Malaria that emerges in sudden epidemics needs to be treated differently from when it is transmitted continuously. Outbreaks often occur in remote rural areas, making them difficult to tackle with routine approaches to the disease. This policy brief outlines an early warning and response system being piloted in several African countries that could serve as a model for countries in Asia and Latin America.
The solution to malaria goes beyond geographical borders, as does the disease. In some areas, where malaria transmission is continuously transmitted (endemic), the disease is best contained by routine prevention and control measures.
But as well as endemic disease, and in other, non-endemic areas, malaria can also occur in sudden outbreaks, or epidemics. This can happen if environmental conditions change and mosquitoes that carry malaria are more able to breed, multiply, and come into contact with people.
Millions of Africans are affected by these epidemics every year, yet policies and procedures for minimising the impact of epidemics are either non-existent or inadequate. Part of the problem is disagreement over precisely what constitutes an epidemic, and its causes. Making matters worse, epidemics often occur in remote rural areas, sometimes among populations on the periphery of civil jurisdiction, thus challenging the normal routine approaches to controlling the disease.
To truly have an impact on the burden of malaria in sub-Saharan Africa, a new strategy needs to be in place that prevents and minimises epidemic malaria. It requires a degree of advanced planning, preparedness and response that has never before been achieved. This policy brief outlines a possible way forward — currently being implemented in sub-Saharan Africa — the Malaria Early Warning and Response System. The system aims to predict more accurately when and where malaria epidemics are likely to occur. If successful, the system would enable local control services to plan and act more swiftly and effectively to minimise casualties. It could also serve as a model for other epidemic-prone regions in Asia and Latin America.
Identifying epidemic malaria
Endemic and epidemic malaria are often treated equally despite needing different methods of prevention and control. Endemic malaria can be prevented and controlled largely through routine measures, whereas epidemics require flexibility and responsiveness to rapid change.
The greatest burden of malaria in Africa is in endemic areas where the disease is continuously present in the community. The environment encourages interactions between the Anopheles mosquito, malaria parasites and human hosts: surface water in which mosquitoes can lay their eggs; humidity for adult mosquito survival; and temperatures that allow both the mosquito and the malaria parasite to develop. When malaria control measures are inadequate, the disease distribution is closely linked with seasonal patterns of the climate and local environment.
The recently growing incidence of malaria in Africa is often wrongly attributed to epidemics. It is instead largely due to more gradual changes in endemic areas, such as demographic changes, increasing resistance of parasites to drugs, declining control infrastructure, co-infection with other diseases, and poverty.
Epidemic malaria tends to occur along the geographical margins of endemic areas, when the conditions supporting the balance between the human, parasite and mosquito vector populations are disturbed. This leads to a sharp but temporary increase in disease incidence. More than 124 million Africans live in such areas, and experience epidemics causing around 12 million malaria episodes and up to 310,000 deaths annually. 
Those most vulnerable to endemic malaria are young children who have yet to acquire immunity to the disease, pregnant women whose immunity is reduced during pregnancy, and non-immune migrants or travellers. By contrast, all age groups are vulnerable to epidemic malaria, because their exposure to disease is so infrequent that they have little immunity. 
In the case of 'classic' or 'true' epidemics, the change is brought about by natural causes such as large-scale climate anomalies in regions where the environment does not normally allow mosquito and parasite development.
Typically these involve desert-fringes (usually too dry to support transmission) or highland-fringes (usually too cool). The climate anomalies are often periodic and temporary, and there is a return to equilibrium. Examples include the epidemics occurring in the semi-arid areas of Southern Africa in 1996-1997, East Africa in 1997-1998, and the West African Sahel in 1999-2000, all of which were associated with wide-scale and unusually heavy rainfall.  Being poorly prepared, health services often become rapidly overwhelmed, leading to perhaps ten times more malaria-related deaths than in non-epidemic years, across all age ranges. [2, 4]
Alternatively, some epidemics are caused by human activities, such as an irrigation scheme in a warm, semi-arid environment; the displacement of people between highland and lowland regions; or a breakdown in pre-existing levels of malaria control.
Rarely, they may also follow the accidental introduction of an exotic mosquito species. These human-induced epidemics are less likely to return to the pre-existing equilibrium and may instead lead to endemic transmission (see figure below).
Figure: three different epidemic scenarios
Top: A 'true' epidemic, i.e. an infrequent event occurring in
areas where the disease does not normally occur. This type of epidemic is often
associated with warm dry regions. This type of epidemic may be cyclical in
Analysing previous epidemics helps understand their causes and offers opportunities for predicting and monitoring the conditions that are likely to give rise to new epidemics. This in turn enables the creation of an epidemic early warning system, which, when combined with a flexible control plan, can lead to epidemic prevention. 
The policy context
National and district malaria control programmes in affected regions are attempting to develop better means of predicting, preventing and controlling malaria epidemics. The countries of the Southern African Development Community, for example, are doing so as part of their commitment to the Abuja Targets for the Roll Back Malaria (RBM) initiative in Africa.  National malaria control services are expected to detect 60 per cent of malaria epidemics within two weeks of onset, and respond to 60 per cent of epidemics within two weeks of their detection. They recognise that to achieve these targets they need better information on where epidemics are most likely to occur, and ideally some indication of when they are likely to happen.
To help guide the process, the RBM technical support network on epidemic prevention and control, together with partnerships between World Health Organization inter-country programmes and national ministries of health has developed a new Malaria Early Warning and Response System (MEWS). [7 – 11]
This system has now been tested in several countries.  It involves new technologies, including seasonal climate forecasting, satellite observation of environmental conditions, geographical information systems and analytical tools designed to help the health sector interpret climate and surveillance data. It also involves new policies and activities for communicating and acting on the information obtained.
The MEWS strategy has five integrated components: 1) vulnerability assessment and monitoring; 2) seasonal climate forecasting; 3) environmental monitoring; 4) sentinel case surveillance, and 5) planning, preparedness and response.
Vulnerability assessment and monitoring
Many factors may increase a population's vulnerability to malaria epidemics, and increase the severity of disease outcome. In sub-Saharan Africa in particular, these include co-infection with other diseases such as HIV-AIDS, and resistance to anti-malarial drugs. Drought, floods, food insecurity and civil unrest can also increase vulnerability by forcing populations to move between areas of differing levels of malaria.
Information from agencies outside the health sector, such as food security and drought monitoring services, may help in assessing and monitoring vulnerability. Although this does not predict an epidemic, it does indicate whether there will be an increase in the risk and probability of more severe disease. With this information, control services can refine and assess their plans and capacity to respond to an epidemic.
Specific assessment tools may help in identifying which populations are most vulnerable and therefore most in need of immediate action. They may also be used to assess the appropriateness of management and policy decisions. 
Seasonal climate forecasting
Recent scientific advances, largely based on improved understanding of the interaction between sea surface temperatures and the atmosphere, have significantly improved our ability to predict seasonal climate several months in advance.  The degree of predictive skill involved varies from region to region, but is generally higher in the tropics.
Due to the inherent uncertainty of such predictions, control services should collaborate with regional and national forecasting services regarding their interpretation and implications for the coming season. This would enable them to prepare methods to prevent infection, procure effective drug supplies, and raise community awareness on personal protection. These could all take place before the main rainy season and three to six months before the peak malaria season.
Control services can also use routine information on relevant environmental variables such as rainfall, temperatures, humidity, vegetation status (indicating soil water availability) and flooding.  Although they have shorter lead-times (one to three months) than seasonal forecasts they are generally more reliable as they are based on direct observation.
Two sources provide the necessary information: periodic summaries from specialist centres (usually estimated from meteorological satellites) which are available via the Internet (e.g. http://iri.columbia.edu); and the national meteorological services’ ground-based weather station network.
The periodic summaries are generally free of charge, whereas data from meteorological stations may come at a price, which could present difficulties for health services in poor countries. Control services are being encouraged to discuss the specific information and support they may need with their national meteorological services. Environmental information would allow control services to focus more on local changes in conditions, which they can address with preventative interventions such as vector control and personal protection, including mosquito nets.
Surveillance and early detection
Epidemics need to be identified as quickly as possible to allow control services to decide how to distribute limited resources. Good sentinel case surveillance systems are essential for detecting any unusual increases in case numbers, and may involve several different methods for different settings.
These may detect the early phase of an epidemic, but if used in isolation they offer control services very short lead-times (one to three weeks) to plan and implement prevention measures.
Once an epidemic threshold has been passed, control services need to consider carefully whether to concentrate on ensuring drug distribution and effective case management, or embark on mosquito control, which requires considerable time to be effective.
Planning, preparedness and response
This most important component of MEWS encapsulates the system's primary purpose: early and effective responses to malaria epidemics, based on advanced planning of control and contingency measures.
Control services can plan their response by assessing the vulnerability of communities living in epidemic-prone regions. To raise confidence and accountability, control services also need to agree which early-warning and detection methods to use, their interpretation, and the choice and costs of preventative control options.
They need to identify contingency measures, the roles and responsibilities of collaborative partners, and also to define when and how much additional national resources they would require.
As time progresses, the development, adoption and use of the warning system should produce knowledge to make it more 'intelligent' and capable of more appropriate and effective responses to a malaria epidemic.
The southern African region, which has a long history of malaria epidemics, now has the most advanced integrated approach to epidemic malaria control based on the evidence of the key determinants of epidemics in the region. [14 – 16]
Experience and evidence for use of an integrated warning system approach within a national malaria control programme has been demonstrated in Botswana over the past few years. 
Other countries in the Southern African Development Community consider that this approach provides a useful framework for planning epidemic preparedness and response strategies. The WHO Southern Africa Inter-Country Programme for Malaria Control (SAMC) has supported them in exploring these tools further. 
As a result the region is now, for the second time running, preparing for the forthcoming (2005-2006) malaria season with a regional meeting at which national and local vulnerability is assessed in the context of the pre-rainy season climate forecasts that have been tailored to the malaria community.
The regional and national control planners recognize that southern African countries vary markedly in how endemic they are, or how many epidemics they have, and also in their capacity for surveillance and control coverage.
Tanzania, for example, is a highly endemic country with about 16-19 million cases per year. Botswana and Swaziland, by contrast, have recorded cases in the thousands and hundreds, respectively.
Zimbabwe's political and economic crisis has recently compromised its control programme, making its population highly vulnerable to epidemics. Meanwhile, Mozambique and Angola, are reconstructing their control programmes, after emerging from long-term conflict.
While an integrated MEWS approach may be relevant to certain districts in all southern African countries, its implementation remains challenging for many. Botswana has significant resources, a stable government and a good health service, and has a long history of mosquito control operations to prevent malaria. Several of its neighbours also have a history of effective mosquito control but, with the exception of South Africa and Swaziland, have neglected to maintain these systems in recent years.
The Global Fund for AIDS, TB and Malaria is tackling this by making significant resources available to national malaria control programmes to revitalise their malaria control infrastructure and strategies. In addition, the WHO Inter-Country Programme Teams are beginning to access funds from the African Development Bank to develop regional support for malaria control, allowing countries to gain value added benefit from the Global Fund's investment.
This is a promising development as cross-border cooperation in malaria control is vital: epidemic-prone areas are defined by environmental zones rather than administrative boundaries. For example, high rainfall in Angola may cause increased stream-flow into Botswana and Namibia, and create extensive breeding sites for mosquitoes. Drought, food security, or a range of other factors, may lead to people migrating across borders from one level of endemicity to another and pose a significant increase in epidemic risk. National epidemic risk maps should be taken to reflect the situation in neighbouring countries.
The successful implementation of the warning system will depend on close communication and co-operation among several partners in the health, environment and climate communities, as well as those involved with vulnerable populations, such as refugee and food agencies. This needs to take place at all levels from local to international.
Increasing our knowledge of environmentally sensitive diseases, strengthening health systems and making them better equipped to cope with a changing environment will help us now and in the future.
The MEWS approach will strengthen overall health information and surveillance systems, as some other diseases are also affected both spatially and temporarily by changes in the climate and environment.
Increasing our resilience to the threats posed by climate-sensitive diseases today may also help increase our ability to resist new threats brought about by any future changes in climate.
Stephen Connor and Madeleine Thomson are research scientists at the International Research Institute for Climate and Society, The Earth Institute at Columbia University, New York, United States.
 Worrall E., Rietveld A., Delacollette C. The burden of malaria epidemics and cost-effectiveness of interventions in epidemic situations in Africa. Am J Trop Med Hyg 71, 136-140 (2004).
 Kiszewski A. E., Teklehaimanot A. A review of the clinical and epidemiologic burdens of epidemic malaria. Am J Trop Med Hyg 71, 128-135 (2004).
 WHO. Malaria epidemics: forecasting, prevention, early warning and control - from policy to practice. World Health Organisation 59 (2004).
 Brown V., et al. Epidemic of malaria in north-eastern Kenya. The Lancet 352, 1356-1357 (1998).
 Connor S., Thomson M., Molyneux D. Forecasting and prevention of epidemic malaria: new perspectives on an old problem. Parassitologia 41, 439-448 (1999).
 Goddard L., et al. Current approaches to seasonal-to-interannual climate predictions. Int J Climatol 21, 1111-1152 (2001).
 WHO. Implementation of The Global Malaria Control Strategy: report of a WHO study group on the implementation of the Global Plan of Action for Malaria Control 1993-2000. World Health Organisation 57 (1993).
 WHO. WHO Expert Committee on Malaria, 20th Report. World Health Organisation 85 (2000).
 WHO. Malaria Control Today: current WHO recommendations. World Health Organisation 75 (2005).
 WHO. Africa Malaria Report. World Health Organisation (2003).
 WHO. Malaria Early Warning Systems, Concepts, Indicators and Partners, A framework for field research in Africa. WHO/Roll Back Malaria/Technical Support Network for Prevention and Control of Malaria (2001).
 Luers A.L., The surface of vulnerability: An analytical framework for examining environmental change. Global Environmental Change-Human and Policy Dimensions 15, 214-223 (2005).
 Ceccato P. et al. Application of Geographical Information System and Remote Sensing in Malaria Risk. Parassitologia 47 (2005).
 Craig M.H. et al. Exploring 30 years of malaria case data in KwaZulu-Natal, South Africa: Part I. The impact of climatic factors. Trop Med Int Health 9, 1247–1257 (2004).
 Craig M.H. et al. Exploring 30 years of malaria case data in KwaZulu-Natal, South Africa: Part II. The impact of non-climatic factors. Trop Med Int Health 9, 1258–1266 (2004).
 Thomson M.C. et al. Rainfall and Sea-Surface temperature monitoring for malaria early warning in Botswana. Am J Trop Med Hyg 73, 214-221 (2005).
 DaSilva J. et al. Improving Epidemic Malaria Planning, Preparedness and Response in Southern Africa. Malaria Journal 3, 37 (2004).