of the two estimates from the validation studies), the estimated infant mortality rate then becomes 140 and the under-five mortality rate becomes 209 per 1000 live births.

Fig. 1. Sensitivity analysis: infant and under-five mortality rates estimated by applying varying proportions of dead girls among unreported girls, Afghanistan
Fig. 1. Sensitivity analysis: infant and under-five mortality rates estimated by applying varying proportions of dead girls among unreported girls, Afghanistan

Discussion

Our findings show that in rural Afghanistan infant and under-five mortality rates are lower than the estimates in the State of the world’s children and that the infant mortality rate is similar to the projected rate for 2005 under the median scenario applied to derive the United Nations Children’s Fund’s best estimates (130 per 1000 live births for 2005)21 and the United States Bureau of the Census estimates, based on census data from 1979 (137 per 1000 live births).9 These national estimates are not directly comparable to the rural estimates from the AHS; however, because approximately 80% of the population of Afghanistan lives in rural areas, national estimates are similar to rural estimates. Because the AHS provides estimated infant and under-five mortality rates from a new primary data source, it fills a gap in the understanding of current levels of mortality in Afghan children. However, the AHS and other available estimates of infant and under-five mortality rates in Afghanistan should be interpreted with caution in light of their limitations. Further research in this area is needed.

The AHS did not cover urban areas or highly unsafe regions of the country. Urban areas may have lower infant and under-five mortality than rural areas, but the urban–rural differentials observed in the NDFGS and 2003 Multiple Indicator Cluster Survey shown in Table 1 indicate that infant and under-five mortality rates at the national level may be only slightly lower than those in rural areas of Afghanistan. Highly unsafe regions of the country that could not be covered in the 2006 AHS may have higher mortality levels than areas that were covered. The use of mortality estimates based on the youngest maternal age groups may introduce bias and instability in the estimates, since a child’s risk of dying is related to the mother’s age.

The skewed sex ratio at birth observed in the AHS has several possible explanations. Preference for sons manifests itself before birth in the form of sex-selective abortions, and after birth as female infanticide or abandonment and unequal treatment or neglect of girls, which lead to higher female mortality. Sex-selective abortion as a cause of the skewed sex ratio is discarded as an explanation for Afghanistan, where few ultrasound machines exist and antenatal care is uncommon. There is little evidence (either from the literature or anecdotally) that infanticide is practiced in Afghanistan. Thus, the skewed sex ratio may stem from a combination of intentional misclassification of girls as boys and underreporting of girls, provided the number of boys reported is correct. Misclassification of girls as boys affects sex-specific mortality rates, but it does not affect overall mortality rates if the total number of children ever born is correct. Anecdotal evidence suggests that households that have more girls than boys may intentionally misclassify girls as boys, since there is some stigma in Afghanistan attached to having only girls. Underreporting of girls could indicate that dead girls are not being counted in the total births. Conservative households may intentionally underreport the number of living girls in the household to keep strangers from knowing about the presence of females. Also, many households refuse to divulge the names of females in the household, and some may refuse to report the presence of living females. Underreporting of girls would introduce error into overall estimates of mortality.

The high ratio of deaths in children of women who gave birth to only boys compared to deaths in children of women who gave birth to only girls indicates that deaths in girls are probably underreported. It suggests that underreporting of liveborn girls is likely to be a greater problem than misclassification of girls as boys. Dead girls are probably underreported and may not even be included among total births. The underreporting of deaths among girls would lead to underestimation of the true mortality rate. Table 6 summarizes possible threats to validity in this study, including their likely direction and magnitude.

While available mortality estimates for Afghanistan need to be interpreted with caution, large gains have been made in Afghanistan’s health sector in recent years. For example, studies have shown marked progress in primary health care, especially in the quality of patient care and the availability of essential drugs and family planning supplies; in antenatal care to pregnant women; in improved health worker skills; and in the number of female health workers providing care throughout the country.22 Coverage with maternal and child health services such as family planning, skilled antenatal care, skilled birth attendance and child immunization has also increased.2325

There is a clear need for robust and current measures of fertility and mortality in Afghanistan, especially in view of widespread changes in the country, including large investments in health services development, and the various obstacles to accurate measurement in the country. Given these obstacles, which make it difficult to accurately estimate mortality retrospectively, a better understanding is needed of the sociocultural determinants of mortality and how they affect the reporting of births and deaths in Afghanistan. Systems for measuring and reporting mortality need to be institutionalized in the country to generate better data for planning and policy-making. The findings from this study also have wider implications for child mortality statistics published elsewhere, such as in the State of the world’s children, which are often accepted as accurate although they can be flawed, especially in countries that lack a well established vital registration system and rigorous methods for regularly measuring mortality. Strengthening expert review of these estimates and investing in studies on mortality-related reporting practices in countries with data limitations may result in more robust estimates.


Acknowledgments

We acknowledge the many contributions to this study made by members of the Monitoring and Evaluation Technical support team from the Johns Hopkins Bloomberg School of Public Health and the Indian Institute of Health Management Research, and by colleagues from the Ministry of Public Health, Afghanistan. We would also like to thank the 250 surveyors who collected data in difficult circumstances.

At the time this study was conducted, Dr Peter Hansen was affiliated with the Department of International Health, Johns Hopkins Bloomberg School of Public Health.

Funding:

This work was supported by the Third Party Evaluation Contract (MoPH/AFG/GCMU/19/04) between the Government of Afghanistan and the Johns Hopkins Bloomberg School of Public Health with the Indian Institute of Health Management Research.

Competing interests:

None declared.

References

Share