THE NEW ENGLAND JOURNAL OF MEDICINE
Vol. 327, No. 13, Pg. 931-936, Sept. 24, 1992SPECIAL ARTICLE
-----------------EFFECT OF THE GULF WAR ON INFANT AND CHILD MORTALITY IN IRAQ
ALBERTO ASCHERIO, M.D., D.P.H., ROBERT CHASE, M.D., C.C.F.P., TIM COTE, M.D., M.P.H., GODELIEAVE DEHAES, M.D., ERIC HOSKINS, M.D., JILALI LAAOUEJ, M.D., MEGAN PASSEY, M.B., B.S., M.P.H., SALEH QADERI, M.B., B.S., SAHER SHUQAIDEF, M.B., B.S., DR.P.H., MARY C. SMITH, M.SC., AND SARAH ZAIDI, M.SC.
Abstract Methods Results Discussion References About the Authors Abstract Background. Increased malnutrition and morbidity among Iraqi children after the onset of the Persian Gulf war have been reported by several fact-finding missions. The magnitude of the effect of the war and the economic embargo on child mortality remains uncertain, however.
Methods. We conducted a survey of 271 clusters of 25 to 30 households each, chosen as a representative sample of the Iraqi population. The households were selected and the interviews conducted by an international team of public health professionals independent of Iraqi authorities. In each household all women 15 to 49 years of age were interviewed, and the dates of birth and death of all children born on or after January 1, 1985, were recorded.
Results. The study population included 16,076 children, 768 of whom died during the period surveyed (January 1, 1985, to August 31, 1991). The age-adjusted relative mortality for the period after the war began, as compared with the period before the war, was 3.2 (95 percent confidence interval, 2.8 to 3.7). No material change in the relative risk was observed after adjustment for region of residence, maternal education, and maternal age. The increase in mortality after the onset of the war was higher among children 1 to less than 12 months old (relative risk, 4.1; 95 percent confidence interval, 3.3 to 5.2) and among those 12 to less than 60 months old (relative risk, 3.8; 95 percent confidence interval, 2.6 to 5.4) than among those less than 1 month old (relative risk, 1.8; 95 percent confidence interval, 1.4 to 2.4). The association between the war and mortality was stronger in northern Iraq (relative risk, 5.3) and southern Iraq (relative risk, 3.4) than in the central areas (relative risk, 1.9) or in Baghdad (relative risk, 1.7).
Conclusions. These results provide strong evidence that the Gulf war and trade sanctions caused a threefold increase in mortality among Iraqi children under five years of age. We estimate that an excess of more than 46,900 children died between January and August 1991. (N Engl J Med 1992;327:931-6.)
WHEN Iraq invaded Kuwait in August 1990, international trade sanctions were rapidly instituted against Iraq. On January 16, 1991, the war to liberate Kuwait began, and within six weeks Allied forces declared a cease-fire. Civilian revolts against the Iraqi government in March and April resulted in the displacement of an estimated 2 million people in the northern (Kurdish) and southern (Shiite) regions of Iraq.
Little objective information is available about the effect of the economic sanctions, war, and civilian uprisings on the health of civilians. An international team that visited Iraq in April 1991 reported epidemics of cholera, typhoid, and gastroenteritis and found that mortality rates had increased two to three times among children admitted to hospitals in Baghdad and Basra.1 Other studies have reported on the health conditions in Kurdish refugee camps2,3 and the nutritional status of children in the south.4 However, all previous estimates of the effect of the war on health have been based on studies of selected population groups that may not represent the entire country. We report here the results of a survey of mortality conducted in a nationwide sample of households between August 25 and September 5, 1991. The survey was made independently of the Iraqi government by a team of international researchers.
Study Population
The survey was based on interviews with clusters of households selected according to a sampling plan that involved several stages. The framework for the sample was provided by the Iraqi census of 1987, with adjustments for the displacement of populations in northern and southern Iraq during March and April 1991. Data on these displacements were provided by the United Nations High Commissioner for Refugees. First, the population of the entire country was divided into 19 regional groups, 18 of which corresponded to the administrative divisions that constitute the Republic of Iraq and 1 of which represented the camps of refugees under United Nations protection in the north (the neutral territories). Within each administrative division, an urban and a rural area were then defined on the basis of the census. The sample was designed to include 299 clusters of 25 to 30 households each, distributed among the regional groups in proportion to their population.
For the purposes of the survey, a household was defined as a dwelling, house, apartment, or tent serving as the residence of at least one woman who had given birth to a live-born child on January 1, 1985, or thereafter. In each regional group, the method of the Expanded Programme of Immunization was used for the sampling.5 In the first stage, districts were selected for which the probability of being included in the sample was proportional to the size of the population. In the second stage, 1 household in the district was selected at random, and then 25 to 30 households in the vicinity of that household were identified to form the cluster. The team collected information on 271 such clusters. The remaining 28 clusters originally intended for the sample were not included, largely because of time constraints, and the selection of those that were omitted can reasonably be considered to have been random. The entire sample was selected, and the survey executed, independently of the Iraqi government and of international organizations present in Iraq.
Data Collection
Most of the interviewers were volunteers recruited in Jordan among students at Jordan University, Irbid University, and the Jordanian University of Science and Technology. All the interviewers had attended or were attending a university, graduate school, or medical school either in Jordan or overseas. They were fluent in Arabic and English, and some had previous experience conducting similar surveys in Jordan. Of the 31 interviewers, 21 were women. In Kurdish-speaking areas, female teachers were hired as interpreters to help conduct the survey. The interviewers were organized into seven teams, each supervised by a public health professional (among the supervisors, there were two each from Belgium, Jordan, and the United States and one each from Canada and Pakistan). One team had two supervisors. The teams were trained in Amman over a three-day period. The training focused on the accurate collection and reporting of dates of birth and death and was complemented by a one-day pilot survey in Baghdad.
In each household all women 15 to 49 years of age who reported at least one live birth on or after January 1, 1985, were interviewed. During the interview, a one-page form was used to record the dates of birth and death of all children born on or after January 1, 1985. Birth certificates or immunization cards (available for 66.5 percent of the children) were used to verify the date of birth. If this documentation was unavailable, the mother was asked the exact date of the birth. The date of death was based on the mother's report. When dates were uncertain, they were estimated. The data collected also included information on maternal education and, for each child, information on the presence of diarrhea, defined as the occurrence of three loose stools in a 24-hour period during the past 48 hours. In the case of children who died, the cause of death as reported by the mother was recorded. Refusals to participate in the survey were not recorded, but they were estimated to number under 50.
A total of 16,172 live births and 803 deaths were reported by the women interviewed. Ninety-six children (0.6 percent) were excluded from the study because the year of birth or death was missing; of these children, 35 had died (4.4 percent of all deaths). Thus, 16,076 live births and 768 deaths were included in the analyses. In the primary analysis, children for whom the month of birth (n = 905) or death (n = 160) was missing were assigned a month at random. The assignment of missing months of death took into account the child's approximate age at death as reported by the mother (data were available for 63 children) and the distribution of ages at death for all children for whom information was complete.
Statistical Analysis
Neonatal mortality (deaths between birth and the age of one month) was estimated by dividing the number of deaths among children less than one month of age by the number of live births during the same period. The analyses of postneonatal mortality (1 to <12 months) and child mortality (12 to <60 months) were based on rates calculated with person-time used as the denominator. Each child was considered to contribute person-time from birth until either death or the date of the survey.
Mortality rates were calculated for each four-month period from January 1985 through August 1991 in order to assess trends and seasonal variations. The effect on mortality of the Gulf war, including the civil disruption and the economic sanctions, was estimated by calculating relative mortality, defined as the ratio of the average mortality rates on or after January 1, 1991, to the corresponding rates before January 1,1991. Because births before 1985 were not included in the survey, child mortality before 1988 reflected mortality among children 12 to less than 24 months of age, rather than 12 to less than 60 months, and therefore it was not included in the computation of the mortality rate used as a reference for the age group of 12 to less than 60 months.
Confidence intervals were calculated with the assumption of a simple Poisson distribution and also by applying the Mantel-Haenszel procedure to the 271 clusters in the sample, with each cluster treated as a separate stratum.6 The two methods gave virtually identical results for each age group, and therefore only the results based on the simple Poisson distribution are included. The age-specific mortality rates were transformed into the probabilities of dying before the age of one year and before the age of five years with life tables generated directly from the data.7
The number of excess deaths related to the war was based on a projected population in Iraq of 18.92 million in the middle of 1990, with 18 percent of the population under the age of five years, a growth rate of 3.5 percent, and a crude birth rate of 43 per 1000.8 Excess deaths were calculated as the difference between the estimated and the expected numbers of deaths among Iraqi children under five years of age during the first eight months of 1991. For this calculation, the expected and estimated numbers of deaths in each age group were obtained by applying the age-specific mortality rates before and after the onset of the war, respectively, to the estimated population of Iraq in the same age group on January 1, 1991; estimates of the mean population counts for the ages of 1 to less than 12 months and of 12 to less than 60 months were based on the respective proportions among the children in the survey. The prewar birth rate was used to estimate the number of live births during the first eight months of 1991, because there is no evidence that fertility rates changed after the beginning of the war. Relative mortality was also calculated according to strata of maternal education, region of residence, and sex. Maternal education was classified in three levels: none (i.e., the mother had no formal schooling and was illiterate), primary (i.e., the mother attended primary school), and secondary or university (i.e., the mother attended secondary school or university). Region of residence was classified into five areas: Baghdad, central (al-Anbar, Babel, Diala, and Salah el-Din), south (al-Muthana, Basra, Karbala, Maysan, Najaf, and Qadissiyah), north (Irbil, Dohuk, Ninevah, Sulaymaniyah, and al-Tamim), and neutral territories (Penjwin and Said Sadiq). Neutral territories was considered a separate region because Kurdish persons displaced as a consequence of the Gulf war resided there under the protection of the United Nations High Commissioner for Refugees.
To adjust the relative mortality data for age, education, area of residence, and maternal age, we used Poisson regression models.9,10 In these analyses, age was categorized into five groups (<1 month, 1 to <6 months, 6 to <12 months, 12 to <24 months, and 24 to <60 months), and maternal age into six groups (from 15 to 44 years, in 5-year intervals). For each age category, all the person-time and all deaths occurring between January 1, 1985, and December 31, 1990, were used to compute base-line rates. All the covariates were represented in the regression models by indicator variables. To assess whether the effect of the war on mortality was modified by other factors, we added to the regression model indicator variables for the interactions of war with age, maternal education, and region of residence. Statistical tests were based on the likelihood ratio; two-sided P values lower than 0.05 were considered significant.
The sociodemographic characteristics of the children included in the study are shown in Table 1. There were no clear trends in infant mortality between January 1985 and December 1990 (Fig. 1 and 2). Similarly, no trends were observed in child mortality between January 1988 and December 1990 (Fig. 3). The average mortality rates in the years before the war were similar, whether they were computed from January to August or from January to December.
When the period after the onset of the war was compared with that before 1991, the relative mortality was 1.8 for neonates in the first month of life, 4.1 for infants 1 to less than 12 months of age, and 3.8 for children 12 to less than 60 months of age (Table 2). When these figures are extrapolated to the entire population of Iraq, a similar increase in mortality would have resulted in approximately 46,900 excess deaths among children under five years of age during the first eight months of 1991. The probabilities of dying in the first year of life or in the first five years were 32.5 and 43.2 per 1000 live births, respectively, before January 1, 1991. In a cohort of children who were hypothetically at risk from birth to the age of five years, given the mortality rates prevailing after the onset of the war, the probabilities of dying in the first year of life or the first five years would have been 92.7 and 128.5 per 1000, respectively.
An increased risk of death in 1991 as compared with the previous years was seen for each region and each level of maternal education, except for children under one month of age in the central region (Table 3). Regional differences in mortality that were present before the war were maintained or exacerbated by the conflict. In the neutral territories of the north (the refuge of the displaced Kurds), the mortality rate before the war for the group 1 to less than 12 months old was more than 3.5 times higher than that in Baghdad, whereas after the war mortality in the north was more than 11 times that in Baghdad (Table 3). Similarly, the infant-mortality rate among children born to illiterate mothers was 2.5 times as high before the war and 3.5 times as high after the war as the rate among children born to mothers with a secondary education (Table 3).
In the regression analyses, the age-adjusted relative mortality after the war was 3.2 (95 percent confidence interval, 2.8 to 3.7), and it changed only slightly after adjustment for region of residence and maternal education (relative risk, 3.4; 95 percent confidence interval, 2.9 to 3.9). These estimates did not change materially after adjustment for maternal age. When interaction terms were added to the regression model, the effect of war on mortality was significantly different between age groups (P<0.001) and between regions (P<0.001), but not between levels of maternal education. When the period after the war was compared with the period before the war, the estimated age-specific relative risks were 1.8 for neonates and between 4.3 and 5.2 for the other age groups. Among regions, the lowest relative risks were seen in Baghdad (1.7) and the central region (1.9), followed by the southern region (3.4), the neutral territories (5.1), and the northern region (5.3).
Information on the cause of death was available for 583 children (75.9 percent). The age-adjusted mortality rate from diarrhea rose from 2.1 per 1000 person-years before the onset of the war to 11.9 per 1000 person-years after the onset of the war. The age-adjusted mortality rate from injuries rose from 0.55 per 1000 person-years before the war to 2.25 per 1000 person-years after the onset of the war. Before the war, 20.7 percent of deaths were due to diarrhea and 8.8 percent to injuries; after the onset of the war, the comparable proportions were 38.0 percent and 7.2 percent.
In this survey conducted in a representative sample of Iraqi households, we found that infant and child mortality increased more than threefold in the period from January through August 1991, as compared with the average rates during the previous six years. This increase corresponds to an excess of about 46,900 deaths among Iraqi children under five years of age.
As in any survey, these results may be affected by several sources of error. The framework for the sample, based on the 1987 census and corrected for major population displacements, may reflect the actual distribution of the population only approximately. Increases in mortality were observed in all regions, however, and thus major error from this source seems unlikely. Two sources of potential bias are the dates of death missing for a substantial proportion of the children in our sample and the possible omission of deaths occurring before 1990. Families that might have moved out of Iraq because of the conflict would not have been included in the survey. Moreover, the exclusion from the sample of children born to women who died during the war may cause an underestimation of the actual effect of the war on child survival. However, most of the deaths for which we have incomplete dates occurred before 1991, and although any exclusion of these children from the analyses would produce an underestimation of mortality before 1991 (and therefore an overestimation of the increase in mortality after the onset of the war), the random assignment of the month of death does not affect the allocation of deaths to the period before or after January 1, 1991, and therefore it is unlikely to affect the estimates of relative risk. In addition, our estimates of mortality rates for infants and children before 1991 are similar to those in surveys conducted before the beginning of the conflict.
As compared with our estimates of 32.5 deaths per 1000 live births, the estimates of infant mortality in Iraq for 1987 by the United Nations Children's Fund (UNICEF) were 25 (obtained directly by the technique of birth historyll) and 41 (obtained indirectly by the Brass method12) deaths per 1000 live births.13 The Gulf Child Health Survey calculated infant-mortality rates for 1988 of 29 and 36 deaths per 1000 live births by the direct and indirect methods, respectively.14 Comparable figures for the mortality rate for children under five years of age are 43.2 per 1000 live births (in this study), 52 per 1000 live births (UNICEF estimates) and 46 per 1000 live births (Gulf Child Health Survey). An apparent increase in recent mortality could have been caused by a systematic bias toward reports of recent deaths, in which children who died many years before were omitted from the report to the interviewers. The lack of a secular trend in the prewar mortality rates is evidence against this. Also, the interviewers and respondents may have worked more diligently to report and document 1991 deaths than they did for earlier deaths. This explanation, however, fails to account plausibly for the magnitude of the differences in mortality before and after the war.
Our data demonstrate the link between the events that occurred in 1991 (war, civilian uprising, and economic embargo) and the subsequent increase in mortality. The destruction of the supply of electric power at the beginning of the war, with the subsequent disruption of the electricity-dependent water and sewage systems, was probably responsible for the reported epidemics of gastrointestinal and other infections.1 These epidemics were worsened by the reduced accessibility of health services and decreased ability to treat severely ill children.1 Increased malnutrition, partly related to the rising prices of food,15 may also have contributed to the increased risk of death among infants and children. The effect of the war has been greater among groups that had higher base-line mortality rates, suggesting that poverty and lower educational level increased children's vulnerability to the crisis. In northern and southern Iraq, the situation was exacerbated by the civilian uprisings and the subsequent flight of 2 million Kurds and Shiites into mountains and marshes at a climatically inhospitable time.
The hypothesis that the excess mortality caused by the war was due to infectious diseases and to the decreased quality and availability of medical care, food, and water is consistent both with the increase in the proportional mortality from diarrhea and with the shift in the age pattern of mortality, characterized by a lower proportional contribution of neonatal deaths to mortality among persons under the age of five years after the onset of the war. This pattern resembles that observed in the less-developed countries, where diarrhea and respiratory infections account for most deaths in infancy and childhood.16
War is never good for health,17,18 but the full effect of war and economic sanctions on morbidity and mortality is difficult to assess, and the number of civilian casualties caused indirectly is likely to be underestimated. The vulnerability of children to war waged at low intensity in Angola and Mozambique has recently been reported.19 During the Gulf war, it was suggested that by using high-precision weapons with strategic targets, the Allied forces were producing only limited damage to the civilian population. The results of our study contradict this claim and confirm that the casualties of war extend far beyond those caused directly by warfare.
We are indebted to the people of Iraq who participated in the study and to Mouna Abdel Hamid, Ghassan Abusitteh, Nisreen Alami, Kholoud Athamneh, Eham Athamneh, Ohoud Bata, Hussam Bushnaq, Waleed Gharaibeh, Samar Hassan, Hussein Jafar, Saied Jaradat, Natalie Kakish, Ghada Konash, Lamees Marji, Ismail Matalka, Tahani Momani, Lulwa Mutawi, Taline Najjar, Faruq al-Omari, Rania Orabi, Vassar Qatarneh, Rita Qumsieh, Dana Sajdi, Rasha Sayegh, Shereen Shaheen, Faten ai-Taber, Raja Tubaishat, Shurooq Tubaishat, Nadine Toukan, Rana Tumaira, Mahmoud Turk, and Mua'tasem Ubeidat for performing the interviews; to Chris Cook for creating the data base; to Philippe Jacob for computer assistance; to Elizabeth Benjamin, Kurtis Doebbler, Steven Donzinger, Jean Dreze, Roger Normand, and Sarah Leah Whitson for making this mission possible; and to Drs. Allan Hill, Walter Willett, Lincoln Chen, Michel Garenne, Chung-cheng Hsieh, and Alexander Walker for their advice and encouragement.REFERENCES
1. The Harvard Study Team. The effect of the Gulf crisis on the children of Iraq. N Engl J Med 1991;325:977-80.
2. Public health consequences of acute displacement of Iraqi citizens -- March-May 1991. MMWR Morb Mortal Wkly Rep 1991;40:443-6.
3. Medecins Sans Frontieres. Kurdistan: helping an abandoned people. Alert. Fall-Winter. 1991-1992:1-4.
4. Sato N, Obeid O, Brun T. Malnutrition in southern Iraq. Lancet 1991;338:1202.
5. Henderson RH, Sundaresan T. Cluster sampling to assess immunization coverage: a review of experience with a simplified sampling method. Bull World Health Organ 1982;60:253-60.
6. Rothman KJ. Modem epidemiology. Boston: Little. Brown. 1986:171.
7. Newell C. Methods and models in demography. New York: Guildford Press. 1988:71-81.
8. Demographic yearbook 1990. New York: United Nations, 1992.
9. Breslow NE, Day NE. Statistical methods in cancer research. Vol. 2. The design and analysis of cohort studies. Lyon, France: International Agency for Cancer Research, 1987:160-76. (IARC scientific publications no. 32.)
10. Baker RJ, Nelder JA. The GLIM system: release 3. Oxford, England: Numerical Algorithm Group, 1978.
11. David PH, Beisharat L, Hill AG, Bennet S. Measuring childhood mortality: a guide for simple surveys. Amman, Jordan: UNICEF, 1990:1-26.
12. Manual X: indirect techniques for demographic estimation. New York: United Nations, 1983.
13. Iraq: immunization, diarrhoeal diseases, maternal and childhood mortality survey. Evaluation Series no. 9. Amman, Jordan: UNICEF Regional Office for the Middle East and North Africa, 1990.
14. National child health survey. Baghdad, Iraq: Ministry of Health, 1990.
15. Dreze J, Ghazdar H. Hunger and povery in Iraq. Vol. 32 of Wider Series. London: London School of Economics, 1991:1-63.
16. Adamchak DJ, Flint WC. A note on percent neonatal-postneonatal mortality vs infant mortality rates as correlates of socio-economic development. Pop Rev 1980;24:37-40.
17. Stein Z, Susser M, Saenger G, Marolla M. Famine and human development: the Dutch hunger winter of 1944-45. New York: Oxford University Press, 1975.
18. Garfield RM, Neugut AI. Epidemiologic analysis of warfare: a historical review. JAMA 1991;266:688-92.
19. UNICEF. Children in the frontline. New York: United Nations, 1988.
About the AuthorsFrom the Departments of Epidemiology (A.A., M.C.S.) and Population Sciences and International Health (S.Z.), Harvard School of Public Health, Boston; the National Institutes of Health, Bethesda, Md. (T.C.); Johns Hopkins University, Baltimore (S.S.); the Department of Clinical Epidemiology and Biostatistics, McMaster University, Toronto (R.C.); Medical Aid for the Third World, Brussels, Belgium (G.D., J.L.); Oxford University, Oxford, United Kingdom (E.H.); Papua New Guinea Institute of Medical Research, Goroka, Papua New Guinea (M.P.); and the Jordan University of Science and Technology, Irbid, Amman, Jordan (S.Q.). Address reprint requests to Ms. Zaidi at the Center for Population and Development Studies, Harvard School of Public Health, 9 Bow St., Cambridge, MA 02138.
Supported in part by the United Nations Children's Fund (UNICEF).
Copyright © 1992 Massachusetts Medical Society
*** FAIR USE NOTICE: In accordance with Title 17 U.S.C. Section 107, this material is distributed without profit to those who have expressed a prior interest in receiving the included information for research and educational purposes. For more information, see http://www.law.cornell.edu/uscode/17/107.shtml. ***