able to provide richer information because it also examined whether
the greenness–physical activity association was of different magnitude
for residents living or not in the smart growth community
(interaction of effects). Such an interaction in this enhanced
momentary analysis allowed the authors to assess whether the
overall community design (smart growth or not) influenced the
extent to which spending time in greenspaces was associated with
increased physical activity levels. The estimated interaction revealed
that the greenness–physical activity association was slightly stronger
in the smart growth community, even if the confidence intervals
were overlapping.
Finally, the fourth reviewed study (Zenk et al., 2011) (Detroit,
2008–2009) analyzed GPS data for 120 adults and older adults
(30 s intervals) and accelerometer data for 97 of these participants
(1 min epochs). Zenk and colleagues primarily relied on GPS
to improve the assessment of environmental exposures, i.e., in the
perspective of a contextual expology as denominated in a recent
publication of ours (Chaix et al., 2012c). The notion of expology is
sometimes used in environmental epidemiology and toxicology to
refer to the characterization of individual risks of exposure; in our
view, ‘‘contextual expology’’ is a subdiscipline of neighborhood
and health research interested in the multiple places and times of
exposure to contexts. The objective of Zenk et al. was to overcome
the mischaracterization of environmental exposures of studies
focused on residential neighborhoods. The authors’ aim was not
to assess the actual contexts of behavior but the (potential)
spatial accessibility to services such as fast-food outlets, supermarkets,
or parks that may influence health behavior or health. In
their analyses conducted at the individual level, exposures were
assessed in ‘‘daily path areas’’ derived by buffering all GPS points
at 0.5 miles. Interestingly, a comparison of measures of exposures
concluded that exposures in the residential neighborhood correlated
weakly with exposures in the GPS-based activity space.
Fast-food outlet density measured in the residential neighborhood
was associated with none of the three dietary intake outcomes
(intake of saturated fat, fruits and vegetables, and whole
grains). However, fast-food outlet density in the daily path area
(around each GPS point) was positively related to saturated fat
intake and negatively associated with whole grain intake. The
comment of the authors that ‘‘the daily path area may better
capture fast-food outlets that were actually utilized’’ suggests
that such an exposure measure reflects the actual behavioral
contexts rather than potential access as needed for causal inference
on environmental effects.
The study by Zenk is the only reviewed article to discuss this
issue of bias, indicating that ‘‘activity space fast-food outlet
density may be associated with dietary behavior because individuals
who want to consume fast-food seek out environments
with higher fast-food outlet concentration in order to obtain it’’
(p. 1158). Arguing that exposure measures that reflect actual
behavior generate bias, the authors interestingly suggested that
future research should investigate whether the actual use of
resources mediates relationships between the potential access
to resources around daily activity locations (accessibility) and
weight-related behaviors.