Meningitis seasonal pattern in Africa and detection of epidemics: a retrospective study in Niger, 1990-98.

Control of epidemic meningitis is still an unresolved problem in Africa. WHO has promoted the use of surveillance and response following alerts based on weekly threshold levels. In order to avoid any waste of resources related to false-positive alerts, it was decided not to choose too sensitive thresholds. This policy, however, leads to delayed response. The seasonal pattern of epidemics provides a solution to this dilemma. We carried out a retrospective survey of district-level surveillance data in Niger from June 1990 to June 1998. We identified an early and late meningitis season. Following this pattern, we studied the performance of the WHO-recommended threshold as compared to alternative thresholds for identifying early, late and non-epidemic district-years (DYs). (ADY was defined as a 52-week period starting in the last week of June, at the district level). We studied 296 DYs, comprising 50 early epidemic, 38 late epidemic, and 208 non-epidemic DYs. Early epidemics were more often large and accounted for almost 75% of total cases. When applied no later than the first week of March, a highly sensitive alternative threshold resulted in initiation of an alert, with a median of 3 weeks earlier than the standard threshold, with no false-positive alerts, i.e., a specificity of 1.