Abstract
BACKGROUND: Epidemics of meningococcal disease in Africa are commonly detected too late to prevent many cases. We assessed weekly meningitis incidence as a tool to detect epidemics in time to implement mass vaccination. METHODS: Meningitis incidence for 41 subdistricts in Mali was determined from cases recorded in health centres (1989-98) and from surveillance data (1996-98). For incidence thresholds of 5 to 20 cases per 100000 inhabitants per week, we calculated sensitivity and specificity for detecting epidemics, and determined the time lapse between threshold and epidemic peak. FINDINGS: We recorded 9084 meningitis cases. Clinic-based weekly incidence of 5 and 10 cases per 100000 inhabitants detected all meningitis epidemics (sensitivity 100%, 95% CI 93-100), with median threshold-to-peak time of 5 and 3 weeks. Under-reporting reduced sensitivity: only surveillance thresholds of 5 or 7 cases per 100000 inhabitants per week detected all epidemics. Crossing the lower threshold before the 10th calendar week doubled epidemic risk relative to crossing it later (relative risk 2.1, 95% CI 1.4-3.2). At 10 cases per 100000 inhabitants per week, specificity for outbreak prediction was 88%, 95% CI 83-91). For populations under 30000, 3 to 5 cases in one or two weeks predicted epidemics with 85% to 97% specificity. INTERPRETATION: Low meningitis thresholds improve timely detection of epidemics. Ten cases per 100000 inhabitants per week in one area confirm epidemic activity in a region, with few false alarms. An alert threshold of 5 cases per 100000 inhabitants per week allows time to investigate, prepare for an epidemic, and initiate mass vaccination where appropriate. For populations under 30000, the alert threshold is two cases in a week. High quality surveillance is essential.
