A group of economists working with data collected in Kenya's northern arid districts have developed a model to predict severe child malnutrition - an indicator of famine - at least three months in advance. "Our forecasts are likely to be correct more than 75 percent of the time," said Andrew Mude, the leading author of the study.

"The [government-run] Arid Lands Resource Management Project (ALRMP) has been collecting information every month since 1996 from households, such as herd sizes, lactation rates, mortality rates and child nutrition data - specifically mid-upper arm circumference (MUAC)- which we feel is an accurate determinant of the nutritional status of a child," he said.

Mude is with the Nairobi-based International Livestock Research Institute, and one of the five researchers involved in the study, Empirical Forecasting of Slow-Onset Disasters for Improved Emergency Response: an Application to Kenya's Arid North.

The researchers’ definition of a famine is a situation in which more than 20 percent of the children in the area being studied are found to be severely malnourished or wasted.

The participating households were spread over 54 primarily pastoralist communities across four districts in the north: Baringo, Marsabit, Samburu and Turkana. The data also took into account food aid flows.

Africa has a host of early warning systems in place: the USAID-funded Famine early Warning Systems Network (FEWS-NET) provides forecasts for 20 African countries, and the Livestock Information Network and Knowledge system (LINKS) project informs pastoralists in East Africa of impending droughts.

Mude asserts that the new model is more "rigorous, cost-effective, practical and accurate" than existing early warning systems. "The required data can easily be collected by members of the community, and the empirical structure of the model allows the calibration of its accuracy. The more information the model is fed, the more accurate it will become."

Link to response

Early warning experts have been cautious in their response to the new model. "The key issue for early warning is that it needs to link to response. Although this data has been collected for years, the author acknowledges that it is not always used to guide response in the most effective manner, so this needs to be investigated as to why," said Grainne Moloney, Nutrition Project Manager for the UN Food and Agriculture Organisation's Food Security Analysis Unit in Somalia.

"Producing information with this new approach without understanding the previous obstacles to improved response will only solve half the problem," she said, while describing the model as a "promising approach".

Moloney suggested that it be field-tested at a national level in other contexts with similar structures, such as Ethiopia or Zimbabwe, with the focus on linking this information to action.

She also raised concerns about using mid-upper arm circumference and the ALRMP data. "The use of MUAC for routine nutrition surveillance is still quite a new approach, and the Arid Lands project is quite unique, with other countries using the more conventional weight-for-age indicators."

Moloney noted that while MUAC was an "easier, faster and cheaper tool, and has been shown to be a sensitive predictor of mortality in children ... its role in routine nutrition surveillance has not yet been quantified at global level, and research is ongoing."

She also pointed out that the collection of data by the ALRMP has been flawed.

Mude said the issue was that the ALRMP data had not been well stored and so much of it has been lost. “But management information systems have recently been put in place to make data collection efforts more useful”.