Mortality in Iraq associated with the 2003-2011 war and occupation: findings from a national cluster sample survey by the university collaborative Iraq Mortality Study

Abstract

Background: 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 3.74-5.27), 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 48,000-751,000) 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 (2005-2006). 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 The reliance on outdated census data and the long recall period required of participants are limitations of our study.

Conclusions: 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. Please see later in the article for the Editors' Summary.

Figure 1. Raw number of household deaths by year and cause, 2001–2011, reported by the University Collaborative Iraq Mortality Study.

Counts of deaths reported by respondents to the household mortality questionnaire, by year and cause. The survey concluded July 2, 2011, so the final bar reflects data for only half of the year.

Figure 2. Estimates of numbers of deaths per week in Iraq for 2-y intervals, 2001–2011, by cause as reported by households in the University Collaborative Iraq Mortality Study.

National estimate of deaths in Iraq between 2001 and 2011. Crude death rates were estimated separately within 2-y blocks (the first two time intervals are not strictly 2 y long, in order to align the first interval dividing point with the start of the war in March 2003; the survey concluded July 2, 2011, so the final bar reflects only half of the year). The counterfactual (had there been no war) estimate shows the predicted death counts if crude death rates had remained at their average level from 2001–2002 during the war and occupation (in gray). War-related, but not violent, deaths above the normal baseline are in the salmon-colored area. War-related violent deaths are portrayed in red.

Figure 3. Estimated crude death rates (and 95% uncertainty intervals) by sex and 2-y intervals from household reports in the University Collaborative Iraq Mortality Study, with plotted estimates from other studies.

Estimates of the differences in male and female death rates in households over the period 2001–2011, based on interviewing 1,960 household heads across Iraq. Mortality estimates from other studies are plotted as point estimates for comparison purposes –. The first two time intervals are not strictly 2 y long, in order to align the first interval dividing point with the start of the war in March 2003; the survey concluded July 2, 2011, so the final bar reflects only half of the year. PY, person-years.

Figure 4. Raw number of adult deaths by year and cause, 2001–2011, reported by the University Collaborative Iraq Mortality Study.

Counts of deaths reported by respondents to the sibling survival questionnaire, by year and cause. The estimates use the ICSS method to correct for survival bias . The survey concluded July 2, 2011, so the final bar reflects data for only half of the year.

Figure 5. Estimates of numbers of adult deaths per week in Iraq for 2-y intervals, 2001–2011, by cause as reported by siblings in the University Collaborative Iraq Mortality Study.

National estimate of deaths in Iraq between 2001 and 2011. Age-specific mortality rates were estimated separately within 2-y blocks (the first two time intervals are not strictly 2 y long, in order to align the first interval dividing point with the start of the war in March 2003; the survey concluded July 2, 2011, so the final bar reflects only half of the year). The counterfactual (had there been no war) estimate shows the predicted death counts if crude death rates had remained at their average level from 2001–2002 during the war and occupation (in gray). War-related, but not violent, deaths above the normal baseline are in the salmon-colored area. War-related violent deaths are portrayed in red. The estimates use the ICSS method to correct for survival bias .

Figure 6. Estimates of the probability of dying between age 15 and age 60-y intervals, 2001–2011, from sibling histories as reported in the University Collaborative Iraq Mortality Study, with plotted estimates from other studies.

Estimates of adult mortality risk (₄₅ q ₁₅, the risk of dying between the ages of 15 and 59 y), over the period 2001–2011, based on the sibling history survey. The first two time intervals are not strictly 2 y long, in order to align the first interval dividing point with the start of the war in March 2003; the survey concluded July 2, 2011, so the final bar reflects only half of the year. The estimates use the ICSS method to correct for survival bias . Mortality estimates from other studies are plotted as point estimates: IMIRA (part of the Iraq Living Conditions Survey) , the IFHS household survey (IFHS-HH), and the IFHS sibling survey (IFHS-SIB) .

Figure 7. Density of civilian deaths in Baghdad, with the cluster locations of the University Collaborative Iraq Mortality Study as well as 2006 cluster locations of a previous study .

The density map was generated using a kernel density estimation of civilian deaths from the Wikileaks Iraq War Logs release . The kernel density estimation provides a smoothed surface from a point pattern that represents spatial variation in the density of events, in this case civilian deaths. This allows for a crude visual analysis of the variation of events across space in Baghdad as well as the relation of 2006 and 2011 cluster locations to the density of civilian deaths. Analysis is based on geolocated data of all civilian deaths by any means reported in the Wikileaks Iraq War Logs data from 2004 to 2009.