Abstract
Background: Visceral leishmaniasis (VL) is one of the Neglected Tropical Diseases targeted for ‘elimination as a public health problem’ on the Indian Subcontinent by the WHO (<1 case per 10,000 by or before 2020). Due to the implemented intervention strategies, VL incidence rates have been decreasing at subcontinent, country and
... read more
regional levels. However, VL is a very focal disease with spatiotemporal fluctuations in annual incidence rates on smaller scales. This study aims to investigate spatiotemporal autocorrelation (randomness vs. clustered or dispersed patterns) annual VL incidence rates and underlying factors explaining these fluctuations on small scale in Bihar, India. Methods: Longitudinal geocoded data on annual VL incidence rates from 276 hamlets (subunit of a village) between 2007 and 2015 in a highly endemic district of Bihar were available. Incidence rates were mapped with QGIS for each year to visualize spatiotemporal fluctuations. Inverse Distance Weighting (IDW) was applied to interpolate and predict VL incidence rates across the district. Moran’s I Index statistics was used to determine spatial autocorrelation in interpolated incidence data for each year. Both spatiotemporal and spatial clusters of high incidence rates were detected with SaTScan. Temporal autocorrelation between the different years was determined using a Poisson regression analysis. Age and socioeconomic status (measured by asset index), two factors known to influence VL dynamics, were available for the population enrolled in the study in 2007 and 2008. A Chi2 test was used to discover if age and asset index distributions of population inside of spatial clusters or ‘hotspots’, significantly differed from outside of hotspots.
Results: Annual incidence rates per 10,000 capita at sub-district level in Muzaffarpur drop from 12.3 cases in 2007 to 0.9 cases in 2015, with a small peak in 2012. Mapping clearly illustrates the spatial heterogeneity in VL incidence rates among hamlets. IDW interpolated maps showed multiple predicted risk areas for VL, varying per year and location. Moran’s I Index shows that VL cases are spatially clustered in 2007–2012. Five significant spatial and two significant spatiotemporal clusters were detected with SaTScan. Temporal clustering was evident, hotspots remained for approximately 2–6 years. Results showed a shift towards lower socioeconomic status within hotspots. We found no significant difference in age distribution between hotspots and the rest of the study area. Conclusions: The results reveal clear spatiotemporal, spatial and temporal clustering of annual VL incidence rates at hamlet level in Muzaffarpur between 2007 and 2015. Hotspots can partially be explained by lower socioeconomic status. The gained insights into spatiotemporal patterns of VL can help defining future risk areas of the disease and guide decision-making as where to implement additional control strategies.
show less