AUTHOR=Zeimes Caroline Brigitte , Quoilin Sophie , Henttonen Heikki , Lyytikäinen Outi , Vapalahti Olli , Reynes Jean-Marc , Reusken Chantal , Swart Arno N. , Vainio Kirsti , Hjertqvist Marika , Vanwambeke Sophie O.
TITLE=Landscape and Regional Environmental Analysis of the Spatial Distribution of Hantavirus Human Cases in Europe
JOURNAL=Frontiers in Public Health
VOLUME=3
YEAR=2015
URL=https://www.frontiersin.org/journals/public-health/articles/10.3389/fpubh.2015.00054
DOI=10.3389/fpubh.2015.00054
ISSN=2296-2565
ABSTRACT=Background: In Europe, the most prevalent hantavirus, Puumala virus, is transmitted by bank voles and causes nephropathia epidemica in human. The European spatial distribution of nephropathia epidemica is investigated here for the first time with a rich set of environmental variables.
Methods: The influence of variables at the landscape and regional level is studied through multilevel logistic regression, and further information on their effects across the different European ecoregions is obtained by comparing an overall niche model (boosted regression trees) with regressions by ecoregion.
Results: The presence of nephropathia epidemica is likely in populated regions with well-connected forests, more intense vegetation activity, low soil water content, mild summers, and cold winters. In these regions, landscapes with a higher proportion of built-up areas in forest ecotones and lower minimum temperature in winter are expected to be more at risk. Climate and forest connectivity have a stronger effect at the regional level. If variables are staying at their current values, the models predict that nephropathia epidemica may know intensification but should not spread (although southern Sweden, the Norwegian coast, and the Netherlands should be kept under watch).
Conclusion: Models indicate that large-scale modeling can lead to a very high predictive power. At large scale, the effect of one variable on disease may follow three response scenarios: the effect may be the same across the entire study area, the effect can change according to the variable value, and the effect can change depending on local specificities. Each of these scenarios impacts large-scale modeling differently.