Scientific Reports (Oct 2021)

Spatio-temporal analysis and prediction of malaria cases using remote sensing meteorological data in Diébougou health district, Burkina Faso, 2016–2017

  • Cédric S. Bationo,
  • Jean Gaudart,
  • Sokhna Dieng,
  • Mady Cissoko,
  • Paul Taconet,
  • Boukary Ouedraogo,
  • Anthony Somé,
  • Issaka Zongo,
  • Dieudonné D. Soma,
  • Gauthier Tougri,
  • Roch K. Dabiré,
  • Alphonsine Koffi,
  • Cédric Pennetier,
  • Nicolas Moiroux

DOI
https://doi.org/10.1038/s41598-021-99457-9
Journal volume & issue
Vol. 11, no. 1
pp. 1 – 12

Abstract

Read online

Abstract Malaria control and prevention programs are more efficient and cost-effective when they target hotspots or select the best periods of year to implement interventions. This study aimed to identify the spatial distribution of malaria hotspots at the village level in Diébougou health district, Burkina Faso, and to model the temporal dynamics of malaria cases as a function of meteorological conditions and of the distance between villages and health centres (HCs). Case data for 27 villages were collected in 13 HCs. Meteorological data were obtained through remote sensing. Two synthetic meteorological indicators (SMIs) were created to summarize meteorological variables. Spatial hotspots were detected using the Kulldorf scanning method. A General Additive Model was used to determine the time lag between cases and SMIs and to evaluate the effect of SMIs and distance to HC on the temporal evolution of malaria cases. The multivariate model was fitted with data from the epidemic year to predict the number of cases in the following outbreak. Overall, the incidence rate in the area was 429.13 cases per 1000 person-year with important spatial and temporal heterogeneities. Four spatial hotspots, involving 7 of the 27 villages, were detected, for an incidence rate of 854.02 cases per 1000 person-year. The hotspot with the highest risk (relative risk = 4.06) consisted of a single village, with an incidence rate of 1750.75 cases per 1000 person-years. The multivariate analysis found greater variability in incidence between HCs than between villages linked to the same HC. The time lag that generated the better predictions of cases was 9 weeks for SMI1 (positively correlated with precipitation variables) and 16 weeks for SMI2 (positively correlated with temperature variables. The prediction followed the overall pattern of the time series of reported cases and predicted the onset of the following outbreak with a precision of less than 3 weeks. This analysis of malaria cases in Diébougou health district, Burkina Faso, provides a powerful prospective method for identifying and predicting high-risk areas and high-transmission periods that could be targeted in future malaria control and prevention campaigns.