Journal of Physical Activity and Health, 2015, 12, 909 -914 http://dx.doi.org/10.1123/jpah.2014-0039 © 2015 Human Kinetics, Inc.
ORIGINAL RESEARCH
A Longitudinal Study of Objectively Measured Built Environment as Determinant of Physical Activity in Young Adults: The European Youth Heart Study Jasper Schipperijn, Mathias Ried-Larsen, Merete S. Nielsen, Anneli F. Holdt, Anders Grøntved, Annette K. Ersbøll, and Peter L. Kristensen Background: This longitudinal study aimed to examine if a Movability Index (MI), based on objectively measured built environment characteristics, was a determinant for objectively measured physical activity (PA) among young adults. Methods: Data collected from 177 persons participating in the Danish part of the European Youth Hearth Study (EYHS) was used to examine the effect of the built environment on PA. A MI was developed using objectively measured built environment characteristics, and included residential density, recreational facilities, daily destinations and street connectivity. Results: Results showed a positive cross-sectional association between MI and PA. PA decreased from baseline to follow-up. MI increased, primarily due to participants relocating to larger cities. An increase in MI from baseline to follow-up was associated with a reduced decrease in PA for females. Conclusions: Our findings suggest that the built environment is a determinant for PA, especially for females. The found gender differences might suggest the need to develop gender specific environmental indices in future studies. The validity of the measures can be further improved by creating domain specific PA measures as well as domain specific environmental indices and this can potentially reveal more specific built environment determinants for PA. Keywords: movability index, accelerometry, GIS, EYHS
Physical inactivity increases the risk of many adverse health conditions such as coronary heart disease, type 2 diabetes, and breast and colon cancers, and shortens life expectancy.1 Worldwide 31.3% and in Denmark 35.1% of adults (15 years and older) do not meet international physical activity (PA) guidelines.2 A large review of reviews by Bauman et al3 of potential correlates of PA, concluded that the individual level factors age, sex, health status, self-efficacy, and motivation are associated with PA. They furthermore concluded that there is a need to further study the importance of social and physical environmental factors as potential correlates. At present, few consistent environmental correlates have been identified for transport and leisure related PA.3 Reviews by Wendel-Vos et al4 and Ding et al5 suggested that these inconclusive results might be due to differences in definitions of the environmental correlates used, as well as differences in measuring PA. Ding et al5 suggested that more studies with improved conceptualization and better measures are needed to expand the current knowledge base. To overcome some of the measurement issues, and to increase interarea and intercountry comparability, Frank et al6 developed a standardized walkability index. Based on a similar systematic approach, Buck et al7 developed a broader Movability Index (MI) that included street connectivity, destination density and level of urbanization. Inspired by these 2 indexes, a MI was developed in the
present paper for use in a Danish context. Walkability indexes are commonly calculated for network buffers with threshold distances varying between 500 m and 1600 m. The review by Bauman et al3 revealed that most studies had cross-sectional designs, only providing correlations rather than looking at determinants. They suggested that longitudinal observational studies could identify factors that are causal determinants for PA.3 The unique data from the European Youth Heart Study (EYHS) with longitudinal objectively measured PA data, as well as objectively measured built environment variables, were used for the current study. The National Longitudinal Study of Adolescent Health study in the United States did use several objectively measured built environment variables for similar age-groups, but PA was self-reported.8,9 Also the longitudinal studies of other age-groups by Lee et al10 and Coogan et al11 used a combination of self-reported PA and objective built environment measures. The main aims of this study were to examine 1) the relationship between MI and objectively measured PA in adolescents at baseline and young adults at follow-up and 2) the effect of changes in MI from adolescence to young adulthood on objectively measured PA in young adults.
Methods Participants and Design
Schipperijn ([email protected]), Ried-Larsen, Grøntved, and Kristensen are with the Dept of Sport Science and Clinical Biomechanics, University of Southern Denmark, Odense, Denmark. Nielsen is with the Danish Federation for Company Sports, Nyborg, Denmark. Holdt is with the Kalundborg Gymnasium & HF, Kalundborg, Denmark. Ersbøll is with the National Institute of Public Health, University of Southern Denmark, Copenhagen, Denmark.
The EYHS is an international population-based longitudinal study that addresses the biological, environmental, demographic and lifestyle correlates and determinants of cardiovascular disease risk factors in children and adolescents. The study design and the sampling procedures have been described in detail elsewhere.12 In 2003–2004, a random sample of 771 Danish 15-year-old adolescents was invited to participate, of whom 444 adolescents agreed to participate (58% of the sample). A 6-year follow-up was conducted 909
910 Schipperijn et al
in 2009–2010, where all participants were reinvited to participate. A total of 214 persons (48%) agreed to participate. All participant addresses were geocoded in a Geographic Information System (GIS), enabling the inclusion of a range of objectively measured environmental variables. Complete data on exposure and outcome for both time-periods were available for 177 participants and are reported in the current study. The majority (79.6%) of participants moved to a different address from baseline to follow-up. The study was approved by the Regional Scientific Ethical Committee for Southern Denmark and data were collected according the Helsinki declaration. All participants gave a written informed consent.
Downloaded by Washington Univ In St Louis on 09/19/16, Volume 12, Article Number 7
Physical Activity Mean daily total PA was assessed using the Actigraph accelerometer model 7164 at baseline and the Actigraph accelerometer model GT3X or model GT1M at follow-up (ActiGraph, Pensacola, FL). We chose to present PA is as counts per minute (CPM) and not the more commonly used minutes of moderate to vigorous physical activity (MVPA) per day. The main reason for doing so is 2-fold: 1) the ongoing debate regarding appropriate cut-points for moderate activity and the large number of different cut points used makes it difficult to decide on the most appropriate cut point and 2) energy expenditure occurs during activity of all intensities and for that reason the average activity level is a good proxy for mean daily total PA. The participants were instructed to wear the accelerometer for 7 consecutive days and only remove it during showering, swimming or during night-time sleep. All activity files were screened using open-source accelerometer software (Propero v.1.0.18). Consecutive strings of zeros for more than 60 min were defined as “accelerometer not worn” and removed. Subsequently, activity files were only included if they included at least 3 valid days of monitoring including at least 1 weekend day. A valid day should include at least 60% of average daily awake time (9 hrs and 36 min). The activity files were manually screened to identify and remove data from 1) malfunctioning equipment and 2) activity monitors not removed during night-time sleep, as described in Kristensen et al.13 The mean daily total PA was calculated as the mean value of the total PA per day (in the following referred to PA). As there are a number of potential sources of error in the PA data,14 The mean daily total PA was adjusted for within-week variation as described in Kristensen et al,13 more specifically this was done by weighting each week day by 5/7 and each weekend day by 2/7. At baseline, PA was further adjusted for seasonal variation.13 As there was no indication of seasonal variation at follow-up, this adjustment was not performed.
Built Environmental Variables Four built environment variables were calculated: residential density, recreational facilities, daily destinations, and street connectivity. Geographic Data. All objective measurements were derived from
high resolution geographic data from the Danish Geodata Agency (land use) and the Danish Building and Housing Register (building use). Data from 2003–2004 and 2009–2010 were used to analyze the built environment at baseline and follow-up respectively, making it possible to account for changes in the built environment between 2003–2004 and 2009–2010. All measures were computed in ArcGIS 10.1. The geographic unit of analysis was the home-neighborhood for each participant, defined by a 2 km network buffer around each
individual home address. We choose for a 2 km threshold, larger than the commonly used 1 km or 1 mile buffers for walkability measures, as we hypothesized that the area relevant for all types of activities needed to be larger than the area relevant for walking. Movability Index. Inspired by the walkability index developed
by Frank et al6 and the MI developed by Buck et al,7 an MI was developed for use in a Danish context. The MI includes 4 neighborhood characteristics that are thought to influence PA positively: residential density, share of recreational facility area, density of daily destinations, and street connectivity, within each buffer zone. The MI was calculated by taking the sum of the z-scores of each of the 4 built environment variables. The z-scores were calculated based on the mean and SD for the combined values of 2003–2004 and 2009–2010 to be able to compare scores across the 2 time periods and calculate changes. The change in MI from baseline to follow-up (ΔMI) was calculated as MIFollow-up – MIBaseline.
Residential Density. The number of residences in each neighborhood was calculated and served as an indicator for population density. For buildings with multiple residences, each residence was counted separately. Recreational Facilities. Counting the number of recreational
facilities does not take the size of the areas into account and this would lead to a low score in neighborhoods with for example 1 large park, while in reality large parks are considered more attractive for recreation than smaller parks.15,16 For this reason, the share of recreational facility area was calculated. Inspired by Norman et al17 the following land uses were included in the analysis as recreational facilities: sports-clubs and stadiums, parks, woodlands, wet-lands and churchyards.
Daily Destinations. Based on their relevance as regularly reoccurring destinations, the following building use categories were counted as daily destinations: retail, supermarkets, sport-clubs, schools and educational institutions, and cultural facilities such as libraries and theaters. Street Connectivity. Street connectivity was calculated based on
the number of intersections with 3 or more connections. The available data included pedestrian trails and bike paths, the importance of which was stressed by Chin et al.18 Motorways and expressways with no access for bikes or pedestrians were excluded. To avoid double intersections (eg, 2 3-way intersections near each other or double tracked roads), 10 m buffers around each intersection were created and overlapping buffers were merged, and a new buffer centroid was used as new intersection point, as described in Forsyth et al.19
Covariates Parental educational level was assessed by parental self-report using the International Standard Classification of Education (ISCED,20) and this was used as indicator for family-socioeconomic-status (family SES). To reduce problems with an insufficient number of participants in some categories, the ISCED scale was recoded into 3 categories (low: ISCED 0–2, < 10 yrs of education; middle: ISCED 3–4, 10–12 yrs of education; high: ISCED 5–7, > 12 yrs of education). If there were differences in educational level between parents, the highest level of the 2 was used in the analysis. Adiposity was expressed as the sum of 4 skinfold measurements from the biceps, triceps, subcapsular, and suprailiac. Skinfold thickness was measured with a Harpenden caliper. Registration of relocation of the participants was done using their home addresses at baseline and follow-up (relocated/not relocated).
JPAH Vol. 12, No. 7, 2015
Built Environment as Determinant of PA 911
Downloaded by Washington Univ In St Louis on 09/19/16, Volume 12, Article Number 7
Statistical Analyses Descriptive statistics were calculated for both included and excluded participants. Exclusion criteria were 1) drop-outs, 2) drop-ins, and 3) incomplete or invalid data on exposure or outcome measures. A multivariable analysis of variance was used to evaluate the association between PA and MI (and each of the 4 MI components) at baseline and follow-up, respectively, and to examine the effect of changes in MI from baseline to follow-up on PA at follow-up. The association between MI and PA at baseline and follow-up, respectively, were adjusted for differences in gender, and secondarily for family SES and adiposity. The analysis of the associations between changes in MI and PA at follow-up was also adjusted for baseline MI. Beta-values (the slopes) with 95% confidence intervals are reported as well as standardized beta-values. The standardized beta-values indicate the strength of the associations (ie, how many standard deviations PA will change) for each standard deviation increase in the MI variable. Manual inspections of plots of outcomes against exposure and inspection of components-plus-residual plots did not reveal any nonlinear relationships. Variance inflation factors did not reveal any signs of collinearity between covariates. All analyses were checked for modification by gender (P < .05 for interaction). All statistical analyses were performed in STATA 11.2 (STATA Corp. Fort Valton TX) with a 5% significance level.
Results
gender differences were observed in ΔMI. A significant difference in the mean ΔMI from baseline to follow-up was observed between participants who relocated [ΔMI = 3.22 (SE = 3.78)], compared with those who did not [ΔMI = 0.15 (0.13)]. However, no differences were observed in PA, family SES, or adiposity between the 2 groups (P > .1) (data not shown). There were also no differences between the included or excluded participants at baseline (data not shown).
Cross-Sectional Associations Table 2 describes the cross-sectional association between MI (and each of the 4 MI components) and PA at baseline and follow-up. At baseline, MI was positively associated to PA. An increase in MI of 1 unit leads to a 0.15 units increase in PA (P = .036). No gender modification was observed (P > .1). The association was stronger in females compared with males. Of the MI components only the number of intersections was significantly associated to PA at baseline (P = .003). At follow-up, modification by gender was observed (P = .003 for interaction). MI was significantly associated to PA for females (P = .02). No associations between MI or its components and PA were observed in males at follow-up (P > .1) expect for the number of intersections (P = .012). In females, the number of daily destinations was significantly associated to PA (P = .008). Further adjustment for family SES and adiposity did not change the associations substantially.
Population Characteristics
Longitudinal Associations
Characteristics of the participants at baseline and follow-up are presented in Table 1, stratified by gender. On average MI increased 2.4 units from baseline to follow-up for males and 2.7 units for females. PA decreased from 416 CPM to 343 CPM for males, and from 357 CPM to 313 CPM for females. PA for males was significantly higher than for females at baseline, whereas adiposity was significantly higher for females at both baseline and follow-up. No
Table 3 shows the associations between the change in MI and the MI-components from baseline to follow-up (ΔMI) and PA at followup adjusted for baseline MI. In females, an increase of 1 MI unit was associated with a 0.27 unit higher PA level at follow-up (P = .02). An increase in daily destinations from baseline to follow-up was significantly associated with follow-up PA (β = 0.32, P = .008). Including SES and adiposity (model 2) did not change the
Table 1 Population Characteristics at Baseline (2003–2004) and Follow-up (2009–2010), Mean (SD), Percentage (%) and Median (Q1; Q3) 2003–2004 Males
2009–2010 Females
Males
Females
Age
15.7 (0.3)
15.7 (0.4)
21.8 (0.3)
21.7 (0.4)
BMI
20.8 (2.4)
21.3 (3.0)
24.3 (3.6)
28.8 (4.1)
Adiposity (mm)
32.3 (16.5)
53.2 (18.1)*
54.0 (28.5)
73.4 (27.7)*
11/37/52
5/45/50
–
–
Family SES (% low/middle/high) Physical activity (CPM)
416 (160)
357 (139)*
343 (130)
313 (122)
Daily MVPA (minutes)
25.1 (17.7 to 38.2)
16.8 (9.9 to 28.4)*
18.3 (9.1 to 27.1)
13.8 (6.6 to 23.0)*
Daily wear time (hours)
14.3 (1.1)
14.2 (1.1)
14.2 (1.4)
14.0 (1.2)
Accepted days Movability Index Intersections
4.6 (0.8)
4.6 (0.8)
4.6 (0.9)
4.6 (1.0)
–1.59 (2.57)
–1.11 (2.48)
0.84 (3.10)
1.62 (3.06)
–0.50 (0.97)
–0.30 (0.97)
0.25 (0.95)
0.47 (0.86)
Residential density
–0.41 (–0.76 to 0.26)
–0.33 (–0.68 to 0.38)
0.01 (–0.56 to 0.86)
0.37 (–0.23 to 1.00)
Daily destinations
–0.48 (–0.50 to –0.24)
–0.44 (–0.50 to 0.01)
–0.30 (–0.64 to 1.10)
0.18 (–0.52 to 1.19)
–0.13 (1.12)
–0.04 (1.00)
0.02 (0.97)
0.13 (0.94)
Recreational facilities *P < .05 for gender differences
JPAH Vol. 12, No. 7, 2015
912 Schipperijn et al
Table 2 Cross-sectional Associations Between Movability Index and Physical Activity at Baseline and Follow-up Baseline
Follow-up*
Estimate (SE) β (95% CI)
Exposure
Estimate (SE) Std. β
P
β (95% CI)
Std. β
P
All (N = 177)a Movability Index Intersections
9.54 (0.65 to 18.44)
0.15
0.036
1.41 (–4.70 to 7.52)
0.03
0.65
35.56 (12.33 to 58.79)
0.22
0.003
–4.59 (–25.52 to 16.14)
–0.03
0.66
Residential density
15.23 (–7.53 to 38.00)
0.10
0.19
3.56 (–12.33 to 19.45)
0.03
0.66
Daily destinations
–1.30 (–24.38 to 21.78)
–0.01
0.08
11.73 (–4.36 to 27.82)
0.11
0.15
Recreational facilities
19.23 (–2.02 to 40.49)
0.13
0.61
–4.01 (–23.75 to 15.73)
–0.03
0.69
Movability Index
12.01 (0.92 to 23.09)
0.21
0.034
9.34 (1.58 to 17.10)
0.23
0.019
Intersections
37.52 (8.50 to 66.54)
0.25
0.012
27.05 (–0.74 to 54.86)
0.19
0.056
Downloaded by Washington Univ In St Louis on 09/19/16, Volume 12, Article Number 7
Female (N = 101)
Residential density
20.22 (–6.42 to 46.86)
0.15
0.14
17.41 (–2.73 to 37.57)
0.17
0.089
Daily destinations
11.08 (–15.32 to 37.47)
0.08
0.41
26.77 (6.99 to 46.55)
0.26
0.008
Recreational facilities
22.68 (–4.96 to 50.33)
0.16
0.11
5.07 (–20.96 to 31.11)
0.04
0.70
6.48 (–8.30 to 21.26)
0.10
0.39
–8.89 (–18.35 to 0.63)
–0.21
0.067
Male (N = 76) Movability Index Intersections
33.12 (–5.45 to 71.71)
0.20
0.091
–39.25 (–69.50 to –9.00)
–0.29
0.012
Residential density
15.58 (–18.12 to 49.28)
0.11
0.36
–15.21 (–45.96 to 15.52)
–0.11
0.33
Daily destinations
–26.09 (–69.71 to 17.54)
–0.14
0.24
–10.83 (–37.55 to 15.87)
–0.09
0.42
Recreational facilities
15.58 (–18.12 to 49.28)
0.11
0.36
–15.22 (–45.95 to 15.12)
–0.11
0.33
a Adjusted
for gender. * P = .003 for gender interaction.
associations substantially, but decreased the p-values slightly. No significant association was observed between a change in MI and PA at follow-up in men (P > .1). For intersection density, a negative association was significant and an increase of 1 unit in the number of intersections from baseline to follow-up was associated with a decrease of 0.34 units in PA at follow-up (P = .02). Further adjustment for family SES and adiposity (model 2) did not substantially alter the associations. Further adjustment for baseline PA level (data not shown) did not change the observations noticeably. As relocation affected the PA level, the longitudinal analyses were adjusted for relocation status (relocated/not relocated), but this did not change the associations.
Discussion The main strengths of this study were the novel combination of a longitudinal study design and the use of objectively measured PA data, as well as detailed measured built environment characteristics. The main findings of the study were that objectively measured PA and MI were associated at both baseline and follow-up, which is similar to findings from cross-sectional studies (eg21–24). Furthermore, an increase in MI from baseline to follow-up was, for females, associated with a reduced decrease in PA at follow-up. These findings indicate that the residential environment people live in is a casual determinant for their PA behavior. The results show a distinct gender difference in the association between ∆MI and PA which is surprising as there were no
significant differences in MI or PA at follow-up for males and females. This difference is most likely explained by the fact that there was no association between MI and PA for males at follow-up, while this association was present for females. A reason for these gender differences could be that PA in males is influenced by other built environment factors than in females, or that PA in males is less influenced by environmental factors. The results imply gender related effect modification, as also suggested by Boone-Heinonen and Gordon-Larsen,8 who in a longitudinal study found that a higher street connectivity was associated with slightly less PA for women, but not for men.8 In a longitudinal study by Lee et al,10 no relation was found between PA for men and urban sprawl, however few men in this study relocated, and the urban sprawl measure used was rather crude. The gender difference deserves more attention in future studies. Most of the changes in MI in this study were due to participants relocating (79.6% moved to a different address), primarily to go studying, and for that reason, the change in PA behavior could perhaps also be related to other changes in the respondents’ lives occurring simultaneously. Moving away from the parental home to start an independent life is a big change for many young adults and it cannot be ruled out that other factors related to relocation could explain our results. Then again, we feel that it is not unreasonable to assume that the environment young adults relocate to, and in particular the number of opportunities that environment offers for a more active lifestyle, has some effect on their behavior. This is supported by fact that adjustment for relocation did not change the
JPAH Vol. 12, No. 7, 2015
Built Environment as Determinant of PA 913
Table 3 Association Between the Change in Movability Index From Baseline to Follow-up and Adult Physical Activity Model 1
Model 2
Estimate (SE) Exposure
β (95% CI)
Estimate (SE) Std. β
P
β (95% CI)
Std. β
P
All (N = 177) ∆ Movability Index
3.02 (–3.14 to 9.19)
0.09
0.35
3.71 (–2.55 to 9.96)
0.11
0.35
Intersections
–1.71 (–22.84 to 19.41)
–0.02
0.87
1.07 (–20.43 to 22.52)
0.01
0.92
Residential density
5.59 (–10.46 to 21.64)
0.06
0.49
7.10 (–9.23 to 23.43)
0.08
0.39
Daily destinations
14.31 (–1.88 to 30.50)
0.16
0.08
16.85 (0.42 to 33.28)
0.19
0.045
Recreational facilities
5.47 (–15.75 to 26.67)
0.05
0.61
4.04 (–17.50 to 25.58)
0.04
0.71
Females (N = 101) Downloaded by Washington Univ In St Louis on 09/19/16, Volume 12, Article Number 7
∆ Movability Index
9.27 (1.33 to 17.18)
0.27
0.02
10.15 (2.08 to 18.21)
0.30
0.014
Intersections
24.45 (–3.92 to 52.82)
0.23
0.09
28.24 (–0.80 to 57.27)
0.26
0.06
Residential density
16.34 (–4.06 to 36.74)
0.20
0.12
19.06 (–1.78 to 39.90)
0.23
0.07
Daily destinations
27.40 (7.20 to 47.61)
0.32
0.008
31.24 (10.64 to 51.84)
0.37
0.003
Recreational facilities
8.70 (–19.60 to 37.00)
0.08
0.54
5.57 (23.87 to 35.01)
0.05
0.71
–7.75 (–17.39 to 1.88)
–0.22
0.11
–7.55 (–17.46 to 2.36)
–0.21
0.13
Intersections
–36.40 (–67.15 to –5.65)
–0.34
0.02
–35.47 (–67.10 to –3.83)
–0.33
0.03
Residential density
–11.00 (–36.94 to 14.94)
–0.11
0.40
–10.43 (–36.81 to 15.95)
–0.11
0.43
Daily destinations
–8.11 (–35.01 to 18.79)
–0.08
0.55
–6.56 (–33.92 to 20.80)
–0.07
0.63
Recreational facilities
–11.55 (–44.89 to 21.80)
–0.11
0.49
–13.28 (–47.63 to 21.08)
–0.12
0.44
Males (N = 76) ∆ Movability Index
Note. Model 1 is adjusted for baseline movability index and gender. Model 2 is further adjusted for family SES and adiposity.
associations. New habits are created when starting a new life stage and, in theory, these habits are easier to establish if both the physical and social environment are supportive of them. A limitation of the study was that it was not possible to distinguish between PA in different domains (such as walking, recreation, or transport) based on accelerometer data alone, and this also meant that it was not possible to address the relative importance of the different settings for different persons. The same mean CPM could stem from activity in very different settings. Furthermore, since only overall PA could be assessed, an environmental index that describes the environmental suitability for all components of PA needed to be developed. The composite nature of the MI meant that environmental factors positively associated with one type of activity could be negatively associated with activity in another domain. The lack of domain specific PA data as well as domain specific environmental data might have led to reduced overall associations. In addition, as the observed gender differences potentially indicate the association between the environment and PA might be different for different people, suggesting that the environmental factors included in an index might need to be differentiated by gender, age, or other sociodemographic factors. In their conceptual model for longitudinal studies of the obesogenic environment, Boone-Heinonen and Gordon-Larsen9 suggested the need for more focus on the interactions between respondents’ sociodemographic characteristics and the environmental characteristics of the neighborhood they live in. Another limitation of the study is the relatively small sample size, as well as the fact that female respondents from families with a lower SES were less likely to participate in the follow-up study;
however, there were no significant differences in PA, BMI, or adiposity between included and excluded participants. Furthermore, there might have been an element of positive self-selection to participate in the EYHS study, where more active participants might have been more likely to participate. For future studies, a setup in which accelerometer measurements are combined with GPS measurements could be used to distinguish between PA in different settings/domains. In combination with domain-specific, and perhaps also gender-specific, environmental indexes, this will increase the conceptual validity of the measures.
Conclusion In summary, this study is among the first studies looking at the longitudinal relation between changes in objectively assessed built environment with changes in objectively measured physical activity. Even though the measures can and should be further improved in the future, the findings indicate that built environment indeed plays a role in explaining individual differences in physical activity levels, in particular among females. The gender differences need to be further explored. Acknowledgments We would like to thank the Danish EYHS team for their support, and we would like to thank all the participants who contributed to the study. The Danish part of the European Youth Heart Study was funded by the Danish Council for Strategic Research (grant number 2101-08-0058); The Danish
JPAH Vol. 12, No. 7, 2015
914 Schipperijn et al
Heart Foundation, the Danish Health Fund, the Danish Ministry of the Interior and Health, TrygFonden, the Danish Medical Research Council, and the Ministry of Culture Council for Sports Science Research.
Downloaded by Washington Univ In St Louis on 09/19/16, Volume 12, Article Number 7
References 1. Lee IM, Shiroma EJ, Lobelo F, et al. Effect of physical inactivity on major non-communicable diseases worldwide: an analysis of burden of disease and life expectancy. Lancet. 2012;380(9838):219–229. PubMed doi:10.1016/S0140-6736(12)61031-9 2. Hallal PC, Bauman AE, Heath GW, Kohl HW, 3rd, Lee IM, Pratt M. Physical activity: more of the same is not enough. Lancet. 2012;380(9838):190–191. PubMed doi:10.1016/S01406736(12)61027-7 3. Bauman AE, Reis RS, Sallis JF, et al. Correlates of physical activity: why are some people physically active and others not? Lancet. 2012;380(9838):258–271. PubMed doi:10.1016/S01406736(12)60735-1 4. Wendel-Vos W, Droomers M, Kremers S, Brug J, van Lenthe F. Potential environmental determinants of physical activity in adults: a systematic review. Obes Res. 2007;8(5):425–440. 5. Ding D, Sallis JF, Kerr J, Lee S, Rosenberg DE. Neighborhood environment and physical activity among youth a review. Am J Prev Med. 2011;41(4):442–455. PubMed doi:10.1016/j.amepre.2011.06.036 6. Frank LD, Sallis JF, Saelens BE, et al. The development of a walkability index: application to the Neighborhood Quality of Life Study. Br J Sports Med. 2010;44(13):924–933. PubMed doi:10.1136/ bjsm.2009.058701 7. Buck C, Pohlabeln H, Huybrechts I, et al. Development and application of a moveability index to quantify possibilities for physical activity in the built environment of children. Health Place. 2011;17(6):1191– 1201. PubMed doi:10.1016/j.healthplace.2011.08.011 8. Boone-Heinonen J, Gordon-Larsen P. Life stage and sex specificity in relationships between the built and socioeconomic environments and physical activity. J Epidemiol Community Health. 2011;65(10):847– 852. PubMed doi:10.1136/jech.2009.105064 9. Boone-Heinonen J, Gordon-Larsen P. Obesogenic environments in youth: concepts and methods from a longitudinal national sample. Am J Prev Med. 2012;42(5):e37–e46. PubMed doi:10.1016/j. amepre.2012.02.005 10. Lee IM, Ewing R, Sesso HD. The built environment and physical activity levels: the Harvard Alumni Health Study. Am J Prev Med. 2009;37(4):293–298. PubMed doi:10.1016/j.amepre.2009.06.007 11. Coogan PF, White LF, Adler TJ, Hathaway KM, Palmer JR, Rosenberg L. Prospective study of urban form and physical activity in the Black Women’s Health Study. Am J Epidemiol. 2009;170(9):1105–1117. PubMed doi:10.1093/aje/kwp264 12. Riddoch C, Edwards D, Page A, et al. The European Youth Heart study—cardiovascular disease risk factors in children: rationale,
aims, study design, and validation of methods. J Phys Act Health. 2005;2(1):115–129. 13. Kristensen PL, Moller NC, Korsholm L, Wedderkopp N, Andersen LB, Froberg K. Tracking of objectively measured physical activity from childhood to adolescence: the European youth heart study. Scand J Med Sci Sports. 2008;18(2):171–178. PubMed doi:10.1111/j.16000838.2006.00622.x 14. Kristensen PL, Korsholm L, Moller NC, Wedderkopp N, Andersen LB, Froberg K. Sources of variation in habitual physical activity of children and adolescents: the European youth heart study. Scand J Med Sci Sports. 2008;18(3):298–308. PubMed doi:10.1111/j.16000838.2007.00668.x 15. Schipperijn J, Stigsdotter UK, Randrup TB, Troelsen J. Influences on the use of urban green space—a case study in Odense, Denmark. Urban for Urban Gree. 2010;9(1):25–32. doi:10.1016/j. ufug.2009.09.002 16. Giles-Corti B, Broomhall MH, Knuiman M, et al. Increasing walking: how important is distance to, attractiveness, and size of public open space? Am J Prev Med. 2005;28(2, Suppl 2):169–176. PubMed doi:10.1016/j.amepre.2004.10.018 17. Norman GJ, Nutter SK, Ryan S, Sallis JF, Calfas KJ, Patrick K. Community design and access to recreational facilities as correlates of adolescent physical activity and body-mass index. J Phys Act Health. 2006;3(supplement):s118–s128. 18. Chin GK, Van Niel KP, Giles-Corti B, Knuiman M. Accessibility and connectivity in physical activity studies: the impact of missing pedestrian data. Prev Med. 2008;46(1):41–45. PubMed doi:10.1016/j. ypmed.2007.08.004 19. Forsyth A, D’Sousa E, Koepp J, et al. Twin Cities Walking Study— environment and physical activity: GIS protocols version 4.1, June 2007. 2007; http://www.designforhealth.net/pdfs/GIS_Protocols/ MinnGIS_Ver4_1_FINAL.pdf. 20. UNESCO. ISCED: International Standard Classification of Education. 1997; www.uis.unesco.org/Education/Pages/international-standardclassification-of-education.aspx 21. De Meester F, Van Dyck D, De Bourdeaudhuij I, Deforche B, Sallis JF, Cardon G. Active living neighborhoods: is neighborhood walkability a key element for Belgian adolescents? BMC Public Health. 2012;12:7. PubMed doi:10.1186/1471-2458-12-7 22. Owen N, Cerin E, Leslie E, et al. Neighborhood walkability and the walking behavior of Australian adults. Am J Prev Med. 2007;33(5):387–395. PubMed doi:10.1016/j.amepre.2007.07.025 23. Sallis JF, Saelens BE, Frank LD, et al. Neighborhood built environment and income: examining multiple health outcomes. Soc Sci Med. 2009;68(7):1285–1293. PubMed doi:10.1016/j.socscimed.2009.01.017 24. Van Dyck D, Cardon G, Deforche B, Sallis JF, Owen N, De Bourdeaudhuij I. Neighborhood SES and walkability are related to physical activity behavior in Belgian adults. Prev Med. 2010;50(Suppl 1):S74– S79. PubMed doi:10.1016/j.ypmed.2009.07.027
JPAH Vol. 12, No. 7, 2015
ORIGINAL RESEARCH
A Longitudinal Study of Objectively Measured Built Environment as Determinant of Physical Activity in Young Adults: The European Youth Heart Study Jasper Schipperijn, Mathias Ried-Larsen, Merete S. Nielsen, Anneli F. Holdt, Anders Grøntved, Annette K. Ersbøll, and Peter L. Kristensen Background: This longitudinal study aimed to examine if a Movability Index (MI), based on objectively measured built environment characteristics, was a determinant for objectively measured physical activity (PA) among young adults. Methods: Data collected from 177 persons participating in the Danish part of the European Youth Hearth Study (EYHS) was used to examine the effect of the built environment on PA. A MI was developed using objectively measured built environment characteristics, and included residential density, recreational facilities, daily destinations and street connectivity. Results: Results showed a positive cross-sectional association between MI and PA. PA decreased from baseline to follow-up. MI increased, primarily due to participants relocating to larger cities. An increase in MI from baseline to follow-up was associated with a reduced decrease in PA for females. Conclusions: Our findings suggest that the built environment is a determinant for PA, especially for females. The found gender differences might suggest the need to develop gender specific environmental indices in future studies. The validity of the measures can be further improved by creating domain specific PA measures as well as domain specific environmental indices and this can potentially reveal more specific built environment determinants for PA. Keywords: movability index, accelerometry, GIS, EYHS
Physical inactivity increases the risk of many adverse health conditions such as coronary heart disease, type 2 diabetes, and breast and colon cancers, and shortens life expectancy.1 Worldwide 31.3% and in Denmark 35.1% of adults (15 years and older) do not meet international physical activity (PA) guidelines.2 A large review of reviews by Bauman et al3 of potential correlates of PA, concluded that the individual level factors age, sex, health status, self-efficacy, and motivation are associated with PA. They furthermore concluded that there is a need to further study the importance of social and physical environmental factors as potential correlates. At present, few consistent environmental correlates have been identified for transport and leisure related PA.3 Reviews by Wendel-Vos et al4 and Ding et al5 suggested that these inconclusive results might be due to differences in definitions of the environmental correlates used, as well as differences in measuring PA. Ding et al5 suggested that more studies with improved conceptualization and better measures are needed to expand the current knowledge base. To overcome some of the measurement issues, and to increase interarea and intercountry comparability, Frank et al6 developed a standardized walkability index. Based on a similar systematic approach, Buck et al7 developed a broader Movability Index (MI) that included street connectivity, destination density and level of urbanization. Inspired by these 2 indexes, a MI was developed in the
present paper for use in a Danish context. Walkability indexes are commonly calculated for network buffers with threshold distances varying between 500 m and 1600 m. The review by Bauman et al3 revealed that most studies had cross-sectional designs, only providing correlations rather than looking at determinants. They suggested that longitudinal observational studies could identify factors that are causal determinants for PA.3 The unique data from the European Youth Heart Study (EYHS) with longitudinal objectively measured PA data, as well as objectively measured built environment variables, were used for the current study. The National Longitudinal Study of Adolescent Health study in the United States did use several objectively measured built environment variables for similar age-groups, but PA was self-reported.8,9 Also the longitudinal studies of other age-groups by Lee et al10 and Coogan et al11 used a combination of self-reported PA and objective built environment measures. The main aims of this study were to examine 1) the relationship between MI and objectively measured PA in adolescents at baseline and young adults at follow-up and 2) the effect of changes in MI from adolescence to young adulthood on objectively measured PA in young adults.
Methods Participants and Design
Schipperijn ([email protected]), Ried-Larsen, Grøntved, and Kristensen are with the Dept of Sport Science and Clinical Biomechanics, University of Southern Denmark, Odense, Denmark. Nielsen is with the Danish Federation for Company Sports, Nyborg, Denmark. Holdt is with the Kalundborg Gymnasium & HF, Kalundborg, Denmark. Ersbøll is with the National Institute of Public Health, University of Southern Denmark, Copenhagen, Denmark.
The EYHS is an international population-based longitudinal study that addresses the biological, environmental, demographic and lifestyle correlates and determinants of cardiovascular disease risk factors in children and adolescents. The study design and the sampling procedures have been described in detail elsewhere.12 In 2003–2004, a random sample of 771 Danish 15-year-old adolescents was invited to participate, of whom 444 adolescents agreed to participate (58% of the sample). A 6-year follow-up was conducted 909
910 Schipperijn et al
in 2009–2010, where all participants were reinvited to participate. A total of 214 persons (48%) agreed to participate. All participant addresses were geocoded in a Geographic Information System (GIS), enabling the inclusion of a range of objectively measured environmental variables. Complete data on exposure and outcome for both time-periods were available for 177 participants and are reported in the current study. The majority (79.6%) of participants moved to a different address from baseline to follow-up. The study was approved by the Regional Scientific Ethical Committee for Southern Denmark and data were collected according the Helsinki declaration. All participants gave a written informed consent.
Downloaded by Washington Univ In St Louis on 09/19/16, Volume 12, Article Number 7
Physical Activity Mean daily total PA was assessed using the Actigraph accelerometer model 7164 at baseline and the Actigraph accelerometer model GT3X or model GT1M at follow-up (ActiGraph, Pensacola, FL). We chose to present PA is as counts per minute (CPM) and not the more commonly used minutes of moderate to vigorous physical activity (MVPA) per day. The main reason for doing so is 2-fold: 1) the ongoing debate regarding appropriate cut-points for moderate activity and the large number of different cut points used makes it difficult to decide on the most appropriate cut point and 2) energy expenditure occurs during activity of all intensities and for that reason the average activity level is a good proxy for mean daily total PA. The participants were instructed to wear the accelerometer for 7 consecutive days and only remove it during showering, swimming or during night-time sleep. All activity files were screened using open-source accelerometer software (Propero v.1.0.18). Consecutive strings of zeros for more than 60 min were defined as “accelerometer not worn” and removed. Subsequently, activity files were only included if they included at least 3 valid days of monitoring including at least 1 weekend day. A valid day should include at least 60% of average daily awake time (9 hrs and 36 min). The activity files were manually screened to identify and remove data from 1) malfunctioning equipment and 2) activity monitors not removed during night-time sleep, as described in Kristensen et al.13 The mean daily total PA was calculated as the mean value of the total PA per day (in the following referred to PA). As there are a number of potential sources of error in the PA data,14 The mean daily total PA was adjusted for within-week variation as described in Kristensen et al,13 more specifically this was done by weighting each week day by 5/7 and each weekend day by 2/7. At baseline, PA was further adjusted for seasonal variation.13 As there was no indication of seasonal variation at follow-up, this adjustment was not performed.
Built Environmental Variables Four built environment variables were calculated: residential density, recreational facilities, daily destinations, and street connectivity. Geographic Data. All objective measurements were derived from
high resolution geographic data from the Danish Geodata Agency (land use) and the Danish Building and Housing Register (building use). Data from 2003–2004 and 2009–2010 were used to analyze the built environment at baseline and follow-up respectively, making it possible to account for changes in the built environment between 2003–2004 and 2009–2010. All measures were computed in ArcGIS 10.1. The geographic unit of analysis was the home-neighborhood for each participant, defined by a 2 km network buffer around each
individual home address. We choose for a 2 km threshold, larger than the commonly used 1 km or 1 mile buffers for walkability measures, as we hypothesized that the area relevant for all types of activities needed to be larger than the area relevant for walking. Movability Index. Inspired by the walkability index developed
by Frank et al6 and the MI developed by Buck et al,7 an MI was developed for use in a Danish context. The MI includes 4 neighborhood characteristics that are thought to influence PA positively: residential density, share of recreational facility area, density of daily destinations, and street connectivity, within each buffer zone. The MI was calculated by taking the sum of the z-scores of each of the 4 built environment variables. The z-scores were calculated based on the mean and SD for the combined values of 2003–2004 and 2009–2010 to be able to compare scores across the 2 time periods and calculate changes. The change in MI from baseline to follow-up (ΔMI) was calculated as MIFollow-up – MIBaseline.
Residential Density. The number of residences in each neighborhood was calculated and served as an indicator for population density. For buildings with multiple residences, each residence was counted separately. Recreational Facilities. Counting the number of recreational
facilities does not take the size of the areas into account and this would lead to a low score in neighborhoods with for example 1 large park, while in reality large parks are considered more attractive for recreation than smaller parks.15,16 For this reason, the share of recreational facility area was calculated. Inspired by Norman et al17 the following land uses were included in the analysis as recreational facilities: sports-clubs and stadiums, parks, woodlands, wet-lands and churchyards.
Daily Destinations. Based on their relevance as regularly reoccurring destinations, the following building use categories were counted as daily destinations: retail, supermarkets, sport-clubs, schools and educational institutions, and cultural facilities such as libraries and theaters. Street Connectivity. Street connectivity was calculated based on
the number of intersections with 3 or more connections. The available data included pedestrian trails and bike paths, the importance of which was stressed by Chin et al.18 Motorways and expressways with no access for bikes or pedestrians were excluded. To avoid double intersections (eg, 2 3-way intersections near each other or double tracked roads), 10 m buffers around each intersection were created and overlapping buffers were merged, and a new buffer centroid was used as new intersection point, as described in Forsyth et al.19
Covariates Parental educational level was assessed by parental self-report using the International Standard Classification of Education (ISCED,20) and this was used as indicator for family-socioeconomic-status (family SES). To reduce problems with an insufficient number of participants in some categories, the ISCED scale was recoded into 3 categories (low: ISCED 0–2, < 10 yrs of education; middle: ISCED 3–4, 10–12 yrs of education; high: ISCED 5–7, > 12 yrs of education). If there were differences in educational level between parents, the highest level of the 2 was used in the analysis. Adiposity was expressed as the sum of 4 skinfold measurements from the biceps, triceps, subcapsular, and suprailiac. Skinfold thickness was measured with a Harpenden caliper. Registration of relocation of the participants was done using their home addresses at baseline and follow-up (relocated/not relocated).
JPAH Vol. 12, No. 7, 2015
Built Environment as Determinant of PA 911
Downloaded by Washington Univ In St Louis on 09/19/16, Volume 12, Article Number 7
Statistical Analyses Descriptive statistics were calculated for both included and excluded participants. Exclusion criteria were 1) drop-outs, 2) drop-ins, and 3) incomplete or invalid data on exposure or outcome measures. A multivariable analysis of variance was used to evaluate the association between PA and MI (and each of the 4 MI components) at baseline and follow-up, respectively, and to examine the effect of changes in MI from baseline to follow-up on PA at follow-up. The association between MI and PA at baseline and follow-up, respectively, were adjusted for differences in gender, and secondarily for family SES and adiposity. The analysis of the associations between changes in MI and PA at follow-up was also adjusted for baseline MI. Beta-values (the slopes) with 95% confidence intervals are reported as well as standardized beta-values. The standardized beta-values indicate the strength of the associations (ie, how many standard deviations PA will change) for each standard deviation increase in the MI variable. Manual inspections of plots of outcomes against exposure and inspection of components-plus-residual plots did not reveal any nonlinear relationships. Variance inflation factors did not reveal any signs of collinearity between covariates. All analyses were checked for modification by gender (P < .05 for interaction). All statistical analyses were performed in STATA 11.2 (STATA Corp. Fort Valton TX) with a 5% significance level.
Results
gender differences were observed in ΔMI. A significant difference in the mean ΔMI from baseline to follow-up was observed between participants who relocated [ΔMI = 3.22 (SE = 3.78)], compared with those who did not [ΔMI = 0.15 (0.13)]. However, no differences were observed in PA, family SES, or adiposity between the 2 groups (P > .1) (data not shown). There were also no differences between the included or excluded participants at baseline (data not shown).
Cross-Sectional Associations Table 2 describes the cross-sectional association between MI (and each of the 4 MI components) and PA at baseline and follow-up. At baseline, MI was positively associated to PA. An increase in MI of 1 unit leads to a 0.15 units increase in PA (P = .036). No gender modification was observed (P > .1). The association was stronger in females compared with males. Of the MI components only the number of intersections was significantly associated to PA at baseline (P = .003). At follow-up, modification by gender was observed (P = .003 for interaction). MI was significantly associated to PA for females (P = .02). No associations between MI or its components and PA were observed in males at follow-up (P > .1) expect for the number of intersections (P = .012). In females, the number of daily destinations was significantly associated to PA (P = .008). Further adjustment for family SES and adiposity did not change the associations substantially.
Population Characteristics
Longitudinal Associations
Characteristics of the participants at baseline and follow-up are presented in Table 1, stratified by gender. On average MI increased 2.4 units from baseline to follow-up for males and 2.7 units for females. PA decreased from 416 CPM to 343 CPM for males, and from 357 CPM to 313 CPM for females. PA for males was significantly higher than for females at baseline, whereas adiposity was significantly higher for females at both baseline and follow-up. No
Table 3 shows the associations between the change in MI and the MI-components from baseline to follow-up (ΔMI) and PA at followup adjusted for baseline MI. In females, an increase of 1 MI unit was associated with a 0.27 unit higher PA level at follow-up (P = .02). An increase in daily destinations from baseline to follow-up was significantly associated with follow-up PA (β = 0.32, P = .008). Including SES and adiposity (model 2) did not change the
Table 1 Population Characteristics at Baseline (2003–2004) and Follow-up (2009–2010), Mean (SD), Percentage (%) and Median (Q1; Q3) 2003–2004 Males
2009–2010 Females
Males
Females
Age
15.7 (0.3)
15.7 (0.4)
21.8 (0.3)
21.7 (0.4)
BMI
20.8 (2.4)
21.3 (3.0)
24.3 (3.6)
28.8 (4.1)
Adiposity (mm)
32.3 (16.5)
53.2 (18.1)*
54.0 (28.5)
73.4 (27.7)*
11/37/52
5/45/50
–
–
Family SES (% low/middle/high) Physical activity (CPM)
416 (160)
357 (139)*
343 (130)
313 (122)
Daily MVPA (minutes)
25.1 (17.7 to 38.2)
16.8 (9.9 to 28.4)*
18.3 (9.1 to 27.1)
13.8 (6.6 to 23.0)*
Daily wear time (hours)
14.3 (1.1)
14.2 (1.1)
14.2 (1.4)
14.0 (1.2)
Accepted days Movability Index Intersections
4.6 (0.8)
4.6 (0.8)
4.6 (0.9)
4.6 (1.0)
–1.59 (2.57)
–1.11 (2.48)
0.84 (3.10)
1.62 (3.06)
–0.50 (0.97)
–0.30 (0.97)
0.25 (0.95)
0.47 (0.86)
Residential density
–0.41 (–0.76 to 0.26)
–0.33 (–0.68 to 0.38)
0.01 (–0.56 to 0.86)
0.37 (–0.23 to 1.00)
Daily destinations
–0.48 (–0.50 to –0.24)
–0.44 (–0.50 to 0.01)
–0.30 (–0.64 to 1.10)
0.18 (–0.52 to 1.19)
–0.13 (1.12)
–0.04 (1.00)
0.02 (0.97)
0.13 (0.94)
Recreational facilities *P < .05 for gender differences
JPAH Vol. 12, No. 7, 2015
912 Schipperijn et al
Table 2 Cross-sectional Associations Between Movability Index and Physical Activity at Baseline and Follow-up Baseline
Follow-up*
Estimate (SE) β (95% CI)
Exposure
Estimate (SE) Std. β
P
β (95% CI)
Std. β
P
All (N = 177)a Movability Index Intersections
9.54 (0.65 to 18.44)
0.15
0.036
1.41 (–4.70 to 7.52)
0.03
0.65
35.56 (12.33 to 58.79)
0.22
0.003
–4.59 (–25.52 to 16.14)
–0.03
0.66
Residential density
15.23 (–7.53 to 38.00)
0.10
0.19
3.56 (–12.33 to 19.45)
0.03
0.66
Daily destinations
–1.30 (–24.38 to 21.78)
–0.01
0.08
11.73 (–4.36 to 27.82)
0.11
0.15
Recreational facilities
19.23 (–2.02 to 40.49)
0.13
0.61
–4.01 (–23.75 to 15.73)
–0.03
0.69
Movability Index
12.01 (0.92 to 23.09)
0.21
0.034
9.34 (1.58 to 17.10)
0.23
0.019
Intersections
37.52 (8.50 to 66.54)
0.25
0.012
27.05 (–0.74 to 54.86)
0.19
0.056
Downloaded by Washington Univ In St Louis on 09/19/16, Volume 12, Article Number 7
Female (N = 101)
Residential density
20.22 (–6.42 to 46.86)
0.15
0.14
17.41 (–2.73 to 37.57)
0.17
0.089
Daily destinations
11.08 (–15.32 to 37.47)
0.08
0.41
26.77 (6.99 to 46.55)
0.26
0.008
Recreational facilities
22.68 (–4.96 to 50.33)
0.16
0.11
5.07 (–20.96 to 31.11)
0.04
0.70
6.48 (–8.30 to 21.26)
0.10
0.39
–8.89 (–18.35 to 0.63)
–0.21
0.067
Male (N = 76) Movability Index Intersections
33.12 (–5.45 to 71.71)
0.20
0.091
–39.25 (–69.50 to –9.00)
–0.29
0.012
Residential density
15.58 (–18.12 to 49.28)
0.11
0.36
–15.21 (–45.96 to 15.52)
–0.11
0.33
Daily destinations
–26.09 (–69.71 to 17.54)
–0.14
0.24
–10.83 (–37.55 to 15.87)
–0.09
0.42
Recreational facilities
15.58 (–18.12 to 49.28)
0.11
0.36
–15.22 (–45.95 to 15.12)
–0.11
0.33
a Adjusted
for gender. * P = .003 for gender interaction.
associations substantially, but decreased the p-values slightly. No significant association was observed between a change in MI and PA at follow-up in men (P > .1). For intersection density, a negative association was significant and an increase of 1 unit in the number of intersections from baseline to follow-up was associated with a decrease of 0.34 units in PA at follow-up (P = .02). Further adjustment for family SES and adiposity (model 2) did not substantially alter the associations. Further adjustment for baseline PA level (data not shown) did not change the observations noticeably. As relocation affected the PA level, the longitudinal analyses were adjusted for relocation status (relocated/not relocated), but this did not change the associations.
Discussion The main strengths of this study were the novel combination of a longitudinal study design and the use of objectively measured PA data, as well as detailed measured built environment characteristics. The main findings of the study were that objectively measured PA and MI were associated at both baseline and follow-up, which is similar to findings from cross-sectional studies (eg21–24). Furthermore, an increase in MI from baseline to follow-up was, for females, associated with a reduced decrease in PA at follow-up. These findings indicate that the residential environment people live in is a casual determinant for their PA behavior. The results show a distinct gender difference in the association between ∆MI and PA which is surprising as there were no
significant differences in MI or PA at follow-up for males and females. This difference is most likely explained by the fact that there was no association between MI and PA for males at follow-up, while this association was present for females. A reason for these gender differences could be that PA in males is influenced by other built environment factors than in females, or that PA in males is less influenced by environmental factors. The results imply gender related effect modification, as also suggested by Boone-Heinonen and Gordon-Larsen,8 who in a longitudinal study found that a higher street connectivity was associated with slightly less PA for women, but not for men.8 In a longitudinal study by Lee et al,10 no relation was found between PA for men and urban sprawl, however few men in this study relocated, and the urban sprawl measure used was rather crude. The gender difference deserves more attention in future studies. Most of the changes in MI in this study were due to participants relocating (79.6% moved to a different address), primarily to go studying, and for that reason, the change in PA behavior could perhaps also be related to other changes in the respondents’ lives occurring simultaneously. Moving away from the parental home to start an independent life is a big change for many young adults and it cannot be ruled out that other factors related to relocation could explain our results. Then again, we feel that it is not unreasonable to assume that the environment young adults relocate to, and in particular the number of opportunities that environment offers for a more active lifestyle, has some effect on their behavior. This is supported by fact that adjustment for relocation did not change the
JPAH Vol. 12, No. 7, 2015
Built Environment as Determinant of PA 913
Table 3 Association Between the Change in Movability Index From Baseline to Follow-up and Adult Physical Activity Model 1
Model 2
Estimate (SE) Exposure
β (95% CI)
Estimate (SE) Std. β
P
β (95% CI)
Std. β
P
All (N = 177) ∆ Movability Index
3.02 (–3.14 to 9.19)
0.09
0.35
3.71 (–2.55 to 9.96)
0.11
0.35
Intersections
–1.71 (–22.84 to 19.41)
–0.02
0.87
1.07 (–20.43 to 22.52)
0.01
0.92
Residential density
5.59 (–10.46 to 21.64)
0.06
0.49
7.10 (–9.23 to 23.43)
0.08
0.39
Daily destinations
14.31 (–1.88 to 30.50)
0.16
0.08
16.85 (0.42 to 33.28)
0.19
0.045
Recreational facilities
5.47 (–15.75 to 26.67)
0.05
0.61
4.04 (–17.50 to 25.58)
0.04
0.71
Females (N = 101) Downloaded by Washington Univ In St Louis on 09/19/16, Volume 12, Article Number 7
∆ Movability Index
9.27 (1.33 to 17.18)
0.27
0.02
10.15 (2.08 to 18.21)
0.30
0.014
Intersections
24.45 (–3.92 to 52.82)
0.23
0.09
28.24 (–0.80 to 57.27)
0.26
0.06
Residential density
16.34 (–4.06 to 36.74)
0.20
0.12
19.06 (–1.78 to 39.90)
0.23
0.07
Daily destinations
27.40 (7.20 to 47.61)
0.32
0.008
31.24 (10.64 to 51.84)
0.37
0.003
Recreational facilities
8.70 (–19.60 to 37.00)
0.08
0.54
5.57 (23.87 to 35.01)
0.05
0.71
–7.75 (–17.39 to 1.88)
–0.22
0.11
–7.55 (–17.46 to 2.36)
–0.21
0.13
Intersections
–36.40 (–67.15 to –5.65)
–0.34
0.02
–35.47 (–67.10 to –3.83)
–0.33
0.03
Residential density
–11.00 (–36.94 to 14.94)
–0.11
0.40
–10.43 (–36.81 to 15.95)
–0.11
0.43
Daily destinations
–8.11 (–35.01 to 18.79)
–0.08
0.55
–6.56 (–33.92 to 20.80)
–0.07
0.63
Recreational facilities
–11.55 (–44.89 to 21.80)
–0.11
0.49
–13.28 (–47.63 to 21.08)
–0.12
0.44
Males (N = 76) ∆ Movability Index
Note. Model 1 is adjusted for baseline movability index and gender. Model 2 is further adjusted for family SES and adiposity.
associations. New habits are created when starting a new life stage and, in theory, these habits are easier to establish if both the physical and social environment are supportive of them. A limitation of the study was that it was not possible to distinguish between PA in different domains (such as walking, recreation, or transport) based on accelerometer data alone, and this also meant that it was not possible to address the relative importance of the different settings for different persons. The same mean CPM could stem from activity in very different settings. Furthermore, since only overall PA could be assessed, an environmental index that describes the environmental suitability for all components of PA needed to be developed. The composite nature of the MI meant that environmental factors positively associated with one type of activity could be negatively associated with activity in another domain. The lack of domain specific PA data as well as domain specific environmental data might have led to reduced overall associations. In addition, as the observed gender differences potentially indicate the association between the environment and PA might be different for different people, suggesting that the environmental factors included in an index might need to be differentiated by gender, age, or other sociodemographic factors. In their conceptual model for longitudinal studies of the obesogenic environment, Boone-Heinonen and Gordon-Larsen9 suggested the need for more focus on the interactions between respondents’ sociodemographic characteristics and the environmental characteristics of the neighborhood they live in. Another limitation of the study is the relatively small sample size, as well as the fact that female respondents from families with a lower SES were less likely to participate in the follow-up study;
however, there were no significant differences in PA, BMI, or adiposity between included and excluded participants. Furthermore, there might have been an element of positive self-selection to participate in the EYHS study, where more active participants might have been more likely to participate. For future studies, a setup in which accelerometer measurements are combined with GPS measurements could be used to distinguish between PA in different settings/domains. In combination with domain-specific, and perhaps also gender-specific, environmental indexes, this will increase the conceptual validity of the measures.
Conclusion In summary, this study is among the first studies looking at the longitudinal relation between changes in objectively assessed built environment with changes in objectively measured physical activity. Even though the measures can and should be further improved in the future, the findings indicate that built environment indeed plays a role in explaining individual differences in physical activity levels, in particular among females. The gender differences need to be further explored. Acknowledgments We would like to thank the Danish EYHS team for their support, and we would like to thank all the participants who contributed to the study. The Danish part of the European Youth Heart Study was funded by the Danish Council for Strategic Research (grant number 2101-08-0058); The Danish
JPAH Vol. 12, No. 7, 2015
914 Schipperijn et al
Heart Foundation, the Danish Health Fund, the Danish Ministry of the Interior and Health, TrygFonden, the Danish Medical Research Council, and the Ministry of Culture Council for Sports Science Research.
Downloaded by Washington Univ In St Louis on 09/19/16, Volume 12, Article Number 7
References 1. Lee IM, Shiroma EJ, Lobelo F, et al. Effect of physical inactivity on major non-communicable diseases worldwide: an analysis of burden of disease and life expectancy. Lancet. 2012;380(9838):219–229. PubMed doi:10.1016/S0140-6736(12)61031-9 2. Hallal PC, Bauman AE, Heath GW, Kohl HW, 3rd, Lee IM, Pratt M. Physical activity: more of the same is not enough. Lancet. 2012;380(9838):190–191. PubMed doi:10.1016/S01406736(12)61027-7 3. Bauman AE, Reis RS, Sallis JF, et al. Correlates of physical activity: why are some people physically active and others not? Lancet. 2012;380(9838):258–271. PubMed doi:10.1016/S01406736(12)60735-1 4. Wendel-Vos W, Droomers M, Kremers S, Brug J, van Lenthe F. Potential environmental determinants of physical activity in adults: a systematic review. Obes Res. 2007;8(5):425–440. 5. Ding D, Sallis JF, Kerr J, Lee S, Rosenberg DE. Neighborhood environment and physical activity among youth a review. Am J Prev Med. 2011;41(4):442–455. PubMed doi:10.1016/j.amepre.2011.06.036 6. Frank LD, Sallis JF, Saelens BE, et al. The development of a walkability index: application to the Neighborhood Quality of Life Study. Br J Sports Med. 2010;44(13):924–933. PubMed doi:10.1136/ bjsm.2009.058701 7. Buck C, Pohlabeln H, Huybrechts I, et al. Development and application of a moveability index to quantify possibilities for physical activity in the built environment of children. Health Place. 2011;17(6):1191– 1201. PubMed doi:10.1016/j.healthplace.2011.08.011 8. Boone-Heinonen J, Gordon-Larsen P. Life stage and sex specificity in relationships between the built and socioeconomic environments and physical activity. J Epidemiol Community Health. 2011;65(10):847– 852. PubMed doi:10.1136/jech.2009.105064 9. Boone-Heinonen J, Gordon-Larsen P. Obesogenic environments in youth: concepts and methods from a longitudinal national sample. Am J Prev Med. 2012;42(5):e37–e46. PubMed doi:10.1016/j. amepre.2012.02.005 10. Lee IM, Ewing R, Sesso HD. The built environment and physical activity levels: the Harvard Alumni Health Study. Am J Prev Med. 2009;37(4):293–298. PubMed doi:10.1016/j.amepre.2009.06.007 11. Coogan PF, White LF, Adler TJ, Hathaway KM, Palmer JR, Rosenberg L. Prospective study of urban form and physical activity in the Black Women’s Health Study. Am J Epidemiol. 2009;170(9):1105–1117. PubMed doi:10.1093/aje/kwp264 12. Riddoch C, Edwards D, Page A, et al. The European Youth Heart study—cardiovascular disease risk factors in children: rationale,
aims, study design, and validation of methods. J Phys Act Health. 2005;2(1):115–129. 13. Kristensen PL, Moller NC, Korsholm L, Wedderkopp N, Andersen LB, Froberg K. Tracking of objectively measured physical activity from childhood to adolescence: the European youth heart study. Scand J Med Sci Sports. 2008;18(2):171–178. PubMed doi:10.1111/j.16000838.2006.00622.x 14. Kristensen PL, Korsholm L, Moller NC, Wedderkopp N, Andersen LB, Froberg K. Sources of variation in habitual physical activity of children and adolescents: the European youth heart study. Scand J Med Sci Sports. 2008;18(3):298–308. PubMed doi:10.1111/j.16000838.2007.00668.x 15. Schipperijn J, Stigsdotter UK, Randrup TB, Troelsen J. Influences on the use of urban green space—a case study in Odense, Denmark. Urban for Urban Gree. 2010;9(1):25–32. doi:10.1016/j. ufug.2009.09.002 16. Giles-Corti B, Broomhall MH, Knuiman M, et al. Increasing walking: how important is distance to, attractiveness, and size of public open space? Am J Prev Med. 2005;28(2, Suppl 2):169–176. PubMed doi:10.1016/j.amepre.2004.10.018 17. Norman GJ, Nutter SK, Ryan S, Sallis JF, Calfas KJ, Patrick K. Community design and access to recreational facilities as correlates of adolescent physical activity and body-mass index. J Phys Act Health. 2006;3(supplement):s118–s128. 18. Chin GK, Van Niel KP, Giles-Corti B, Knuiman M. Accessibility and connectivity in physical activity studies: the impact of missing pedestrian data. Prev Med. 2008;46(1):41–45. PubMed doi:10.1016/j. ypmed.2007.08.004 19. Forsyth A, D’Sousa E, Koepp J, et al. Twin Cities Walking Study— environment and physical activity: GIS protocols version 4.1, June 2007. 2007; http://www.designforhealth.net/pdfs/GIS_Protocols/ MinnGIS_Ver4_1_FINAL.pdf. 20. UNESCO. ISCED: International Standard Classification of Education. 1997; www.uis.unesco.org/Education/Pages/international-standardclassification-of-education.aspx 21. De Meester F, Van Dyck D, De Bourdeaudhuij I, Deforche B, Sallis JF, Cardon G. Active living neighborhoods: is neighborhood walkability a key element for Belgian adolescents? BMC Public Health. 2012;12:7. PubMed doi:10.1186/1471-2458-12-7 22. Owen N, Cerin E, Leslie E, et al. Neighborhood walkability and the walking behavior of Australian adults. Am J Prev Med. 2007;33(5):387–395. PubMed doi:10.1016/j.amepre.2007.07.025 23. Sallis JF, Saelens BE, Frank LD, et al. Neighborhood built environment and income: examining multiple health outcomes. Soc Sci Med. 2009;68(7):1285–1293. PubMed doi:10.1016/j.socscimed.2009.01.017 24. Van Dyck D, Cardon G, Deforche B, Sallis JF, Owen N, De Bourdeaudhuij I. Neighborhood SES and walkability are related to physical activity behavior in Belgian adults. Prev Med. 2010;50(Suppl 1):S74– S79. PubMed doi:10.1016/j.ypmed.2009.07.027
JPAH Vol. 12, No. 7, 2015
