Forecasting malaria epidemics

Malaria epidemics are thought to be occurring more frequently than in the past. The reason for this increased frequency is widely debated. However, the most often cited factors are abnormal climate, modifications to the environment, increased parasite resistance to antimalarials and population movements.

An estimated 100 million people live in malaria epidemic-prone areas in Africa alone. Epidemic-prone areas are areas where little malaria usually occurs, such as highland zones, arid areas and desert fringes. Because transmission is low, people in these areas are not exposed to malaria as often as people in higher transmission settings, and thus develop little or no natural immunity against malaria parasites. As a result, malaria epidemics often lead to very high levels of morbidity and mortality.

The early detection and containment of malaria epidemics is one of the four key priority areas for the Roll Back Malaria (RBM) campaign. In the RBM framework, epidemics are ideally detected within two weeks of onset. However, even if malaria epidemics are detected within this period, there is rarely enough time to act to reduce their impact.

What is a malaria epidemic?

The definition of a malaria epidemic depends on the local malaria situation. It is not possible to develop a global definition or threshold for malaria epidemics because malaria cases in epidemic-prone areas are seasonal, and vary greatly from year to year. Historical malaria records at health centres can be analysed to ascertain the ‘normal’ pattern for that area. If cases of malaria in a season increase greatly from the expected pattern, this may indicate an epidemic. The definition of where an epidemic can occur is another important, but difficult, question. For example, no malaria transmission is expected in areas above 1,750m, which means that just one malaria case in such an area could constitute an epidemic. However, this is too much of a generalisation as there is no firm line between high transmission on one side, and low transmission on the other.

Although difficult, defining when a malaria epidemic occurs is still vitally important. If an epidemic is declared, resources can be mobilised and emergency measures put in place. These measures will be costly if it turns out that there is no epidemic. On the other hand, caution in declaring an epidemic will result in a late response, which will have no effect other than to reassure people that something is being done.

How is malaria affected by climate?

Temperature, humidity and rainfall are vital factors in regulating the development of both mosquitoes and parasites. Temperature increases the development rate of the parasite during its lifecycle in the mosquito. Below 19 degrees centigrade for P. falciparum, and 15–16°C for the other species, parasites are unlikely to complete their cycle and hence spread further to humans. Temperature also affects the ability of mosquitoes to transmit malaria. Temperatures between 22°C and 30°C increase the potential lifespan of the mosquitoes, and increase the frequency of blood meals taken by the females, to up to one meal every 48 hours. Higher temperatures also increase the development of larval mosquitoes, shortening the amount of time it takes the mosquito to develop from egg to adult. Rainfall generally leads to a proliferation of new breeding sites for mosquitoes. However, rain can also destroy existing breeding sites; heavy rains can flush larvae and eggs out of breeding pools. Conversely, exceptional drought conditions can turn streams into pools, making them ideal for breeding. So, if rainfall, temperature and humidity can be measured, this could yield information on the level at which mosquitoes are breeding and surviving in a given area.

Measuring the weather

Many countries in Africa have networks of weather stations measuring climate for agricultural purposes. However, local weather data is not always available or accurate, and other, indirect, climate monitoring tools are needed. Such information can now be provided from satellites. The measurement of vegetation and terrain through satellites has advanced rapidly since the 1970s, and measuring the factors that affect the development of mosquitoes is now possible from satellite-derived data. Some of these factors are described below.

Cold Cloud Duration

The cloud towers which cause rainfall in the tropics are called cumulonimbus, from the Latin cumulus meaning ‘heaped’, and nimbus meaning ‘rain’. Clouds with the coldest top surface produce the heaviest rainfall. It has become possible to estimate the amount of rainfall by measuring the temperatures of cloud tops using infrared images derived from satellites.

Certain temperature ranges will cause clouds to precipitate into rainfall. Generally, when the cloud-top temperature falls below approximately minus 40°C, some rainfall is occurring below. Such clouds are called cold clouds, and are visible from space. Satellites are able to measure the temperature of the cloud tops by remote sensing. The amount of time a cloud-top temperature remains below a defined threshold is known as the Cold Cloud Duration (CCD). Using CCD, we can develop an overview of the rainfall pattern over wide areas.

Vegetation status

The growth of vegetation (specifically primary vegetation) is related to rainfall in tropical areas. The relationship between rainfall and vegetation means that, by measuring vegetation through remote sensing, we can estimate both the timing and the amount of rainfall. Changes in the vegetation index over time have been shown to be a good indictor of the level of malaria cases.

Hydrology

Rainfall alone may not be a good indicator of malaria transmission risk in warm semi-arid areas. What happens to rain after it has fallen is an important factor, which can be measured by satellite or through examining the water levels of rivers or dams.

Putting it together: what is done with all the information?

By combining information collected through satellites, weather stations and health centre data, researchers have developed statistical models to help predict malaria transmission. Statistical models are mathematical representations, or simulations, of an actual situation or process. These can be either quite simple or, as is normally the case, they can be complex models incorporating many variables. Statistical models have been created which incorporate CCD, vegetation indices, meteorological conditions and historical malaria case data to describe possible relationships between these factors and whether these relationships can tell us anything about future malaria transmission. Using these models, in some areas, malaria cases can be predicted up to a month in advance of an epidemic.

Is epidemic forecasting the answer to all our problems?

The short answer is no. Malaria epidemic forecasting is part of a wider framework, which involves identifying geographical areas or populations of epidemic risk, forecasting epidemics at a regional or national level, and providing early warning and detection of malaria epidemics.

Long-range forecasting

Long-range forecasting is a general system which uses weather predictions to determine whether the climate conditions for a malaria epidemic will occur. This forecasting would not be carried out at local levels, but at a higher level, to alert regional and national health authorities of the onset of abnormal weather conditions.

Early-warning systems at local level

With an alert of abnormal weather conditions ahead, health authorities at national level should begin to assess the situation locally. Rainfall, temperature and humidity should be monitored and analysed to determine whether abnormalities predicted at a regional level are occurring in the country itself. Unusual climatic events, such as heavier than average rainfall, prolonged drought or higher than average temperatures, could alert authorities to an impending epidemic. These warning systems could give health authorities 1–2 months’ prior notice of an epidemic.

Early detection of an epidemic

An early warning of an epidemic should also trigger more intense collection of data at health centres. Reports should confirm the occurrence of an epidemic past the normal levels. Having had prior alerts at regional and then national levels through forecasting and early-warning systems, the authorities should be able to put response measures in place.

What does forecasting mean for programming?

In the future, forecasting will mean more and better preparation for health programmes. A long-range forecast will alert regional coordinators to the possibility of epidemic conditions in their areas of work. This should filter down to health coordinators on the ground, so that data at health centres is monitored more frequently. Programmes that collect data on a monthly basis should step this up to weekly data collection, ensuring that health facility staff are using standard case definitions correctly.

Somewhere along the line, health authorities and health agencies will have to seriously consider obtaining and using locally available agricultural data in a formal way. Before long-range forecasts are in place, these will be the next best thing to determining the onset of epidemics.

Some tools already exist. The Mapping Malaria Risk in Africa (MARA) initiative uses epidemiological, entomological and weather data to develop maps which outline areas with different malaria situations throughout Africa. While MARA has provided maps down to country level, these are still considered a crude overview of the position. Nonetheless, these maps are valuable in determining whether health programmes are taking place in areas of high or low transmission. There are also WHO guidelines on epidemic thresholds for malaria. These are relatively easy to follow, but they require at least five years’ worth of malaria data from health centres.

Forecasting problems with forecasting

One of the major problems in detecting epidemics is the collection of quality data which can be analysed to determine the past and present malaria situation. Very few countries in Africa have good surveillance systems, where case definitions are standardised and reporting is completed and entered on time for analysis. In these cases, malaria epidemics are detected one or two months after they have begun, or not detected at all. Epidemic forecasting encompasses good surveillance at health centre level, and as a result implies a need for drastic improvements in surveillance systems.

Another difficulty in declaring an epidemic is the additional burden this imposes on a health system. Health authorities have in the past been reluctant to declare epidemics. Providing free health care to the population of an epidemic area for the duration of an epidemic is expensive. In countries where resources are scarce and cost-recovery provides a substantial portion of health centre revenues, providing free health care creates major problems for health ministries. Alerts can be provided to health authorities, but unless formal epidemic response systems have been put in place, the cascade of activities which are to follow alerts will not occur.

A final question is one which will inevitably be addressed to donors: with the possibility of a malaria epidemic, will donors invest in stockpiles of antimalarials, spray equipment, insecticide and additional health and logistical personnel? How accurate will systems have to be before donors invest in predictions?

Caroline Lynchspent five years working on malaria control in complex emergencies in various countries, including Burundi, the Democratic Republic of Congo, East Timor, Liberia and Somalia. She is currently a doctoral student at the London School of Hygiene and Tropical Medicine (LSHTM), looking at malaria epidemiology in highland zones, and a part-time consultant working mainly with the Malaria Consortium. Her email address is: mailto:Caroline.Lynch@lshtm.ac.uk

References and further reading

J. S. Cox, M. H. Craig, D. Le Sueur and B. L. Sharp, Mapping Malaria Risk in the Highlands of Africa (Durban: MARA/HIMAL, 1999), http://www.who.int/malaria/cmc_upload/0/000/012/178/cox1.htm.

Malaria Epidemics: Forecasting, Prevention, Early Detection and Control (Leysin: WHO, 2003), http://www.sahims.net.
The website of the MARA malaria mapping initiative: http://www.mara.org.za.

Field Guide for Malaria Epidemic Assessment and Reporting. Draft for Field Testing (Geneva: WHO, 2004), http://mosquito.who.int/cmc_upload/0/000/016/569/FTest.pdf.

Malaria Early Warning Systems: Concepts, Indicators and Partners: A Framework for Field Research in Africa (Geneva: WHO, 2001), http://mosquito.who.int/cmc_upload/0/000/014/807/mews2.pdf.