Current survey-based techniques used to estimate violent war deaths have been criticised for their potential biases and inaccuracies. For instance, surveys estimating war mortality on the basis of household deaths, such as those recently done in Iraq, were alleged to be statistically invalid, and to incorporate politically motivated over-reporting of deaths.

But the alternative technique used during most ongoing conflicts including Iraq—passive data collection from eyewitnesses and media reports—is also subject to major biases, including the fact that high levels of war deaths occur in dangerous areas where eyewitnesses are least likely to go.

To overcome some of these biases, Ziad Obermeyer and colleagues from the Institute for Health Metrics and Evaluation in Seattle, designed a new method of estimating violent war related deaths using data on the siblings of respondents in large household surveys conducted in peacetime.

According to the researchers, by comparing passive surveillance data of violent war deaths (mainly from eyewitnesses and the media) in 13 countries over the past 50 years, to peacetime data collected after conflicts in the UN's World Health Surveys, they were able to provide more accurate data on war deaths. For example, say the researchers, the new technique avoids the constraints imposed by active combat, and using siblings' histories rather than household deaths, reduces double counting and exaggeration of deaths.

They estimate that 5.4 million deaths occurred as a result of war in the 13 countries studied between 1955 and 2002, from 7000 in the Democratic Republic of Congo to 3.8 million in Vietnam.

The researchers point out that these estimates are on average three times higher than those obtained from previous reports. For example, they estimate that 378 000 people died a violent death as result of war each year between 1985 and 1994, compared to previous estimates of 137 000 people during this time.

The largest differences, they say, were in Bangladesh where, during the conflict for its independence, they estimate there were 269 000 violent war deaths, compared to previous estimates of 58 000. And in Zimbabwe where they estimated 130 000 deaths, compared to previous estimates of 28 000.

Importantly, say the authors, these new data do not support the prevailing view that war deaths are declining and have been since the mid-twentieth century, or that recent wars have killed relatively few people thanks to technological and strategic innovations designed to minimise civilian deaths. In light of the substantial differences in estimates, conclude the authors, these claims need to be re-evaluated.

Even these figures are still likely to underestimate the importance of conflict as a cause of death because they only address violent deaths, cautions Professor Richard Garfield from Columbia University in an accompanying editorial. "In the poorest countries, where most conflicts now occur, a rise in deaths from infectious diseases often dwarf the number of violent deaths during a conflict".

Garfield argues that the "promising method" pioneered by Obermeyer and colleagues will force us to re-evaluate our assumptions about these deaths "The importance of war as a public health problem and a social problem makes this imperative", he concludes.

Source: BMJ-British Medical Journal