The geographic-level aspects involve a range of dimensions, ranging from physical characteristics of the area – such as location and climate (Bloom and Sachs 1998, APUD Wagstaff, 2001), to the infrastructure offered (Macintyre et. al., 2002) such as health services (quantity and quality), sanitation, water supply, roads, and so forth. One interesting hypothesis is that the presence of favorable aspects, like low crime rates, street cleanness and lightening, recreation places, among other “amenities” in the region under analysis mitigate the effects of unfavorable individual circumstances over health; whereas the adverse ones, such as pollution, lack of sanitation or low accessibility to urban facilities, amplify the already perverse influence of deprived individual characteristics on health status (Macintyre et. al., 2002; Kennedy et al., 1998). In California, for instance, Haan, Kaplan and Camacho (1987) found that people from poor areas experienced higher mortality rates (after controlling proper age, race and sex) than the population from non-poverty areas. Such risk of death persisted even after socioeconomic and behavioral adjustments. Similar findings have been reported by Humphries and Carr-Hill (1991); Jones and Duncan (1995) and Duncan et.al. (1993), supporting the hypothesis of the social environment´s influence over health, independently of the individual-level.
A large amount of empirical evidences has reported significant geographical variations in a wide range of health outcomes, as for instance, in coronary disease mortality (Diez Roux et.al,1997); morbidity (Jones and Ducan,1995); depression (Yen and Kaplan, 1999); and behavior patterns, such as alcohol consumption (Ecob and Macintyre, 2000 APUD Macintyre et. al., 2002), diet and physical activities (Cubbin et. al., 2006; Karvonen and Rimpela, 1996, APUD Ellaway and Macyntire, 2009) and self reported health (Jones and Duncan, 1995; Humphreys and Hill, 1991).
There is an intense debate on international literature about the adequate spatial scale for health analysis. It is often argued that contextual associations with health are most investigated using data referent to small areas (such as neighborhoods and communities); mainly in virtue of data availability (Diez Roux, 2008) and homogeneity of socioeconomic information (Curtis and Jones, 1998). As a result, there is an ‘easily -noted’ prevalence of administrative areas and neighborhoods, to investigate health outcomes through multilevel models. However, arguments in favor of analyzing larger areas also exist. As pointed out by Diez Roux (2008), neighborhoods “may not be the most relevant contexts for many health outcomes”, i.e., there is also the necessity to research other “policy-relevant unities”.
Following this reasoning, Curtis and Jones (1998) argue that effects of environmental factors, such as water and air quality, climate and degrees of urbanization, may not be well identified utilizing small area information,...