At what geographic scale should dengue surveillance and control activities be carried out? Risk factors, including measures of vector densities, can predict risk differently at different geographic scales. Geographic scale is especially important because of the modifiable areal unit problem (MAUP). MAUP refers to variation in results when data are combined into sets of increasingly larger areal units or alternative combinations of base units at equal or similar scales (Openshaw and Taylor, 1979). Both phenomena are common problems for dengue surveillance and control programs because data are most commonly reported for areal units defined by political rather than epidemiological boundaries. Historically, most Ae. aegypti ecologists have characterized temporal, rather than spatial, patterns in mosquito abundance (Sheppard et al., 1969; Gould et al., 1970; Yasuno and Pant, 1970). Recent studies utilized a myriad of spatial analytical tools, including point pattern analysis (Gatrell et al., 1996; Getis, 1999). The utility of these analytical tools are two-fold. First, they characterize spatial autocorrelation patterns in variables of interest. Using a practical example, we can ask if vector densities in households are more highly correlated with those in neighboring houses than houses farther away. Autocorrelation can be measured at different distances and the scale at which autocorrelation is no longer significant would represent the minimum geographic unit for which surveillance and control schemes should be applied. Recent studies demonstrate that entomological risk should be measured at a household scale (Getis et al., 2003; Morrison et al., 2004a), but the distribution of infested houses does not follow a normal distribution (Alexander et al., 2006). Consequently, sample sizes need to be high for prospective epidemiological studies and evaluation of vector interventions. Second, spatial analyses can reveal underlying patterns in different variables. For example, one can ask whether clustering patterns of dengue cases are primarily due to natural variation in Ae. aegypti population densities at households or whether clusters are merely the result of some a priori heterogeneity in the region where the study was conducted (Gatrell et al., 1996). In this way, specific foci of transmission can potentially be identified or evaluated in relation to proximity to specific features of interest, such as village meeting places, schools, or markets. In the case of dengue, not enough is known about the role of human movement in defining the geographic scale of transmission. Although there is clear evidence of clustering of dengue cases within households (Morrison et al., 1998), how human movement patterns affect the scale of dengue transmission remains a major knowledge gap. Defining the appropriate geographic scale for measuring entomological risk and DV transmission, which will not necessarily be the same, will be an important new contribution to dengue surveillance and control (Getis et al., 2003).
After the capacity to account for inherent variation in dengue risk has been improved, it will be necessary to use that information to mitigate public health