Tuesday, March 09, 2010

Estimating the number of civilian deaths from armed conflicts: How to fill the information void?

A research paper on estimating civilian deaths during armed conflict has recently been published in The Lancet medical journal. The paper looks at conflict in the Darfur region of Sudan and a new method for estimating civilian fatalities. This involved building a statistical model from the results of surveys that measured mortality and met certain quality criteria. In all, 107 surveys reported mortality data from Darfur between 2003 and 2008 and the researchers were able to include 63 of these results in their statistical model. [The Lancet]

In an accompanying commentary the limitations of the available data and analysis approach is discussed. Checchi points out that although the number of surveys available from Darfur seems large, they included only 16% of person-time at risk. This, he argues, is an indicator of the information gap that remains to be filled in nearly all large scale crises. Further, "..this limitation suggests the usefulness of mapping information coverage in real time to draw attention to regions where information is lacking or outdated, and to coordinate efforts to gather data. A corollary initiative would be to track the number and location of conflict-affected people in real time." [The Lancet]

What is striking is the paucity of mortality survey data collected on the conflicts covered by this site, namely Afghanistan and Iraq, compared to that in Darfur. Between 2007-2009, the period in Afghanistan when fighting escalated markedly, no mortality survey data has been gathered. Attempts to document civilian fatalities rely on case-by-case investigations by organisations such as the Afghanistan Independent Human Rights Commissions [AIHRC]. Such approaches are almost inevitably going to result in partial coverage and under-estimation of the fatality burden. In addition, fatalities represent just the tip of the iceberg in terms of the total casualty burden, and there seems to be no attempt at all to try and estimate this larger figure in the civilian population.

As Checchi says in his introduction:
"Imagine that you are helping a population cope with the health effects of armed conflict, but have no information about whether their health is improving or not, or whether your programmes adequately address their burden of disease.

...Systematic measurement, analysis, and programmatic use of essential health indicators (mortality rate, prevalence of malnutrition, and coverage of essential services such as vaccination, and water and sanitation) remain the exceptions, despite being predicated in various manuals, policy documents, and meetings."
There are, of course, many reasons why it is so difficult for public health professionals to collect and document information of the health impacts of conflicts such as Afghanistan. However, in cases where initiation of the conflict has been actioned or facilitated by the UK government it must surely have a special responsibility to ensure that reliable data on health impacts are collected and made publicly available, so that humanitarian relief efforts can be appropriately planned. When access by independent scientists is made impossible by insecurity then that role, arguably, should be temporarily assigned to the military. Embedded-journalists are of course notorious for producing one-sided and partial coverage of war news. Could embedded-epidemiologists, following transparent and internationally recognised approaches produce credible data? Its a big question, but given the massive information gap that currently exists, it must surely be worth exploring.

(NB The side bar links to other key research articles and data sources have been updated, with the removal of some dead links and re-focusing of others on the most relevant pages.)