Using GPS and activity tracking to reveal the influence of adolescents' food environment exposure on junk food purchasing.
Sadler, Richard C. ; Clark, Andrew F. ; Wilk, Piotr 等
Some of the most critical public health issues facing Canadians (e.g., obesity, heart disease, stroke, hypertension and type 2 diabetes) are linked to poor nutrition. (1) Among other variables, poor dietary habits have contributed to adverse health outcomes among Canadian adolescents: nearly one third are overweight or obese. (2) Obesity and other diet-related health issues arise not only from individual-level factors (e.g., genetics, lifestyle) but also from the characteristics of our local environments that discourage healthy diets, such as the presence of unhealthy food outlets. (3-6) Neighbourhood food environments can have a particularly strong influence on children, including adolescents, who tend to be more restricted geographically than adults and who are therefore more captive to their local built environments, including food outlets, recreational spaces and transportation infrastructure. (7,8) Understanding the local food environment is therefore important to encouraging healthy lifestyles among adolescents. (9)
Despite increasing acknowledgements of the importance of the built environment for health, it remains poorly conceptualized in much public health work. Researchers at the forefront of the field have advocated the combination of GPS(global positioning system)-derived activity spaces with activity/food diaries to better link junk food exposure and purchasing behaviours. (10-12) In this paper, we evaluate the relationship between junk food exposure and purchasing behaviour among adolescents during the school day, while controlling for sex, mode of transportation, and direction of the trip between home and school.
Addressing bias in geospatial proxies
Sadler and Gilliland (10) showed how geospatial proxies rather than direct measurements continue to be used to evaluate exposure to junk food. Most proxies have included calculating the density of junk food outlets in a child's home or school neighbourhood, or both. (3-5,13,14) Such density estimates are typically calculated within a buffered walkable zone (typically between 400 and 1600 m and measured along the street network or "as the crow flies") around home and/or school, or within the boundaries of a more arbitrary administrative unit, such as the census tract or census dissemination area where the home or school is located. Each method is susceptible to the modifiable areal unit problem, because any observed association may change depending on which scale of areal unit is employed. (15) Williams et al. (16) have cautioned against the use of such metrics because "predominant exposure measures may not account for what individual children actually experience" (p. 359).
Because of the lack of certainty in measurement, the population-level modelling approach common to most studies has been critiqued for creating biases in classification. For instance, user-defined activity spaces have been characterized as subjective constructions of travel surveys and food store listings. (17) While the use of activity spaces overcomes the limitations of grounding exposure to one location, researchers have advocated for more detailed individual-level neighbourhood assessments. (10,14-16)
Objectively measured aspects of the built environment, such as GPS tracking of activity spaces, offers an advance in the level of certainty in approximating a child's food environment. GPS tracking is more accurate for quantifying activity spaces than estimations by parents or participant self-report. (18) In one case, Harrison et al. (19) showed that GPS trips accounted for 50% more food outlets on children's trips home from school when compared with assumed trips.
New work emphasizes the need to focus "on combining GIS [geographic information systems]-based objective measurement of the community food environment with self-report measures" (p. 13). (20) Others refer to this as a need for "ego-centred definitions of areas that approximate individuals' local activity spaces" (p. 227). (21) A recent observational study using self-report measures found that children who ride home from school in private automobiles eat more snacks and candy than those who walk. (22) As that study did not use GPS tracking to delineate and characterize the food environment through which children travelled, important questions remain about the role of exposure.
The objective of this study is to explore the nature of the relationship between junk food purchasing (JFP) and the level of exposure to junk food outlets (JFOs). We achieved this objective by assessing the magnitude of this relationship and considered three trip-level control variables: 1) mode of transportation (active, bus and car); 2) trip direction (to school and from school); and 3) child's biological sex (females and males).
METHODS
Data collection
Data were collected as part of the Spatial Temporal Environment and Activity Monitoring (STEAM) Project (steamproject.ca). The STEAM Project compiled demographic, behavioural and GPS tracking data on 932 adolescents aged 9-13 years from communities in southwestern Ontario. The central aim of STEAM is to explore and assess how the physical (built and natural) environment influences adolescents' activity patterns and food consumption habits. This study was conducted with approval from the University of Western Ontario's Non-Medical Research Ethics Board (REB#: 17918S).
Data were collected over four years (2010-2013), each child being observed for one week in the spring and one week in the fall. The current study uses data from 511 adolescents from 25 elementary schools in Middlesex County and the City of London (Middlesex-London), which are characterized by a broad range of built forms and social environments. Adolescents completed socio-demographic questionnaires and daily activity diaries each week, answering questions about physical activity, eating habits and social/familial engagements. Adolescents indicated for each day whether they had purchased something on their way to or from school and the location of that purchase.
GPS tracks for every child were collected every second between the child leaving for school and the child returning home. Data derived from these GPS tracks included the mode of transportation (e.g., walk, bike, bus, car), time of day, and a key field to link with child-specific individual-level characteristics.
Every JFO in the region was extracted from the Middlesex-London public health inspector's food outlet database and geocoded (using principles of accuracy as discussed in Healy and Gilliland (23)) in a GIS (ArcGIS 10.1, Environmental Systems Research Institute, Redlands, CA). Given average road widths and typical viewsheds within our community, 50 m buffers were calculated around every unhealthy food source to help measure exposure. (10,19) These buffers were combined in GIS with GPS data on trips to give a measure of the number of minutes each adolescent was exposed to junk food sites on each trip to and from school.
The final data used in this study combined the individual GPS trips and modelled exposure values for each trip with data from the activity diaries, which indicated any JFP along the trip to and/or from school. These combined data provide the ability to analyze the relationship between exposure and JFP on a trip-by-trip basis. The study uses the trips to and from school because adolescents indicate that they have the most autonomy during these times.24 This final set of trips for adolescents exposed to junk food were used for the final analysis, as a trip without exposure provided no opportunity to purchase junk food.
Analysis methods
The dependent variable, JFP, is a binary variable indicating whether junk food was purchased or not on the trip. It is calculated for each trip to and from school for each adolescent on the basis of the activity diary entries. Junk food was considered unhealthy food items purchased from fast food or variety stores, pizza places and ice cream shops.
The key independent variable, exposure to JFOs, was defined as the number of minutes during which a child was exposed (i.e., within 50 m) to fast food, variety stores, pizza places or ice cream shops (ranges from 0 sec to 350 min). This variable was truncated at 17 min (1020 sec) to account for significant outliers of exposure.
Three control variables were hypothesized to influence the relationship between exposure and food purchasing: biological sex, mode of transportation, and direction of the trip. Sex was defined as male and female. Mode of transportation was defined for each trip, including active modes (bike, walk, scooter or skateboard), car, and bus (school bus or city transit). The direction of the trip was either to or from school.
This study uses multilevel logistic regression, a commonly used technique with a binary outcome variable that takes into account clustering within the data. Individual trips are not independent but nested within adolescents; data on individual trips from a given adolescent are expected to be more alike than data collected from another adolescent. The study first assessed the overall effect of JFO exposure on JFP. The strength of this relationship was then examined separately for each category of one of the three control variables. Since this was an exploratory analysis, no multivariate models or models with interaction effects were tested. Both the linear and the quadratic effects were evaluated; none of the quadratic terms, however, were significant at the p = 0.05 level. To ease interpretability of the effects, the predicted probabilities of JFP were computed from across a range of exposure values (from "up to 1 min" to "up to 17 min") and plotted separately for each category of the three control variables. Predicted probabilities were computed from multilevel logistic regression models using both the intercepts and corresponding regression coefficients.
Child-level variances in JFP were also estimated by specifying the intercepts in all multilevel regression models as random. The amount of variation in JFP across adolescents was assessed by the intra-class correlation coefficient (ICC) and median odds ratio (MOR). The ICC was calculated by dividing the cluster-level variance by the total variance, representing the proportion of variance attributed to differences among adolescents. To calculate the ICC for the binary variable, the trip-level (level one) variance was fixed to the variance of the standard logistic distribution. (25) The MOR converts the cluster-level variance to an odds ratio scale and, as a consequence, it can be compared directly with odds ratios for fixed effects. (25) A large MOR implies large variation across adolescents, whereas a value of 1 suggests no such variation. The multilevel models are computed using Mplus. (26)
RESULTS
Descriptive statistics
In total, we observed 7,499 individual trips from 654 adolescents in Middlesex-London with both GPS and food purchasing data from recall diaries; 4,588 of these trips, from 511 children, involved exposure to junk food and were therefore retained for analysis. Of these 4,588 trips, 224 (4.9%) involved JFP. The average number of observed trips per child was 9, ranging from 1 to 20, and the prevalence of JFP among adolescents ranged from 0.0% to 100.0%. Additional characteristics included: more females (58.7%) than males (41.3%); more valid trips from (51.6%) than to (48.4%) school; and a modal split of 39.0% by bus, 30.8% by active modes and 30.2% by car.
Factors associated with junk food purchasing
Overall Effect
The results from the multilevel logistic regression analysis reported in Table 1 indicate that as exposure (measured in minutes) increased, the odds that junk food was purchased on that trip increased significantly. The OR for JFP associated with a 1-min increase in exposure to JFOs was 1.174 (95% CI [confidence interval] 1.14-1.21). The trip-level results from a multilevel model should be interpreted as ORs for within-cluster comparisons; they compare two trips observed in the same child. Figure 1a indicates that the probability of JFP increased from approximately 1.7% for trips with exposure of less than 1 min to 15.8% for trips with exposure of between 16 and 17 min. Finally, JFP across adolescents varied significantly, corresponding to an ICC of 0.499 and an MOR of 5.613. Thus, about 50% of the variance in the likelihood of JFP was due to differences between adolescents, and the MOR of 5.613 suggests a substantial difference between two trips with the same level of exposure but made by different, randomly chosen adolescents.
Mode of Transportation
Stratification of the focal relationship by the mode of transportation indicated that the relationship between exposure to JFOs and JFP was statistically significant and positive for data involving trips made by active modes of transportation and for trips made by car, but not significant for trips made by bus. The ORs for JFP associated with a 1-min increase in exposure for active and car travel were 1.13 (95% CI 1.06-1.20) and 1.22 (95% CI 1.16-1.28) respectively. Figure 1b indicates that the rate of increase in the predicted probabilities of JFP was much steeper for trips made by car than for trips made by active modes of transportation, increasing from 2.7 times more likely at 5 min to 4.4 times more likely at 15 min. Finally, the variation in JFP across adolescents was lowest for trips made by car (ICC = 0.459; MOR = 4.914) and the highest for active trips (ICC = 0.610; MOR = 8.705), the trips made by bus falling in the middle (ICC = 0.541; MOR = 6.546).
Trip Type
Stratification by route to or from school indicated that the relationship between JFO exposure and JFP was statistically significant and positive for data involving trips made to and from school. These effects translate into ORs of 1.22 (95% CI 1.12-1.33) and 1.12 (95% CI 1.08-1.16) respectively. Figure 1c shows that, although the effect of exposure was slightly more pronounced for trips to school, trips from school were much more likely to be associated with JFP than trips to school at all levels of exposure. Because the odds of JFP varied by the amount of time spent exposed on a trip to or from school, we cannot give one value for the magnitude of this difference. We can, however, estimate that the average ratio between the two was 4.19, ranging from 6.56 at 1 min to 1.81 at 16 min. In terms of the cross-cluster variance, the variation in JFP was more pronounced in trips to school (ICC = 0.513; MOR = 5.907) than trips from school (ICC = 0.445; MOR = 4.707).
Biological Sex
The results from the multilevel logistic regression models exploring the role of biological sex indicate that, for females, the OR for JFP associated with a 1-min increase in exposure was 1.19 (95% CI 1.15-1.24). For males, the corresponding OR was 1.12 (95% CI 1.06-1.19). Figure 1d shows that trips made by females had a higher likelihood of being linked to JFP at all levels of exposure than trips made by males, but the gap between the two widened with the increased level of exposure, from 2.5 times more likely at 5 min to 3.0 times more likely at 15 min. Finally, the variation in JFP was higher for trips made by males (ICC = 0.583; MOR= 7.734) than for trips made by females (ICC = 0.439; MOR = 4.617).
[FIGURE 1 OMITTED]
DISCUSSION
In this study of adolescents aged 9-13 years in Middlesex-London, ON, nearly 1 in 20 trips (4.9%) made to and from school involved the purchase of unhealthy junk food. Furthermore, a significant positive relationship existed between adolescents' duration of exposure to unhealthy food outlets (i.e., fast food restaurants and variety stores) between home and school and the likelihood of JFP. This finding corroborates a previous study of adolescents in the same geographic area, which suggested that the availability or density of JFOs in a child's home or school neighbourhood increases the likelihood of junk food purchasing. (3)
While previous research has established a relationship between the presence of JFOs and purchasing/consumption, (3,5,7) such findings are inconsistent. (27) This inconsistency may be a result of using areal unit measures as proxies for exposure, which are unable to directly connect the presence of JFOs to the actual routes that adolescents take to and from school. This study has advanced a novel method to connect the GPS-derived routes of adolescents' individual trips between home and school to their junk food exposure. This is particularly important because an individual's trips and activities rarely coincide with the arbitrary spatial boundaries used in previous research (e.g., buffers, census tracts or postal codes). (3-5,13,14) Our research, therefore significantly improves upon the accuracy of daily JFO exposure among children/adolescents en route to/from school. By accurately measuring exposure to JFOs, we are able to fully understand how a child's individual and trip characteristics may alter the relationship between exposure and purchases.
Although this study found a significant relationship between exposure and purchasing, the results show considerable unexplained variance due to differences among adolescents. Exploring these differences by stratifying the data by sex, direction of trip (to or from school), and mode of travel revealed that the relationship remained significant for all categories of stratification, with the exception of adolescents who travel by bus. This lack of a relationship is due to school district policy, which prohibits children from exiting a school bus along the route to/ from school before their designated stop. Any purchases would therefore be conducted only while traveling between a bus stop and home.
Conversely, not only was the relationship between exposure to JFOs and JFP positive and statistically significant for trips made by automobile, but the rate of increase in the predicted probabilities of JFP was also much steeper for trips made by automobile than for trips made by active modes (i.e., walking or biking). This finding is an important contribution to the academic literature and useful for the development of interventions, as it indicates that the influence of exposure on adolescents' JFP is actually greater when adolescents are driven to/from school under adult supervision than when they walk or bike. This may be a result of time-crunched parents buying food for their adolescents "on the go" and parents bending to the will of the child requesting junk food. (28) While exposure also significantly influences JFP among walkers, this should not discourage parents or public health promoters from advocating that adolescents walk to school, especially because active travel has well-known physical and mental health benefits and helps the child develop independence and a sense of environmental competence. (29)
Consistent with previous research on the built environment and adolescents' health-related behaviours, (29,30) this study revealed different patterns of behaviour along the journey to school in the morning versus the journey home from school in the afternoon. Although the effect of exposure was significant for trips in both directions, the odds of JFP by adolescents were much higher on the journey home from school versus to school, and the odds narrowed the longer a child travelled. This finding is likely related to the adolescent having more flexible time on the way home after school compared with the morning, when he or she has to reach school for a set time; it could also be due to daily eating patterns and adolescents just being hungrier after a long school day.
Analysis revealed that females were more strongly influenced by exposure to JFOs than males, being between 1.6 and 3.5 times more likely to make a JFP. Trips made by females had a higher likelihood of being linked to JFP at all levels of exposure than trips made by males. The gap in the likelihood of JFP between trips by females and males widened with the increased level of exposure. This finding is consistent with a previous study of adolescents in London, ON, which found that females were 1.5 times more likely than males to have self-purchased (without parents) fast food at least once per week. (3) It is unclear why the females in our study were more likely to purchase junk food and were more influenced by exposure. Females may have greater access to their own spending money, as it is common in Canadian culture for adolescent girls to start earning money earlier through babysitting. (31)
Limitations
Although this study offers a significant advance by using objective methods of observing adolescents' actual routes between home and school to assess direct JFO exposure, some limitations exist. Researchers have cautioned that even GPS data can have limitations, as they track only where the child has travelled over the course of data collection and may not encompass the totality of their potential exposure. (19) Chaix et al. (12) argue that biases related to selective daily mobility may prohibit accurate assessment of environmental effects. To limit the potential burden on research subjects, most studies using GPS tracking limit data collection to short periods (typically one week); it can be argued that one week of tracking spatial behaviours is not enough time to assess how potential environmental exposures may affect chronic diseases (e.g., obesity, type 2 diabetes, cancer). Nevertheless, it can be argued that GPS tracking can be an ideal tool for assessing how exposure to environmental features such as JFOs influences the likelihood of making a JFP (often an impulse activity).
CONCLUSION
This is one of the first studies to empirically establish a relationship between objective, GPS-derived measurement of direct JFO exposure and adolescents' JFP. While causal relations cannot be inferred and the data are not necessarily generalizable to other age groups or geographic settings, the study highlights important implications for municipal planners, school board officials and other decision-makers involved in the regulation, development and management of adolescents' environments. In particular, municipalities should embed specific bylaws and policies restricting the concentration of JFOs close to schools, as passed in London, UK. (32) School board officials should also consider potential JFO exposure when making decisions on the siting of new schools and the closing of existing neighbourhood schools, which typically results in longer average commutes for students and greater exposure to JFOs. Furthermore, public health agencies at all levels (i.e., municipal, provincial and federal) should work together, in concert with municipal economic development organizations (e.g., business improvement areas) and private sector stakeholders on the supply side of the food system (e.g., food producers, restaurant owner associations, retailers) to introduce effective economic incentives to encourage greater availability, visibility and knowledge of healthier food options in local food stores and restaurants.
In addition to highlighting the need to reduce junk food exposure in neighbourhood food environments, we also reiterate what other studies have shown about the importance of parents as role models for their adolescents when it comes to healthy eating. (33) The findings indicate that junk food exposure has the greatest impact on JFP when adolescents are being driven in a car (i.e., being accompanied by an adult). This finding points to the need for further education to improve food literacy regarding overconsumption of generally unhealthy fast food. Nevertheless, public health practitioners and researchers should not single out adolescents as inherently poor decision-makers; we cannot forget how common junk food consumption is across North American society. Like their adult counterparts, adolescents report eating junk food because of the convenience and taste. (33) Furthermore, "to give up eating what teens call 'junk food' would be to give up much more than the food itself. This speaks to the importance of changing social norms around healthful eating." (p. S42) (34) Thus, we need to continue to make it easier and more attractive to eat healthily; this is an area where food "apps" for smartphones have proven to be effective at behaviour change. (35) Given the immediate and long-term health issues associated with poor dietary habits among adolescents, it is imperative that more innovative research be conducted on how to ameliorate the negative impacts of junk food exposure in adolescents' environments, particularly strategies and interventions that promote lifelong healthy behaviours.
REFERENCES
(1.) Yusuf S, Hawken S, Ounpuu S, et al. Obesity and the risk of myocardial infarction in 27,000 participants from 52 countries: A case-control study. Lancet 2005; 366:1640-1649. PMID: 16271645. doi: doi:10.1016/S0140-6736 (05)67663-5.
(2.) Statistics Canada. Overweight and obese youth (self-reported), 2014. Statistics Canada Health Fact Sheets, 2015; 82-625, http://www.statcan.gc.ca/pub/82-625-x/2015001/article/14186-eng.htm.
(3.) He M, Tucker P, Gilliland J, Irwin JD, Larsen K, Hess P. The influence of local food environments on adolescents' food purchasing behaviors. Int J Environ Res Public Health 2012; 9(4): 1458-1471. PMID: 22690205. doi: 10.3390/ ijerph9041458.
(4.) He M, Tucker P, Irwin JD, Gilliland J, Larsen K, Hess P. Obesogenic neighbourhoods: The impact of neighbourhood restaurants and convenience stores on adolescents' food consumption behaviours. Public Health Nutr 2012; 15(12): 2331-2339. PMID: 22390896. doi: 10.1017/S1368980012000584.
(5.) Gilliland JA, Rangel CY, Healy MA, Tucker P, Loebach JE, Hess PM, et al. Linking childhood obesity to the built environment: A multi-level analysis of home and school neighbourhood factors associated with body mass index. Can J Public Health 2012; 103(9): eS15-eS21. PMID: 23618083. doi: 10.17269/ cjph.103.3283.
(6.) Gilliland J. (2010). The built environment and obesity: trimming waistlines through neighborhood design. In: Bunting T, Filion P, Walker R, eds. Canadian Cities in Transition. Oxford, UK: Oxford University Press, pp. 391-410.
(7.) Bowman SA, Gortmaker SL, Ebbeling CB, Pereira MA, Ludwig DS. Effects of fast-food consumption on energy intake and diet quality among children in a national household survey. Pediatrics 2004; 113(1): 112-118. doi: 10.1542/ peds.113.1.112.
(8.) Loebach J, Gilliland J. Child-led tours to uncover children's perceptions and use of neighborhood environments. Children Youth Environ 2010; 20(1): 52-90. doi: 10.7721/chilyoutenvi.20.1.0052.
(9.) Sallis JF, Glanz K. The role of built environments in physical activity, eating, and obesity in childhood. Fut Children 2006; 16(1): 89-108. doi: 10.1353/foc. 2006.0009.
(10.) Sadler RC, Gilliland JA. Comparing children's GPS tracks with geospatial proxies for exposure to junk food. Spatial Spatio-Temp Epidemiol 2015; 14:55-61.
(11.) Shearer C, Rainham D, Blanchard C, Dummer T, Lyons R, Kirk S. Measuring food availability and accessibility among adolescents: Moving beyond the neighbourhood boundary. Social Sci Med 2015; 133(May): 322-330. doi: 10. 1016/j.socscimed.2014.11.019.
(12.) Chaix B, Meline J, Duncan S, Merrien C, Karusisi N, Perchoux C, et al. GPS tracking in neighborhood and health studies: A step forward for environmental exposure assessment, a step backward for causal inference? Health Place 2013; 21: 46-51. PMID: 23425661. doi: 10.1016/j.healthplace. 2013.01.003.
(13.) Morland K, Diez-Roux AV, Wing S. Supermarkets, other food stores, and obesity: The Atherosclerosis Risk in Communities Study. Am J Prev Med 2006; 30(4): 333-339. PMID: 16530621. doi: 10.1016/j.amepre.2005.11.003.
(14.) Kestens Y, Daniel M. Social inequalities in food exposure around schools in an urban area. Am J Prev Med 2010; 39(1): 33-40. PMID: 20537844. doi: 10.1016/]. amepre.2010.03.014.
(15.) Lovasi S, Grady S, Rundle A. Steps forward: Review and recommendations for research on walkability, physical activity and cardiovascular health. Public Health Rev 2012; 33(2): 484-506. PMID: 25237210.
(16.) Williams J, Scarborough P, Matthews A, Cowburn G, Foster C, Roberts N, et al. A systematic review of the influence of the retail food environment around schools on obesity-related outcomes. Obes Rev 2014; 15(5): 359-374. PMID: 24417984. doi: 10.1111/obr.12142.
(17.) Kestens Y, Lebel A, Chaix B, Clary C, Daniel M, Pampalon R, et al. Association between activity space exposure to food establishments and individual risk of overweight. PloS one 2012; 7(8): e414-18. PMID: 22936974. doi: 10.1371/ journal.pone.0041418.
(18.) Boruff BJ, Nathan A, Nijenstein S. Using GPS technology to (re)-examine operational definitions of 'neighbourhood' in place-based health research. IntJHealth Geogr 2012; 11(1): 22. PMID: 22738807. doi: 10.1186/1476-072X11-22.
(19.) Harrison F, Burgoine T, Corder K, van Sluijs EM, Jones A. How well do modelled trips to school record the environments children are exposed to? A cross-sectional comparison of GIS-modelled and GPS-measured trips to school. IntJ Health Geogr 2014; 13(5). PMID: 24529075. doi: 10.1186/1476072X-13-5.
(20.) Engler-Stringer R, Le H, Gerrard A, Muhajarine N. The community and consumer food environment and children's diet: A systematic review. BMC Public Health 2014; 14(1): 522. PMID: 24884443. doi: 10.1186/14712458-14-522.
(21.) Leal C, Chaix B. The influence of geographic life environments on cardiometabolic risk factors: A systematic review, a methodological assessment and a research agenda. Obes Rev 2011; 12(3): 217-230. PMID: 20202135. doi: 10.1111/j.1467-789X.2010.00726.x.
(22.) Madsen KA, Cotterman C, Thompson HR, Rissman Y, Rosen NJ, Ritchie LD. Passive commuting and dietary intake in fourth and fifth grade students. Am J Prev Med 2015; 48(3): 292-299. PMID: 25547928. doi: 10.1016/j.amepre.2014. 09.033.
(23.) Healy MA, Gilliland JA. Quantifying the magnitude of environmental exposure misclassification when using imprecise address proxies in public health research. Spat Spatiotemporal Epidemiol 2012; 3(1): 55-67. PMID: 22469491. doi: 10.1016/j.sste.2012.02.006.
(24.) Loebach J, Gilliland J. Free range kids? Using GPS-derived activity spaces to examine children's independent neighborhood activity and mobility. Environ Behav 2014; 1-33. doi: 10.1177/0013916514543177.
(25.) Larsen K, Merlo J. Appropriate assessment of neighborhood effects on individual health: Integrating random and fixed effects in multilevel logistic regression. Am J Epidemiol 2005; 161(1): 81-88. PMID: 15615918. doi: 10.1093/aje/kwi017.
(26.) Muthen LK, Muthen BO. Mplus User's Guide [computer program]. Los Angeles, CA: Muthen and Muthen, 2015.
(27.) Van Hulst A, Barnett TA, Gauvin L, Daniel M, Kestens Y, Bird M, et al. Associations between children's diets and features of their residential and school neighbourhood food environments. Can J Public Health 2012; 103(9): eS48-eS54. PMID: 23618089.
(28.) Jabs J, Devine CM. Time scarcity and food choices: An overview. Appetite 2006; 47(2): 196-204. PMID: 16698116. doi: 10.1016/j.appet.2006.02.014.
(29.) Larsen K, Gilliland J, Hess P. Route based analysis to capture the environmental influences on a child's mode of travel between home and school. Annals of the Association of American Geographers 2012; 102(6): 1348-1365. doi: 10.1080/00045608.2011.627059.
(30.) Larsen K, Gilliland J, Hess PM, Tucker P, Irwin J, He M. The influence of the physical environment and sociodemographic characteristics on children's mode of travel to and from school. Am J Public Health 2009; 99(3): 520-526. PMID: 19106422. doi:10.2105/AJPH.2008.135319.
(31.) Breslin FC, Koehoorn M, Cole DC. Employment patterns and work injury experience among Canadian 12 to 14 year olds. Can J Public Health 2008; 201-205. PMID: 18615942. doi: 10.17269/cjph.99.1630.
(32.) Fraser LK, Edwards KL, Cade J, Clarke GP. The geography of fast food outlets: A review. Int J Environ Res Public Health 2010; 7(5): 2290-2308. PMID: 20623025. doi:10.3390/ijerph7052290.
(33.) Neumark-Sztainer D, Story M, Perry C, Casey MA. Factors influencing food choices of adolescents: Findings from focus-group discussions with adolescents. J Am Dietetic Assoc 1999; 99(8): 929-937. doi: 10.1016/S00028223(99)00222-9.
(34.) Story M, Neumark-Sztainer D, French S. Individual and environmental influences on adolescent eating behaviors. J Am Dietetic Assoc 2002; 102(3): S40-S51. doi: 10.1016/S0002-8223(02)90421-9.
(35.) Gilliland J, Sadler R, Clark A, O'Connor C, Milczarek M, Doherty S. Using a smartphone application to promote healthy dietary behaviours and local food consumption. BioMed Res Int, 2015. doi: 10.1155/2015/841368.
Richard C. Sadler, PhD, [1] Andrew F. Clark, PhD, [2] Piotr Wilk, PhD, [3] Colleen O'Connor, PhD, [4] Jason A. Gilliland, PhD [5]
Author Affiliations
[1.] Department of Family Medicine, Division of Public Health, Michigan State University, East Lansing, MI
[2.] Department of Geography, University of Western Ontario, London, ON
[3.] Department of Epidemiology & Biostatistics, Department of Paediatrics, University of Western Ontario, London, ON
[4.] Division of Food & Nutrition Sciences, Brescia University College, London, ON
[5.] Department of Geography, Department of Paediatrics, School of Health Studies, University of Western Ontario, London, ON
Correspondence: Jason A. Gilliland, PhD, Department of Geography, Department of Paediatrics, School of Health Studies, University of Western Ontario, 1151 Richmond St., London, ON N6A5C2, Tel: 519-661-2111, ext. 81239, E-mail: jgillila@uwo.ca
Acknowledgements: The STEAM (Spatial Temporal Environment and Activity Monitoring) study was jointly funded by the Canadian Institutes of Health Research and the Heart and Stroke Foundation of Canada, with seed funding from the Social Sciences and Humanities Research Council of Canada. Additional support was provided by the Children's Health Research Institute and the Children's Health Foundation. We thank the students, parents, teachers, principals and research boards from the Thames Valley District School Board, the London District Catholic School Board, Conseil scolaire catholique Providence and the Conseil scolaire Viamonde. We would also like to acknowledge the dozens of research assistants from the Human Environments Analysis Laboratory who helped with the STEAM project.
Conflict of Interest: None to declare. Table 1. Results from multilevel logistic regression models for the effects of exposure to junk food on junk food purchasing Model Log-odds SE Wald P Overall effect Intercept -4.071 0.200 20.323 0.000 (n = 4588) Exposure ([beta]) 0.160 0.016 9.822 0.000 Variance ([tau]) 3.271 0.710 4.610 0.000 By mode of transportation Active Intercept -4.280 0.441 9.708 0.000 (n =1414) Exposure ([beta]) 0.121 0.031 3.912 0.000 Variance ([tau]) 5.146 1.787 2.880 0.004 Bus Intercept -5.660 0.683 8.284 0.000 (n =1790) Exposure ([beta]) 0.019 0.051 0.370 0.711 Variance ([tau]) 3.880 1.920 2.021 0.043 Car Intercept -3.612 0.319 11.322 0.000 (n =1384) Exposure ([beta]) 0.197 0.026 7.607 0.000 Variance ([tau]) 2.786 1.087 2.564 0.010 By trip type To school Intercept -5.442 0.695 7.828 0.000 (n = 2221) Exposure ([beta]) 0.197 0.043 4.619 0.000 Variance ([tau]) 3.467 2.071 1.674 0.094 From school Intercept -3.449 0.208 16.591 0.000 (n = 2367) Exposure ([beta]) 0.115 0.018 6.421 0.000 Variance ([tau]) 2.637 0.658 4.007 0.000 By sex Female Intercept -3.886 0.231 16.788 0.000 (n = 2806) Exposure ([beta]) 0.174 0.019 8.951 0.000 Variance ([tau]) 2.572 0.744 3.459 0.001 Male Intercept -4.388 0.375 11.686 0.000 (n =1782) Exposure ([beta]) 0.117 0.030 3.864 0.000 Variance ([tau]) 4.599 1.537 2.993 0.003 Model Odds Confidence ICC interval Overall effect Intercept 0.017 (n = 4588) Exposure ([beta]) 1.174 1.14-1.21 Variance ([tau]) 0.499 By mode of transportation Active Intercept 0.014 (n =1414) Exposure ([beta]) 1.129 1.06-1.20 Variance ([tau]) 0.610 Bus Intercept 0.003 (n =1790) Exposure ([beta]) 1.019 0.92-1.13 Variance ([tau]) 0.541 Car Intercept 0.027 (n =1384) Exposure ([beta]) 1.218 1.16-1.28 Variance ([tau]) 0.459 By trip type To school Intercept 0.004 (n = 2221) Exposure ([beta]) 1.218 1.12-1.33 Variance ([tau]) 0.513 From school Intercept 0.032 (n = 2367) Exposure ([beta]) 1.122 1.08-1.16 Variance ([tau]) 0.445 By sex Female Intercept 0.021 (n = 2806) Exposure ([beta]) 1.190 1.15-1.24 Variance ([tau]) 0.439 Male Intercept 0.012 (n =1782) Exposure ([beta]) 1.124 1.06-1.19 Variance ([tau]) 0.583 Model MOR Overall effect Intercept (n = 4588) Exposure ([beta]) Variance ([tau]) 5.613 By mode of transportation Active Intercept (n =1414) Exposure ([beta]) Variance ([tau]) 8.705 Bus Intercept (n =1790) Exposure ([beta]) Variance ([tau]) 6.546 Car Intercept (n =1384) Exposure ([beta]) Variance ([tau]) 4.914 By trip type To school Intercept (n = 2221) Exposure ([beta]) Variance ([tau]) 5.907 From school Intercept (n = 2367) Exposure ([beta]) Variance ([tau]) 4.707 By sex Female Intercept (n = 2806) Exposure ([beta]) Variance ([tau]) 4.617 Male Intercept (n =1782) Exposure ([beta]) Variance ([tau]) 7.734 SE, standard error; ICC, intra-class correlation coefficient; MOR, median odds ratio.