Using Relative Humidity Forecasts to Manage Meningitis in the Sahel

Meningitis epidemics in the Sahel occur quasi-regularly and with devastating impact. In 2008, for example, eighty-eight thousand people contracted meningitis and over five thousand died. Until very recently, the protection provided by the only available vaccine was so limited and short-lived that the only practical strategy for vaccination was reactive: waiting until an epidemic occurred in the region and then vaccinating in that region to prevent the epidemic's further growth. Even with that strategy, there were still times when demand outpaced available vaccine. While a new vaccine has recently been developed that is effective and inexpensive enough to be used more broadly and proactively, it is only effective against the strain of bacteria that causes the most common kind of bacterial meningitis. As a result, there will likely be continued need for reactive vaccination strategies. It is widely known that meningitis epidemics in the Sahel occur only in the dry season. Our project investigated this relationship, and several independent lines of evidence demonstrate a robust relationship between the onset of the rainy season, as marked by weekly average relative humidity above 40%, and the end of meningitis epidemics. These lines of evidence include statistical analysis of two years of weekly meningitis and weather data across the Sahel, cross-correlation of ten years of meningitis and weather data in the Upper East region of northern Ghana, and high-resolution weather simulations of past meningitis seasons to interpolate available weather data. We also adapted two techniques that have been successfully used in public health studies: generalized additive models, which have been used to relate air quality and health, and a linearized version of the compartmental epidemics model that has been used to understand MRSA. Based on these multiple lines of evidence, average weekly relative humidity forecast two weeks in advance appears consistently and strongly related to the number cases of meningitis in the Sahel. Using currently available forecast models contributed through the WMO Thorpex-Tigge project, and applying quantile regression to enhance their accuracy, we can forecast the average weekly relative humidity to two weeks in advance which allows us to anticipate the end of an epidemic in a region of the Sahel up to four weeks in advance. This would allow public health officials to deploy vaccines to areas in which the epidemics are likely to persist due to continued dryness and avoid vaccinating in areas where the epidemics will end with higher humidity. Our presentation will conclude by introducing the relative humidity decision-information tool developed for use by public-health officials. We will also summarize the results of a weekly meningitis forecast exercise held during the 2011-2012 dry season with public health decision makers from several African countries and the World Health Organization. Finally, we highlight some results of concurrent socio-economic research that suggests other interventions for managing meningitis and helps quantify the economic impact of the disease in Ghana. Overall, while our research has demonstrated an actionable relationship between weather and disease, this relationship is only one factor in a complex and coupled human-natural system which merits continued investigation.