Bulletin of the World Health Organization

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

Kavitha Viswanathan, Stan Becker, Peter M Hansen, Dhirendra Kumar, Binay Kumar, Haseebullah Niayesh, David H Peters & Gilbert Burnham

Volume 88, Number 8, August 2010, 576-583

Table 6. Factors potentially affecting the validity of infant and under-five mortality estimates for Afghanistan

Factor	Likely direction of effect	Likely magnitude of effect
Exclusion of urban areas	Overestimate	The IMR and U5MR estimates for urban areas from the NDFGS were calculated here using raw unadjusted data on total number of women in the different age groups, children born and children surviving, obtained from Trussell & Brown.⁸ The urban IMR (using the same model as the rural estimate in Table 1) was 178 and the urban U5MR was 299 per 1 000 live births for women aged 20–24 years. The urban estimates for IMR and U5MR were both 30% lower than the rural estimates for IMR and U5MR in the same study. Approximately 20% of the Afghanistan population lives in urban areas. If the level of difference found between urban and rural estimates from NDFGS is applied to AHS estimates, the IMR estimate in the AHS would drop from 129 to 121 per 1 000 live births and the U5MR would drop from 191 to 180 per 1 000 live births (a 6% decrease).
Exclusion of highly unsafe areas	Underestimate	Approximately 28% of the rural population in Afghanistan was excluded from the survey due to high levels of insecurity. These insecure areas are likely to have higher mortality than other areas, due to higher incidence of violent conflict, disruption of health services and possible disruptions to food production systems and markets. In the absence of data on levels of mortality in these areas, there is little basis upon which to estimate the magnitude of the effect. If the excluded insecure areas had 30% higher mortality rates than areas included in the survey, the IMR estimate from AHS would increase from 129 to 141 per 1 000 live births and the U5MR estimate would increase from 191 to 209 per 1 000 live births (a 9% increase).
Use of youngest maternal age group	Overestimate	Infants of mothers in the youngest age group tend to have higher mortality than infants of mothers in other age groups. Thus, mortality may be overestimated in the current study, but the magnitude is likely to be small because weights were applied reflecting the relative proportion of married women in each age group (and most were in the 20–24 age group) and the estimated mortality levels for infants from the two age groups are similar.
Misclassification of girls as boys	No effect on overall mortality	The misclassification of girls as boys would have no effect on overall mortality estimates, assuming the total number of children ever born and total number of dead children is correct. It would, however, affect sex-specific mortality estimates, depending on whether the misclassified girls are living or dead. The misclassification of dead girls as boys would contribute to an overestimate of mortality in boys and an underestimate of mortality in girls. The misclassification of living girls as boys would have the opposite effect. The magnitude of the effect depends on the proportion of dead and living girls misclassified as boys, if any.
Underreporting of dead girls	Underestimate	Fig. 1 presents a sensitivity analysis showing the effect of underreporting of dead girls on the mortality estimates. Assuming that the proportion dead among the underreported girls is greater than the proportion dead among the reported girls, this would contribute to an underestimate of mortality. Mortality estimates range from 125 IMR and 184 U5MR with 0% of underreported girls being dead, to 156 IMR and 234 U5MR with 100% of underreported girls being dead. Both scenarios are hypothetical and extremely unlikely. There is little evidence upon which to assume the likely proportion dead among the underreported girls. If half of the underreported girls are dead, the IMR estimate would be 140 and the U5MR estimate would be 209 (a 9% increase).

AHS, Afghanistan Health Survey; IMR, infant mortality rate; NDFGS, Afghanistan Demographic and Family Guidance Survey; U5MR, under-five mortality rate.