Background Effective targeting of malaria control in low transmission areas requires identification of transmission foci or hotspots. We investigated the use of functional data analysis to identify and describe spatio-temporal pattern of malaria incidence in an area with seasonal transmission in west-central Senegal.
Method Malaria surveillance was maintained over 5 years from 2008 to 2012 at health facilities serving a population of 500,000 in 575 villages in two health districts in Senegal. Smooth functions were fitted from the time series of malaria incidence for each village, using cubic B-spline basis functions. The resulting smooth functions for each village were classified using hierarchical clustering (Ward’s method), using several different dissimilarity measures. The optimal number of clusters was then determined based on four cluster validity indices, to determine the main types of distinct temporal pattern of malaria incidence. Epidemiological indicators characterizing the resulting malaria incidence pattern in terms of the timing of seasonal outbreaks, were calculated based on the slope (velocity) and rate of change of the slope (acceleration) of the incidence over time.
Results Three distinct patterns of malaria incidence were identified. A pattern characterized by high incidence, in 12/575 (2%) villages, with average incidence of 114 cases/1000 person-years over the 5 year study period; a pattern with intermediate incidence in 97 villages (17%), with average incidence of 13 cases/1000 person-years; and a pattern with low incidence in 466 (81%) villages, with average incidence 2.6 cases/1000 person-years. Epidemiological indicators characterizing the fluctuations in malaria incidence showed that seasonal outbreaks started later, and ended earlier, in the low incidence pattern.
Conclusion Functional data analysis can be used to classify communities based on time series of malaria incidence, and to identify high incidence communities. Indicators can be derived from the fitted functions which characterize the timing of outbreaks. These tools may help to better target control measures.
