Abstract
Despite improved surveillance capacities and WHO recommendations for subdistrict analysis, routine epidemic surveillance of acute bacterial meningitis in the African meningitis belt remains largely limited to the district level. We evaluated the appropriateness and performance of analyses at higher spatial resolution. We used suspected meningitis surveillance data at health centre (HC) resolution from Burkina Faso from 14 health districts spanning years 2004-2014 and analysed them using spatio-temporal statistics and generative models. An operational analysis compared epidemic signals at district and HC-level using weekly incidence thresholds. Eighty-four percent (N = 98/116) of epidemic clusters spanned only one HC-week. Spatial propagation of epidemic clusters was mostly limited to 10-30 km. During the 2004-2009 (with serogroup A meningitis) and 2010-2014 (after serogroup A elimination) period, using weekly HC-level incidence thresholds of 100 and 50 per 100,000 respectively, we found a gain in epidemic detection and timeliness in 9 (41% of total) and 10 (67%), respectively, district years with at least one HC signal. Individual meningitis epidemics expanded little in space, suggesting that a health centre level analysis is most appropriate for epidemic surveillance. Epidemic surveillance could gain in precision and timeliness by higher spatial resolution. The optimal threshold should be defined depending on the current background incidence of bacterial meningitis.