Summary: We find that the spatial scale/aggregation of an analysis influences the apparent importance of ecological drivers of arboviral (Ross River virus) disease; we urge future epidemiological studies to include multiple spatial scales for a more complete picture of disease drivers.

Background: Zoonotic vector-borne disease prevalence is affected by vector, human and reservoir host factors, which are influenced by habitat and climate; these five components interact on microhabitat to landscape scales but are often analyzed at a single spatial scale.

Methods: We present an information theoretic, multi-scale, multiple regression analysis of the ecological drivers of Ross River virus. We analyze the spatial pattern of 20 years of Ross River virus infections from South Australia (1992-2012; n = 5,261) using variables across these five components of disease ecology at three spatial scales.

Results: We found that covariate importance depended on the spatial scale of the analysis; some biotic variables were more important at fine scales and some abiotic variables were more important at coarser spatial scales. The urban score of an area was most predictive of infections and mosquito variables did not improve the explanatory power of these models.

Conclusions: Through this multi-scale analysis, we identified novel drivers of the spatial distribution of disease and recommend public health interventions. Our results underline that single-scale analyses may paint an incomplete picture of disease drivers, potentially creating a major flaw in epidemiological analyses. Multi-scale, ecological analyses are needed to better understand infectious disease transmission.