Mortality surveys in the Democratic Republic of Congo: humanitarian impact and lessons learned

On 6 February 2000, the New York Times published an in-depth, front-page article on the then 17-month-old conflict in the Democratic Republic of Congo (DRC). The article gave a nuanced account of what had been called ‘Africa’s first world war’, describing in detail the complex history and root causes, the regional politics, the interests of the involved parties and the international diplomatic response. What was most striking from the humanitarian perspective was the article’s clear underestimate of the human impact of the conflict. In particular, the reported death toll of 100,000 failed to convey the true scale or nature of the humanitarian crisis.

A limited mortality survey by the International Rescue Committee (IRC) in eastern DRC estimated that 12,000 excess deaths had occurred in Katana health zone alone by the time of the New York Times report. A toll of 100,000 for the entire country therefore seemed unrealistic. The newspaper’s significant under-reporting prompted the IRC to undertake a more comprehensive mortality survey in eastern DRC. Between 2000 and 2004, the IRC conducted four major surveys in DRC – two of them focused on the eastern provinces, the latter two countrywide. The aim of this article is to describe the value and impact of these surveys, and the key lessons learned.

Documenting the humanitarian impact of war

Perhaps the major contribution of the series of surveys has been an improved understanding of the humanitarian impact of modern-day conflict in DRC and beyond. The first of the comprehensive studies estimated 1.7 million excess deaths in eastern DRC between August 1998 and May 2000. This was the first epidemiologically sound study of mortality in the Congo war, and alerted the international community to a death toll well in excess of that previously reported. The study received widespread – albeit brief – media attention, and its objective data helped to draw humanitarian and political attention to the under-reported crisis.

The three subsequent studies have revealed the conflict to be the deadliest since the Second World War. Although the Congolese war officially ended in December 2002 with the signing of a peace accord, fighting and insecurity have continued in large areas of the east of the country. Up to April 2004, a total of 3.9 million excess deaths had been attributed to the conflict. This figure dwarfs the death tolls of all the high-profile natural disasters and acts of terrorism of the past decade – in fact, it is more than four times the total number of deaths from all such disasters combined over the past ten years.

In addition to documenting the scale of the crisis in DRC, the surveys also provided an insight into its nature by documenting the major causes of mortality. Most deaths were due to easily preventable and treatable diseases – less than 10% of all deaths were directly due to violence (this figure was around 2% for the last two surveys). The vast majority of deaths resulted from malaria, respiratory infections, diarrhoea, measles and malnutrition, reflecting the major social, economic and political disruption caused by the war. The fourth survey also demonstrated the strong statistical association between insecurity and increased mortality, indicating that improvements in security were the most effective means of limiting the death toll.

The demonstrated value of the Congo surveys contributed to a more consistent attempt by various agencies to document the scale and characteristics of war. Valuable mortality studies have since been conducted in Darfur, Iraq and Uganda, and the data has been used to advocate on political, security and humanitarian issues. All of these studies have contributed to a growing body of scientific evidence on the human impact of armed conflict.

Mobilising financial resources

Levels of humanitarian aid and international political engagement in DRC have remained completely out of proportion to need since the onset of the conflict. This situation was particularly bleak prior to 2001, and could in part be explained by the poor understanding of the scale of the crisis. Global aid contributions from all donors to DRC in 2000 totalled a paltry $29 million. Following the release of the 2000 survey results, total humanitarian aid increased by over 500% between 2000 and 2001. The United States’ contribution alone increased by a factor of almost 26. It is probably fair to assert that the mortality data played a significant role in increasing international assistance.

Despite these early, apparently dramatic improvements, aid to DRC in 2006 has still not reached the levels necessary to adequately address humanitarian needs. The release of each IRC survey has been accompanied by calls to increase assistance to the affected population. But this advocacy has never been as effective as it was after the release of the initial survey – in fact, between 2001 and 2004 aid levels appeared to plateau.

Data from the 2004 survey, which demonstrated that 38,000 excess deaths continue to occur per month, helped to keep the humanitarian situation in DRC before donors and was probably a factor in the 45% increase in total aid between 2004 and 2005. The study is also widely referenced in the United Nations’ Action Plan 2006: Democratic Republic of Congo, which calls for a further dramatic increase in aid – to $682 million. Nonetheless, total aid to the affected population is still only $11 per person per year, and compares poorly with crises in Darfur ($189) and Northern Uganda ($102). To date, only 30% of the 2006 Action Plan has been funded.

Influencing policy

The IRC surveys have helped to establish excess mortality as the most useful metric for communicating the scale of humanitarian need during a crisis. Data from surveys are the most widely accepted and frequently quoted measures to describe the humanitarian situation in DRC. This data has also been used by many agencies to advocate for policy change with decision-makers.

Data from the surveys has been referenced in policy documents and pronouncements by the United Nations, the World Bank, the European Union and the governments of Britain and the United States. The UN’s Action Plan refers to the data in making its case for three key areas of intervention: saving lives, building a protective environment and promoting stability. Legislation currently before the US Congress directly quotes data and recommendations from the IRC in its proposed policy on relief, security and political transition in DRC. Since 1999, the UN Security Council has issued 38 resolutions on DRC, and refers to the humanitarian situation regularly in making its demands for progress in both the political and security realms.

Nonetheless, the experience of the significant, but still insufficient, increases in aid to DRC highlights a fact familiar to many public health and humanitarian professionals: the presence of objective data rarely results in swift, effective policy developments. Incremental change is more often the case – as we continue to see with respect to the international community’s response to the political, security and humanitarian situation in DRC. Notwithstanding the policy developments described above, the response of the international community remains out of proportion to the documented need in DRC, and further advocacy is required. It is largely for these reasons that IRC has continued to conduct regular mortality surveys in DRC, to provide updated data on the humanitarian situation.

Another important lesson from the surveys was that the method by which study findings are released can significantly influence the ability to influence policy. After lengthy internal debate, the results of the first survey in 2000 were not released through a peer-reviewed publication, as the lead investigator had recommended. Rather, negotiations with the New York Times led to front-page coverage – in exchange for the Times having the right to release the story. Similarly, the 2001 survey was initially reported on the front page of the Washington Post. For the purposes of securing the attention of policy-makers and the media, this was probably the most effective means of releasing the results of the studies. Subsequent peer-reviewed publication of the third and fourth surveys has helped to validate the methods and findings, as well as leaving an important footnote in the literature and making these studies more accessible over the longer term.

Prioritising, targeting and evaluating interventions

While data from the surveys has been invaluable for advocacy, IRC is primarily a humanitarian agency and has used the data to guide the implementation of its own public health interventions. The data has been used to design, target, evaluate and scale up health activities, especially in the most seriously affected areas in the east. Programme emphasis has been on ensuring access to good-quality primary health care services, through the support of facilities and systems at the health zone administrative level.

Data from the surveys and other programme indicators has been useful in demonstrating the effectiveness and impact of the programmes. In three eastern health zones, for example, IRC was able to document significant reductions in excess mortality over the course of its programme. While IRC does not claim that its health programme was directly responsible for these declines in mortality, major increases in health service utilisation over the same period suggest that the programmes contributed to the general improvement in the health status of the community.

Survey data has also contributed to a further scaling up of health programme interventions, especially in infectious disease control, child survival and reproductive health. IRC health programmes now reach an estimated 1.35 million people in 11 health zones in three provinces.

Contributing to the science and practice of field epidemiology

A significant contribution of the first mortality survey was that it demonstrated that valid, population-based data could still be collected in austere, relatively insecure areas, even when population numbers were unknown. The successful conduct of the study required good local knowledge, strong logistics support and flexibility in data collection methods. This survey used a combination of methods to sample the population, depending on the context. In fact, it was the first major mortality survey to our knowledge to use systematic spatial sampling, and the first to use random points identified with a Global Positioning System (GPS) unit for selecting households. While three of the five districts were surveyed using classic cluster sampling methodology (sampling proportional to population size), the two districts for which the population numbers were unknown were surveyed via spatial sampling methods.

A combination of cluster, spatial and systematic random sampling methods was used for the subsequent surveys. In practice, cluster sampling is the most commonly used methodology for mortality surveys. This method was initially designed and validated to estimate vaccination coverage – it was not developed to measure mortality rates. But, while not ideal for that purpose, it is the most practical method available in humanitarian settings.

One limitation of the cluster methodology is that larger samples are required to provide as precise an estimate of mortality as would be given by other more rigorous, yet impractical, methods, such as simple random sampling (the method used during opinion polls). Epidemiologists often refer to the ‘design effect’ – the factor by which the sample size is increased to give as precise an estimate when using the cluster method compared to simple random sampling. Classically, the design effect is assumed to be two when estimating vaccination coverage or nutritional status. But the IRC surveys have shown that, during mortality surveys, the design effect may be significantly higher, and even larger sample sizes may therefore be required. For the IRC’s 2004 survey, a design effect of four was assumed and a sample size of 19,500 households chosen. This is the largest such survey ever conducted in a humanitarian setting, and again demonstrated that such comprehensive surveys can be conducted, given the appropriate technical and logistic support.

The surveys have also contributed to several important debates within the humanitarian community. As noted, they have played an important role in establishing excess mortality as the most important measurement for estimating humanitarian need. They have helped to stimulate the exploration of improved field epidemiology methods in humanitarian settings – not only of sampling methodologies, but of questionnaire design and cause of death determination. They have also contributed to initiatives to make the design, conduct and reporting of mortality surveys more consistent across agencies and contexts. Finally, they have suggested a mechanism by which the effectiveness of humanitarian response – political, security and relief efforts – can be most effectively monitored over time. That is by tracking the most important indicator of all humanitarian indicators: excess mortality.

Sadly, based on the series of IRC surveys, the humanitarian response in DRC, in all its dimensions, has fallen well short of what has been required to address the documented needs.

Richard J. Brennan (Richard.Brennan@theirc.org) and Michael Despines (Michael.Despines@theirc.org) work for the International Rescue Committee. Leslie F. Roberts (les@a-znet.com) works in the Program on Forced Migration and Health, Mailman School of Public Health, at Columbia University.