Between April and June, 2004, MSF did retrospective cluster surveys among 215,400 internally displaced people in four sites of West Darfur (Zalingei, Murnei, Niertiti, El Geneina).

This study, which was done in a difficult setting, provides epidemiological evidence of this conflict's effect on civilians, confirming the serious nature of the crisis, and reinforcing findings from other war contexts.

The four surveys we present, which were done in difficult and insecure field conditions, are representative of about 215,400 internally displaced people — 43% of the total estimated population of 500,800 internally displaced people in West Darfur in July, 2004. These findings provide data on demographic events affecting a large Darfurian population of internally displaced people in the periods before and after displacement, and clarify the contribution of violence to mortality in this population.

We believe that clear trends emerge from our findings.

Most people in the populations we surveyed fled as a result of attacks on their villages. While in the village or in flight, mortality was extremely high, overwhelmingly because of violence. Although men were at far higher risk of being killed, women and children were also targeted. Separations and disappearances were also common, mostly affecting men. Adding these absentees to men who were killed during the recall period largely explains the skewed nature of the age/sex pyramids.

Surveys done among refugees in Chad also showed raised mortality (again, in great part violence-related), although mortality rates were lower than in Darfur, possibly due to the fact that these groups fled to safety in the early stage of the conflict.

Among internally displaced people settled in camps, violence continued to cause a substantial proportion of deaths. Although we did not systematically record information on the circumstances of these killings, data from MSF health centres suggest a high incidence of shootings, beatings, and rapes (unpublished).

Even when violent causes were excluded, mortality remained unacceptably high after arrival in the camps, with alarming peaks in El Geneina, where little humanitarian aid, apart from irregular food distributions, was dispensed before June, 2004, and where we found a 50% weekly attack rate of diarrhoea among children (data not shown).

Generally, existence was precarious in these hard-to-access, poorly serviced camps; people there were under constant threat from malnutrition and epidemics, and were deprived of most coping mechanisms (farming in particular).

The onset of heavy rains in July, 2004, was expected to hamper transport, interrupting vital food supplies, worsen sanitary conditions, and expose vulnerable populations to seasonal malaria and waterborne diseases. A further increase in mortality rates therefore seemed likely, and was, indeed, shown by a region-wide WHO survey targeting a 3-month postdisplacement period.

Surveys such as these have important and welldescribed limitations. When mortality rates are high, entire households may disappear (survival bias): their experience is thus not reflected in the sampled population, and mortality rate is underestimated. Recall bias is difficult to measure, but when retrospective periods are long, as in our Murnei and Zalingei surveys, less recent deaths might be under-reported, leading to an underestimate of mortality rate.

On the other hand, households might recall traumatic events as having occurred more recently than they actually did, leading to an overestimate of recent mortality.

In our surveys, however, we were able to cross-check months of death with place of death and time of arrival, which probably reduced bias related to date recall.

Incorrect reporting of age could have caused a misrepresentation of the age/sex distribution; however, the all-age/sex ratio does suggest an overall imbalance. Systematic or nonsystematic errors might also have resulted from linguistic barriers, since we do not speak Arabic.

The WHO/Expanded Programme on Immunisation cluster design has not been fully validated as a tool to measure mortality, although no alternative methods have been clearly established for settings, such as ours, where systematic sampling is unfeasible due to the absence of an individual household sampling frame. In particular, design effects were higher than expected because of clustering of violent deaths. Surveys such as ours should plan for design effects of greater than two when measuring causes of death that, like violence, are likely to be very clustered.

Our findings do not in themselves substantiate claims that events in Darfur amount to genocide, not least because this would require demonstration of such an intent on the part of the perpetrators, which is clearly beyond the scope of an epidemiological survey.

Nevertheless, we believe that, in the four sites we surveyed, high mortality and family separations amount to a demographic catastrophe. While our data reflect the striking extent of killings, systematic accounts of other crimes, such as mass rape, have been put forward.

Satellite maps have also suggested widespread village destruction. Together, these findings strongly suggest that between 2003 and 2004, populations in West Darfur experienced massive attacks against life and property.

One of the most serious and long-lasting consequences of such attacks may be widespread mental trauma among survivors and witnesses. Although we did not survey this issue, we believe that it is largely overlooked in the present Darfur relief context.

The Darfur crisis resembles most armed conflicts, small and great, that have affected the world—in particular Africa—in recent years. The victims are mostly civilian.21,22 Women and children are not spared, although in West Darfur the risk of being killed was far higher for men. Displacement also results in excess mortality and loss of livelihoods, creating chronic dependence on aid.25 Aid itself is insufficient and late, often due to a deadly combination of international neglect and warring parties who do not grant humanitarian access to the affected populations when they need it most.

West Darfur's case seems exceptional because of the overwhelming contribution of violence to mortality, resulting in crude mortality rates that were actually higher than mortality rates among children younger than five years, contrary to what is commonly observed.

In humanitarian emergencies, field epidemiology can, in addition to helping to orient and evaluate aid programmes, provide key scientific testimony about past events. Overwhelming epidemiological evidence on the consequences of armed conflicts, including that provided here, should be a tool to advocate respect for international humanitarian law.