Juan Cole / Juan Cole.com & Plos Medicine – 2013-10-18 02:18:10
The American Genocide Against Iraq:
4% of Population Dead as Result of US Sanctions, Wars
Juan Cole / Juan Cole.com
(October 17, 2013) — A new household survey of Iraqis has projected the civilian death toll from the Bush administrationâ€™s invasion and occupation of Iraq at roughly 450,000. Passive information-gathering techniques like logging deaths in the Western press have produced estimates closer to 150,000, but such techniques have been proven to miss a lot of people.
(To my knowledge no one was counting all the deaths reported in the some 200 Arabic-language Iraqi newspapers in the 2000s, so even the passive information-gathering was limited. And, the Wikileaks US military log of civilian deaths did not overlap very much with e.g. Iraq Body Count, so both of them were missing things the other caught.)
Of those extra deaths beyond those who would have died if the US had never invaded, some 270,000 died violently, with US troops responsible for about 90,000 civilian deaths and militias for another 90,000. Of those killed violently, 60 percent were shot, and 12 percent died from car bombs. Some 180,000 died because of the destruction of the public health infrastructure (lack of access to hospital treatment, e.g.).
Despite the horrific total, this estimate for 2003-2011 is smaller than The Lancet study of some years ago, which was done under wartime conditions. The authors admit, however, that the death toll could have been even higher; this total is a projection based on 2000 interviews.
The US/ UN sanctions on Iraq of the 1990s, which interdicted chlorine for much of that decade and so made water purification impossible, are estimated to have killed another 500,000 Iraqis, mainly children. (Infants and toddlers die easily from diarrhea caused by gastroenteritis, which causes fatal dehydration).
So the US polished off about a million Iraqis from 1991 through 2011, large numbers of them children. The Iraqi population in that period was roughly 25 million, so the US killed or created the conditions for the killing of 4% of the Iraqi population.
If Iraq had killed 4% of Americans, it would be 12 million people dead.
Iraq did not attack the United States. It did attack Iran in 1980, but by 1983 the US was an ally in Iraqâ€™s war against Iran. It also attacked Kuwait, which it occupied quite bestially, but it was out by spring 1991. There was no casus belli or legitimate legal cause of war in 2003. Iraqâ€™s main crime appears to have been to be an oil state not compliant with US demands.
All this is horrible enough. Even more horrible is that the US occupation of Iraq sparked a Sunni Arab insurgency, which is still vigorous. Insurgencies typically take 10 to 15 years to subside. Some 5000 Iraqi civilians have been killed so far this year by that insurgency. US occupation is the gift that goes on giving.
Despite the Bush administrationâ€™s violation of the UN charter and its war crimes in Iraq, none of its high officials has faced prosecution. Some of them even have the gall to come on television from time to time to urge more killing.
Mortality in Iraq Associated with
The 2003â€“2011 War and Occupation:
Findings from a National Cluster Sample Survey
By the University Collaborative Iraq Mortality Study
Previous estimates of mortality in Iraq attributable to the 2003 invasion have been heterogeneous and controversial, and none were produced after 2006. The purpose of this research was to estimate direct and indirect deaths attributable to the war in Iraq between 2003 and 2011.
Methods and Findings
We conducted a survey of 2,000 randomly selected households throughout Iraq, using a two-stage cluster sampling method to ensure the sample of households was nationally representative. We asked every household head about births and deaths since 2001, and all household adults about mortality among their siblings. We used secondary data sources to correct for out-migration.
From March 1, 2003, to June 30, 2011, the crude death rate in Iraq was 4.55 per 1,000 person-years (95% uncertainty interval 3.74â€“5.27), more than 0.5 times higher than the death rate during the 26-mo period preceding the war, resulting in approximately 405,000 (95% uncertainty interval 48,000â€“751,000) excess deaths attributable to the conflict.
Among adults, the risk of death rose 0.7 times higher for women and 2.9 times higher for men between the pre-war period (January 1, 2001, to February 28, 2003) and the peak of the war (2005â€“2006). We estimate that more than 60% of excess deaths were directly attributable to violence, with the rest associated with the collapse of infrastructure and other indirect, but war-related, causes.
We used secondary sources to estimate rates of death among emigrants. Those estimates suggest we missed at least 55,000 deaths that would have been reported by households had the households remained behind in Iraq, but which instead had migrated away.
Only 24 households refused to participate in the study. An additional five households were not interviewed because of hostile or threatening behavior, for a 98.55% response rate. The reliance on outdated census data and the long recall period required of participants are limitations of our study.
Beyond expected rates, most mortality increases in Iraq can be attributed to direct violence, but about a third are attributable to indirect causes (such as from failures of health, sanitation, transportation, communication, and other systems). Approximately a half million deaths in Iraq could be attributable to the war.
War is a major public health problem. Its health effects include violent deaths among soldiers and civilians as well as indirect increases in mortality and morbidity caused by conflict. Unlike those of other causes of death and disability, however, the consequences of war on population health are rarely studied scientifically.
In conflict situations, deaths and diseases are not reliably measured and recorded, and estimating the proportion caused, directly or indirectly, by a war or conflict is challenging. Population-based mortality survey methods — asking representative survivors about deaths they know about — were developed by public health researchers to estimate death rates. By comparing death rate estimates for periods before and during a conflict, researchers can derive the number of excess deaths that are attributable to the conflict.
Why Was This Study Done?
A number of earlier studies have estimated the death toll in Iraq since the beginning of the war in March 2003. The previous studies covered different periods from 2003 to 2006 and derived different rates of overall deaths and excess deaths attributable to the war and conflict. All of them have been controversial, and their methodologies have been criticized.
For this study, based on a population-based mortality survey, the researchers modified and improved their methodology in response to critiques of earlier surveys. The study covers the period from the beginning of the war in March 2003 until June 2011, including a period of high violence from 2006 to 2008. It provides population-based estimates for excess deaths in the years after 2006 and covers most of the period of the war and subsequent occupation.
What Did the Researchers Do and Find?
Interviewers trained by the researchers conducted the survey between May 2011 and July 2011 and collected data from 2,000 randomly selected households in 100 geographical clusters, distributed across Iraq’s 18 governorates.
The interviewers asked the head of each household about deaths among household members from 2001 to the time of the interview, including a pre-war period from January 2001 to March 2003 and the period of the war and occupation.
They also asked all adults in the household about deaths among their siblings during the same period. From the first set of data, the researchers calculated the crude death rates (i.e., the number of deaths during a year per 1,000 individuals) before and during the war. They found the wartime crude death rate in Iraq to be 4.55 per 1,000, more than 50% higher than the death rate of 2.89 during the two-year period preceding the war.
By multiplying those rates by the annual Iraq population, the authors estimate the total excess Iraqi deaths attributable to the war through mid-2011 to be about 405,000. The researchers also estimated that an additional 56,000 deaths were not counted due to migration. Including this number, their final estimate is that approximately half a million people died in Iraq as a result of the war and subsequent occupation from March 2003 to June 2011.
The risk of death at the peak of the conflict in 2006 almost tripled for men and rose by 70% for women. Respondents attributed 20% of household deaths to war-related violence. Violent deaths were attributed primarily to coalition forces (35%) and militia (32%). The majority (63%) of violent deaths were from gunshots. Twelve percent were attributed to car bombs.
Based on the responses from adults in the surveyed households who reported on the alive-or-dead status of their siblings, the researchers estimated the total number of deaths among adults aged 15â€“60 years, from March 2003 to June 2011, to be approximately 376,000; 184,000 of these deaths were attributed to the conflict, and of those, the authors estimate that 132,000 were caused directly by war-related violence.
What Do These Findings Mean?
These findings provide the most up-to-date estimates of the death toll of the Iraq war and subsequent conflict. However, given the difficult circumstances, the estimates are associated with substantial uncertainties.
The researchers extrapolated from a small representative sample of households to estimate Iraq’s national death toll. In addition, respondents were asked to recall events that occurred up to ten years prior, which can lead to inaccuracies.
The researchers also had to rely on outdated census data (the last complete population census in Iraq dates back to 1987) for their overall population figures. Thus, to accompany their estimate of 460,000 excess deaths from March 2003 to mid-2011, the authors used statistical methods to determine the likely range of the true estimate. Based on the statistical methods, the researchers are 95% confident that the true number of excess deaths lies between 48,000 and 751,000 — a large range.
More than two years past the end of the period covered in this study, the conflict in Iraq is far from over and continues to cost lives at alarming rates. As discussed in an accompanying Perspective by Salman Rawaf, violence and lawlessness continue to the present day. In addition, post-war Iraq has limited capacity to re-establish and maintain its battered public health and safety infrastructure.
Citation: Hagopian A, Flaxman AD, Takaro TK, Esa Al Shatari SA, Rajaratnam J, et al. (2013) Mortality in Iraq Associated with the 2003â€“2011 War and Occupation: Findings from a National Cluster Sample Survey by the University Collaborative Iraq Mortality Study. PLoS Med 10(10): e1001533. doi:10.1371/journal.pmed.1001533
Academic Editor: Edward J. Mills, University of Ottawa, Canada
Accepted: September 3, 2013;
Published: October 15, 2013
Copyright: 2013 Hagopian et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: Support for this study came from pooled internal resources by the American and Canadian researchers without external funding. No funding bodies had any role in study design, data collection and analysis, decision to publish, or preparation of the manuscriptâ€¦.
Abbreviations: COSIT, Iraqi Central Statistical Organization; ICSS, improved corrected sibling survival; IFHS, Iraq Family Health Survey; IMIRA, Iraq Multiple Indicator Rapid Assessment; PY, person-years; UI, uncertainty interval
Estimates of the number of Iraqi deaths after the US-led invasion in 2003 have varied considerably and are contested â€“. Measuring deaths during war is complex, and methods vary, yet assessing the public health consequences of armed conflict is important .
There are several approaches to measuring mortality during a period of conflict, including registration of vital events, passive surveillance, and population-based surveys . Iraq’s last complete census was in 1987, with a partial census in 1997, after which Iraq experienced a period of extensive demographic change, including internal and external migration. With regard to vital events, death certificates continued to be issued during the conflict, though aggregation and tabulation were affected .
To date, five population-based surveys have attempted to estimate war-related deaths in Iraq. None of these were conducted after 2006, the peak of the conflict and subsequent migration . Two of these studies reported only the violent death rate ,, and three estimated both violent-only and all-cause death rates â€“.
These studies reported widely varying rates of mortality. All attracted various criticisms, including potential bias in sample selection, wide ranges of uncertainty intervals (UIs) related to relatively small sample sizes, and disputes related to statistical methods, the choice of reference populations for calculating rates, and the plausibility of results .
There has been substantial demographic change in Iraq as a result of both internal and cross-border migration throughout the course of the long conflict.
Our study builds on lessons from previous mortality studies in conflict settings and, to our knowledge, provides the first estimates for mortality in Iraq during the years 2006â€“2011. We used both the standard household demographic method (reported household deaths) and the improved corrected sibling survival (ICSS) method , the latter to increase sample size, correct for survival bias â€“, and reduce migration bias.
We analyzed data from both methods to produce nationally representative estimates of conflict-related mortality for both the general population (household survey) and for Iraqi adults, defined  as those aged 15â€“60 y (sibling survey). Additional analyses of secondary data were performed to adjust these estimates to account for migration.
In mid-2011 we conducted a nationally representative cross-sectional survey of all adults living in 2,000 randomly selected households in 100 clusters across Iraq. In retrospective cluster sample mortality surveys, the idea is to use random selection to generate a predetermined number of â€œclusters,â€ or geographically proximate household groups, across the area in question.
These representative households are then queried about their composition and mortality events over a given time period, to allow researchers to generate crude death rates; these rates are then multiplied by the country’s population total to calculate a death estimate . We used a questionnaire that asked all adults in the household about the births and deaths of their siblings, as well as all births and deaths in the household since 2001. The questionnaire is provided as Questionnaire S1.
Setting and Sample Selection
We employed a two-stage cluster sampling method. We used a commercial software product (LandScan) that contained gridded population data at the 1-km2 level in a geographic information system, and we linked it to Google Earth imagery. In the first stage of cluster selection, we randomly selected 100 1-km2 areas using a probability-proportional-to-size approach.
After those areas were selected, we superimposed a smaller grid (10 mÃ—10 m) onto each of the selected areas, and randomly selected one grid cell in each of the 100 clusters. In each small grid cell, we examined the Google Earth image and selected the residential rooftop that most fully fit in the square to serve as the start household . Details are in Figure S1, Text S2, and Questionnaire S1. Our field manual (see Manual S1) established protocols for selecting 19 dwellings adjacent to the starting household.
Our sample size was established building on experience derived from previous studies. By doubling the number of clusters used in two previous mortality studies ,, we were able to reduce the possibility of missing pockets of unusually high or low conflict-related mortality, and by halving the number of households per cluster we were able to keep the operational complexity of conducting the survey manageable, and still visit a reasonable number of households per cluster.
Processes and Timeline
We recruited study collaborators and drafted questionnaires in early 2011. Lead researchers from three North American universities and two Iraqi team leaders met in northern Iraq in March 2011 to revise data collection instruments and survey processes, finalize the field manual, and gain experience finding start households using Google maps.
The two Iraqi team leaders recruited eight medical doctors with experience in community surveys as interviewers. Author W. M. W. conducted training for data collectors in Iraq in March 2011. Weekly (or more frequent) teleconferences were held between the North American team and the lead Iraqi investigator during the design and implementation phases. The entire team met again in Iraq in September 2011 to review and interpret preliminary findings.
Data Collection and Entry
Four two-person teams along with their supervisors (for a total of ten surveyors) surveyed 100 clusters of 20 households between May 13 and July 2, 2011. The supervisor returned to one randomly selected household in each cluster (where he or she had not previously visited) to repeat the survey as a quality check. When in governorates outside Baghdad, persons familiar with locations and local security issues joined the teams to help obtain local approvals and find designated clusters.
An Iraqi events calendar and an age/birth-year chart were created to assist with recalling dates of birth or death. Interviewers asked for causes of death, and coded these from a brief listing of common causes. For war-related deaths, we asked for specific causes (such as gunshots or explosions) and perceived responsible parties (such as coalition forces or criminals).
We trained interviewers to probe for sensitive information about missing or disappeared persons, and about events distant in time among siblings and household members. We compiled qualitative observations about the remoteness and other characteristics of each cluster.
We needed to replace only one cluster (in Kerbela; governorate names per the Iraqi Central Statistical Organization [COSIT]; http://www.cosit.gov.iq/AAS/AAS2012/section_10/1.htm) for security reasons. We were obliged, however, to drop two remote clusters where our teams were strongly advised by community leaders not to visit dwellings (for cultural reasons); instead, household members were invited to a central location for interviews. As this violated the study protocol, these households were dropped from the analysis.
Each paper questionnaire contained a household and a sibling component. After obtaining verbal consent following our human subjects protocol, we interviewed the head of household (or the most senior member present) to complete a household listing. A household was defined as a group of people, not necessarily related, who regularly eat and sleep together in a building with a separate entrance and who share a kitchen.
For the household component, household births and deaths between January 1, 2001, and the interview date were recorded. When deaths were reported, interviewers requested to see death certificates. We recorded whether interviewers were shown the certificate, whether the certificate was reported to be present but not seen, or whether it was absent.
The second component of the questionnaire was a sibling history module (commonly used in Demographic and Health Surveys in developing countries) . By asking all adults in the household to recall and report on each of their siblings (defined as persons born to the same biological mother), we were able to estimate probabilities of death for adults across several decades.
Respondents to the sibling history module included all household members aged 18 y and older, and any married people under 18 y. Where necessary, telephone interviews of absent adult household members were conducted while the interviewers were in the household. We allowed limited proxy reporting for siblings. The mother of adult siblings in the home was allowed to report about her own children.
If an adult in the home was incapable of responding (because of absence, disability, or refusal), his or her relatives reported on that person’s siblings, but only if they said they were fully knowledgeable. Otherwise, the person’s response was marked as missing. If two or more siblings lived together in the same home, we interviewed only one (whichever one was actually present, or, if all were present, the sibling with the nearest next birthday to the date of our visit).
Data were recorded on paper forms, and then entered using EpiData soon after collection (see full dataset at http://ghdx.healthmetricsandevaluation.org/record/mortality-iraq-associated-2003-2011-invasion-and-occupation). Data were immediately uploaded to a website â€œdashboardâ€ to allow all investigators to monitor data collection. We employed algorithms to scan for systematic interviewer error , and observed none. All data records were rechecked against the paper record to identify and correct discrepancies.
We estimated crude death rates for the time periods January 1, 2001â€“February 28, 2003 and March 1, 2003â€“June 30, 2011 by counting deaths occurring in all households in each time period and dividing by the person-years lived within the time period.
We collected month and year of birth and death information and month and year of household formation. When the value was missing for month of death (7%, n = 26), we used June (except for 2011, where we used March for the one case with the month missing).
We calculated UIs at the 95% level for crude death rates for each time period using a bootstrapping method. Uncertainty intervals can be interpreted similarly to confidence intervals. To account for clustering, we first sampled (with replacement) the 98 existing clusters 1,000 times, so that each time we selected 98 clusters — with some of the original clusters sampled more than once, and some not sampled at all.
Next, for each of the 1,000 sets of clusters, we resampled the original number of households (with replacement) within each of the 98 sampled clusters. For each of these 1,000 replicates, we calculated annual crude death rates. The 2.5th and 97.5th percentiles of these 1,000 values served as our lower and upper bounds, respectively ,.
To estimate excess deaths caused by conflict, we calculated the war-related death rate to be the difference between the crude death rate for each time period and the crude rate during the baseline time period (January 1, 2001, to February 28, 2003). To create a war-related death count for the total population, we used the yearly United Nations Population Division estimates  for Iraq multiplied by the war-related crude death rate.
To estimate upper and lower uncertainty bounds, we used the bootstrapping method described above. Because the bootstrap process randomly chooses 1,000 possible scenarios, and we did not limit the assumptions otherwise, the occasional random selection could (and did) show a protective effect of conflict (which served to lower our final death rates).
To assess the effects of clustering on our data (â€œdesign effectsâ€), we compared our two-stage bootstrap estimates of crude mortality to a naÃ¯ve bootstrap: the ratio of the confidence interval of the larger to the smaller constitutes an estimate of the square root of the design effect. These effects of cluster sampling were not particularly large, ranging from 1.19 to 1.54 for each sex by year .
Data about adult mortality using the sibling report method are subject to predictable biases. Sibships that experience a higher mortality risk are underrepresented at the time of the survey, because these siblings are less likely to survive to be able to report (survival bias).
Additionally, larger sibships are overrepresented in the sample, because there are more siblings in the sampling frame. We used the ICSS method to adjust for these biases . Further details are in Text S1.
We calculated mortality rates for 5-y age groups between the ages of 15 and 59 y for the time periods January 1, 2001â€“February 28, 2003; March 1, 2003â€“December 31, 2004; the full years 2005â€“2006, 2007â€“2008, 2009â€“2010; and January 1â€“June 30, 2011. Our summary metric of adult mortality is 45q15, which is the risk that an individual will die before his or her 60th birthday given that he or she has lived to age 15 y.
For example, male 45q15 ranges from below 0.05 in a few countries to above 0.45 in a handful of high-mortality African nations . Uncertainty intervals were calculated using the same bootstrapping method as in the household analysis. Bootstrapping is appropriate for complex methods such as ICSS, where there is no alternative to calculating UIs. We used Stata/IC 12.0 and Python 2.6 for all analyses.
Unlike the adult sibling survival method, there is no accepted method for adjusting household figures to account for households entirely destroyed subsequent to the death of all members, or lost to migration out of the country, especially for households that experienced a death.
The Iraq Family Health Survey (IFHS) study acknowledges this shortcoming in its work as well . There is evidence that the killings in Iraq were disproportionately targeted towards the higher-income intelligentsia, a group typically in a better position to migrate to a safer setting if under attack .
We therefore reviewed a number of secondary data sources to estimate the number of Iraqis who migrated out of the country over the course of the war, to arrive at a total estimate of the missing households that left the country (and were therefore no longer available in our sampling frame). We then divided this total by an estimated household size, and multiplied total households by the average fraction of deaths per household  to estimate the total deaths our household survey would have missed, and added this number to our total death count.
We had review board approval from each participating institution in the study. Methods were reviewed to ensure they complied with the ethical guidelines for epidemiological research set out by the Council for International Organizations of Medical Sciences and other guidance, including the professional responsibility code of the American Association for Public Opinion Research â€“.
An ethicist experienced in international research associated with the Institute of Translational Health Sciences at the University of Washington, Benjamin Wilfond, further reviewed the protocols to ensure the safety of participants and interviewers was adequately protected.
We collected data from 2,000 households in 100 clusters, distributed across Iraq’s 18 governorates. After removing the two clusters previously mentioned, the total household count was 1,960, with an average of 5.34 members per household.
The study population was distributed similarly to Iraq’s estimated total 2009 population as reported by COSIT, which based its estimate on projections from the 1997 census for the 15 southern governorates and on the 1987 census for the three Kurdish governorates.
We compared the proportion of our sample to the proportion of the total population in each governorate as reported by COSIT, and derived an index of dissimilarity of 14% . The percent of recorded deaths with missing â€œcause of deathâ€ data is small. See Table 1.
Interviewers reported that 24 households refused to participate in the study, and five households were not interviewed because of hostile or threatening behavior (resulting in a 98.55% response rate). This low refusal rate is not uncommon for surveys in similar countries .
An additional 188 buildings were occupied by a business or other establishment, rather than a household, and four previously selected start dwellings were found to have been destroyed. In all these cases, replacement households were chosen using our established study protocol, to ensure total households numbered 20 per cluster.
Household Survey Results
The majority of the heads of the 1,960 households were male (85%). At the beginning of the first time period (January 2001), 1,313 of these households were already established, and contained approximately 6,455 members.
A total of 2,735 births and 383 deaths were reported during the study period. Sex and cause of death were reported for 98.4% of deaths (n = 377). Of 10,467 household members at the survey date, 50% were male, and 42% were children under the age of 18 y.
On average, households had existed as a unit for a mean of 19.9 y (and a median of 17 y) at the time of interview. The crude birth rate was 35.5 per 1,000 persons in 2001, and 32.7 in 2010. Estimated wartime crude death rates ranged from 2.0 per 1,000 person-years (PY) for females in 2011 to 7.9 for males in 2005â€“2006; pre-war crude death rates (2001â€“February 2003) were 2.1 per 1,000 PY for females and 3.7 for males.
For the full report, click here.
Please access these websites via the online version of this summary at http://dx.doi.org/10.1371/journal.pmed.1001533
This study is further discussed in a PLOS Medicine Perspective by Salman Rawaf.
The Geneva Declaration on Armed Violence and Development website provides information on the global burden of armed violence.
The International Committee of the Red Cross provides information about war and international humanitarian law (in several languages).
Medact, a global health charity, has information on health and conflict.
Columbia University has a program on forced migration and health.
Johns Hopkins University runs the Center for Refugee and Disaster Response.
University of Washington’s Health Alliance International website also has information about war and conflict.
Posted in accordance with Title 17, Section 107, US Code, for noncommercial, educational purposes.