|
|
| |||||||||||||||||||||||||||||||||||||
In the event of a military invasion and ongoing war, however, the likelihood of obtaining good demographic data plummets. A death registry is unlikely to be developed or maintained, and as conditions deteriorate, it may become increasingly unlikely that bodies can be counted at all. In Iraq, there is also a strong cultural imperative that bodies be put to rest quickly, which may affect the ability to arrive at accurate estimates. Although sentinel populations are commonly monitored to rapidly estimate mortality in developing countries when a registry is not available, the impossibility of finding reliably representative populations in countries engaged in armed conflict and the absence of an accurate population count make it difficult to extrapolate from the rates at sentinel sites to produce reliable national estimates.
A more accurate option, but one that is more dangerous for researchers, is the household survey.1 Even in nonemergency situations, the study design of the survey may be subject to underreporting and may not accurately reflect rates of migration and fertility. Complex computations are required to account for variation among regions and subpopulations. During conflicts, an estimation of the death rate is further complicated by the difficulties involved in creating a valid sampling frame, the problem of reporting bias, and obstacles to accurate ascertainment of causes of death. Researchers often must risk their lives if they wish their estimates to accurately represent the population, and they must spend as much time in dangerous areas as in less dangerous ones to minimize bias.
In this issue of the Journal (pages 484–493), the Iraq Family Health Survey (IFHS) study group reports the results of a household survey conducted in Iraq in 2006 and 2007 for the purpose of estimating mortality between January 2002 and June 2006. The researchers divided the country into 56 mutually exclusive sampling strata and surveyed 9345 households in total (see map). Information on all deaths within a household was sought, in an effort to estimate overall and cause-specific rates of death. The group obtained an estimate of 151,000 violent deaths, with a purported 95% confidence interval of 104,000 to 223,000 — a massive death toll — since the 2003 invasion. Violence was found to be the leading cause of death among Iraqi men between the ages of 15 and 59 years and a leading cause of death among Iraqi adults in general.
|
|
Meanwhile, there is ongoing discussion4 about the validity of the study by Burnham et al. The survey methods have been scrutinized, and observers have put forward convincing arguments both that it does and that it does not overestimate mortality. One of the issues under debate is whether the clusters that were surveyed were nonrandomly distributed owing to "main-street bias" (an oversampling of highly trafficked areas). What cannot be debated is that it was a much smaller study (1849 households in 47 clusters) than that conducted by the IFHS (9345 households in 1086 clusters).
Though the IFHS study group should be commended for its attempt to capture the highest-quality results, uncertainties remain. The survey design, in particular, is certainly open to criticism, and the authors honestly admit the shortcomings of their analysis. For example, sometimes it was problematic or too dangerous to enter a cluster of households, which might well result in an undercount; data from the Iraq Body Count on the distribution of deaths among provinces were used to calculate estimates in these instances. If the clustering of violent deaths wasn't accurately captured, that could also increase uncertainty.
The sampling frame was based on a 2004 count, but the population has been changing rapidly and dramatically because of sectarian violence, the flight of refugees, and overall population migration. Another source of bias in household surveys is underreporting due to the dissolution of some households after a death, so that no one remains to tell the former inhabitants' story. Mortality estimates that are derived from surveying deaths of siblings were also calculated, but this method may also be subject to such underreporting. However, the IFHS group's overall estimate of mortality does reflect uncertainty in the level of underreporting (20 to 50%) and the instability of the projected population during the post-invasion period (1 million to 2 million).
Under the current conditions in Iraq, it is difficult to envision a study that would not have substantial limitations. The circumstances that are required to produce high-quality public health statistics contrast starkly with those under which the IFHS study group worked. Indeed, it must be mentioned that one of the authors of the survey was shot and killed on his way to work.
We cannot begin to explore all the political implications of this work, which will no doubt shape international public opinion regarding the war in Iraq. What we can discuss, however, are the implications of this work for mortality-estimation research in areas in which violent conflict persists. The addition of this study and comparisons between it and others will permit a better understanding of the relationship between study techniques and mortality estimation that can, in turn, provide a basis for improved guidelines to decrease bias in future studies. When alternative data sources become available, the IFHS and the other surveys conducted to date will be verified or refuted, along with their survey techniques. The goal should be to ensure that decisions regarding epidemiologic methods for estimating mortality in high-risk populations are based on the best possible evidence; these estimates are critical for the effective implementation of humanitarian relief efforts.
No potential conflict of interest relevant to this article was reported.
Source Information
Ms. Brownstein is a doctoral student in the Department of Genetics at Yale University, New Haven, CT. Dr. Brownstein is a faculty member in the Children's Hospital Informatics Program, Children's Hospital Boston, and an assistant professor of pediatrics at Harvard Medical School — both in Boston.
An interview with Dr. Mohamed Ali of the Department of Measurement and Health Information Systems at the World Health Organization can be heard at www.nejm.org
This article (10.1056/NEJMp0709003) was published at www.nejm.org on January 9, 2008.
References
| |||||||||||||||||||||||||||||||||||||
This article has been cited by other articles:
HOME | SUBSCRIBE | SEARCH | CURRENT ISSUE | PAST ISSUES | COLLECTIONS | PRIVACY | TERMS OF USE | HELP | beta.nejm.org Comments and questions? Please contact us. The New England Journal of Medicine is owned, published, and copyrighted © 2009 Massachusetts Medical Society. All rights reserved. |
Audio Interview
