Death in Iraq


I suppose that data on mortality rates is never uplifting, but some data made available online last Thursday seems especially sober. Researchers at the Johns Hopkins Bloomberg School of Public Health, in collaboration with a researcher at Al Mustansiriya University in Iraq, estimate that 655,000 Iraqis have died as a consequence of our invasion (Burnham, G, et al, Lancet (2006)). The research was published in the well-regarded medical journal, Lancet. It was led by Les Roberts, who also led a survey of deaths in Iraq published in October 2004 (Roberts, L, et al, Lancet (2004)) that was largely ignored.

The timeliness of the publications is, in all likelihood, not a coincidence — both the 2004 and last week’s articles came out weeks before a national election. Dr. Roberts is a distinguished member of academia and, therefore, is statistically expected to be a critic of the Bush administration. Naturally, this timing is an easy and obvious weakness in terms of the neutrality of the authors and the credibility of the study. These are, after all, epidemiological studies — there are many places for human bias to enter into the data collection either intentionally or not. On the flip side, one can criticize the authors (as Tom Engelhardt and Judith Coburn did) for putting the data out at a time when it can get overshadowed by trivial election distractions. Why fabricate data if your strategy for timely publication can backfire (twice)?

With all of this potential bias on the part of the researchers, it is interesting to note that this study was funded by MIT and Johns Hopkins, not by the typical source of research funding in America: the US government. If it were funded by the state, there would be no chance of bias, right? Come to think of it, why isn’t the government funding an accurate casualty count in Iraq?

There was some scholarly discussion of the 2004 article, in the form of letters to the editor of Lancet (Volume 365, page 1133–4). The authors were criticized for their survey method, known as cluster sampling, and for various areas where researcher bias could have been introduced as well as a specific design flaw that could introduce false overestimates of death. As the authors point out in their response, the cluster sampling method they used is recommended by UN and US agencies. Since then, a team of researchers has shown that the cluster sampling method, when done properly, is as reliable as another leading survey technique (Rose, AMC, et al, Bulletin of the World Health Organization. (2006)). In my opinion, the authors address the other potential biases and design flaws specifically in the original article, although they could be lying. I have no reason to believe that they are lying, but the possibility must always be acknowledged.

These criticisms centered around the validity of the statistics. Statistics says that if you roll one die six times, you will roll each number once. However, if you have played a board game, you know this is not what really happens most of the time. What statistics really says is that if you roll one die six hundred times, you will roll each number close to 100 times. If you really wanted to do this, the actual number of times that you got each number would in fact be very close to 100. So the estimate that there are an extra 655,000 Iraqis dead because of the invasion is simply that: an estimate. An error on this estimate can also be calculated, which can be used as a guide of how accurate that original estimate is likely to be. If the study is done properly, the errors will be small and the estimate will be accurate.

While there may be flaws with the study, the overarching conclusion will not change: many more people have died in Iraq since its occupation than would otherwise be expected. There is one set of raw numbers that backs this up: 547 people died between March, 2003 and June, 2006 in areas where 82 people died between January, 2002 and March, 2003. The post-invasion period of 39 months is about two and three-quarters times as long as the pre-invasion period of 14 months, but 547 is over six and one-half times as many people as 82. If our invasion had no effect in these areas, the number of people dead would be directly proportional to the amount of time that passed, but it is not even close to proportional.

The authors estimate that 19.8 deaths per 1000 people occurred between June 2005 and June 2006. This is almost 2 deaths per 100 people. Since this study was based on interviews of households, there was a standard cross-section of ages and sex that you would find in any family. Are you related to 100 people? Have 2 of your relatives died every year? Between my husband and myself, I can say I know just about 100 of our relatives. Great Uncle Mike passed away last spring, which is the only family death in five years. The annual rate of death in our family is a little lower than what families in Iraq were experiencing before March, 2003 (0.2% in our family versus 0.55% estimated in Iraq).

The post-invasion mortality rate was 7.9 deaths annually per 1000 people in the 2004 study. If Dr. Roberts and his colleagues had somehow affected the politicians of this nation with his last article, and inspired a de-occupation of Iraq, would there be an extra 500,000 people dead? If we do not leave Iraq in 2006, what will the Iraqi mortality rate be in 2008? As you consider whether it is the moral obligation of the US to leave Iraq soil, keep in mind that, of the 547 post-invasion deaths recorded by the researchers, 13 were children under the age of 15 killed by coalition air strikes.

October 18, 2006