Spatio-temporal variation of malaria hotspots in Central Senegal, 2008-2012

Abstract

Background: In malaria endemic areas, identifying spatio-temporal hotspots is becoming an important element of innovative control strategies targeting transmission bottlenecks. The aim of this work was to describe the spatio-temporal variation of malaria hotspots in central Senegal and to identify the meteorological, environmental, and preventive factors that influence this variation.

Methods: This study analysed the weekly incidence of malaria cases recorded from 2008 to 2012 in 575 villages of central Senegal (total population approximately 500,000) as part of a trial of seasonal malaria chemoprevention (SMC). Data on weekly rainfall and annual vegetation types were obtained for each village through remote sensing. The time series of weekly malaria incidence for the entire study area was divided into periods of high and low transmission using change-point analysis. Malaria hotspots were detected during each transmission period with the SaTScan method. The effects of rainfall, vegetation type, and SMC intervention on the spatio-temporal variation of malaria hotspots were assessed using a General Additive Mixed Model.

Results: The malaria incidence for the entire area varied between 0 and 115.34 cases/100,000 person weeks during the study period. During high transmission periods, the cumulative malaria incidence rate varied between 7.53 and 38.1 cases/100,000 person-weeks, and the number of hotspot villages varied between 62 and 147. During low transmission periods, the cumulative malaria incidence rate varied between 0.83 and 2.73 cases/100,000 person-weeks, and the number of hotspot villages varied between 10 and 43. Villages with SMC were less likely to be hotspots (OR = 0.48, IC95%: 0.33-0.68). The association between rainfall and hotspot status was non-linear and depended on both vegetation type and amount of rainfall. The association between village location in the study area and hotspot status was also shown.

Conclusion: In our study, malaria hotspots varied over space and time according to a combination of meteorological, environmental, and preventive factors. By taking into consideration the environmental and meteorological characteristics common to all hotspots, monitoring of these factors could lead targeted public health interventions at the local level. Moreover, spatial hotspots and foci of malaria persisting during LTPs need to be further addressed.

Trial registration: The data used in this work were obtained from a clinical trial registered on July 10, 2008 at www.clinicaltrials.gov under NCT00712374.

Keywords: Geostatistical analyses; Malaria hotspots; Non-linear associations; Spatial clusters; Spatio-temporal dynamic.

Fig. 1

Evolution of weekly malaria incidence and rainfall through transmission periods. Malaria incidence (continuous red curve); High Transmission Periods (HTP, in grey) and Low Transmission Periods (LTP, in white) with their duration (weeks, in black) and cumulative incidence rates (red numbers); total weekly rainfall (in blue)

Fig. 2

Spatio-temporal distribution of hotspot villages and vegetation type during transmission periods from 2008 to 2012. Hotspot villages (red dots) and non-hotspot villages (black dots); vegetation type (land cover: open shrublands in beige, grasslands in orange, croplands in yellow, and mixed vegetation in green) according to transmission periods (LTP, HTP) from 2008 to 2012 in Bambey and Fatick health districts, Senegal. Acknowledgements:the Senegal shapefile were downloaded in gadm website and Landcover rasters extracted from MODIS NASA website.

Fig. 3

Temporal evolution of the risk of being hotspot with 95%CI according to vegetation type. Temporal evolution of the risk of being a hotspot (continuous black curve) with 95% confidence interval (discontinuous black curves) according to each vegetation type: open shrublands (panel a), grasslands (panel b), croplands (panel c), and mixed vegetation (panel d). HTPs and LTPs are indicated in grey and white, respectively. The vertical red lines indicate the dates of interest, at which villages were at risk of being hotspot. The horizontal red lines indicate the zero reference line

Fig. 4

Evolution of the risk of being hotspot with 95%CI according to rainfall and vegetation type. Evolution of the risk of being a hotspot (continuous black curve) with 95% confidence interval (discontinuous black curves) according to weekly rainfall and vegetation type: open shrublands (panel a), grasslands (panel b), croplands (panel c), and mixed vegetation (panel d). Vertical red lines show the amount of rainfall starting from which rainfall became a risk factor

Fig. 5

Spatial distribution of the different hotspot types and the associated spatial risk of being hotspot. The black curves are the contours of bivariate smoothed values; the colour bar is the ascending level of risk indicated by the spline smoothing function values (smoothed values) from blue to red; red dots represent the villages that were a hotspot during all 5 LTPs (Hot5LTP), orange dots those that were a hotspot mainly during HTPs (MajoHotHTP), yellow dots those that were a hotspot mainly during LTPs (MajoHotLTP), blue dots those that were a hotspot equally during HTPs and LTPs (EquaHTPLTP), green dots those that were never a hotspot (NeverHot), brown dots those that were a hotspot only during HTPs (OnlyHotHTP), and black dots those that were a hotspot only during LTPs (OnlyHotLTP)