Journal of Physical Activity and Health, 2014, 11, 1070 -1077 http://dx.doi.org/10.1123/jpah.2012-0159 © 2014 Human Kinetics, Inc.
Official Journal of ISPAH www.JPAH-Journal.com ORIGINAL RESEARCH
Neighborhood Environment and Physical Activity Among Older Women: Findings From the San Diego Cohort of the Women’s Health Initiative Jacqueline Kerr, Greg Norman, Rachel Millstein, Marc A. Adams, Cindy Morgan, Robert D. Langer, and Matthew Allison Background: Few studies of older adults have compared environmental correlates of walking and physical activity in women who may be more influenced by the environment. Environmental measures at different spatial levels have seldom been compared. Findings from previous studies are generally inconsistent. Methods: This study investigated the relationship between the built environment and physical activity in older women from the Women’s Health Initiative cohort in San Diego County (N = 5401). Built environment measures were created for 3 buffers around participants’ residential address. Linear regression analyses investigated the relationship between the built environment features and self-reported physical activity and walking. Results: Total walking was significantly positively associated with the walkability index (β = .050: half-mile buffer), recreation facility density (β = .036: 1-mile buffer), and distance to the coast (β = –.064; P-values < .05). Total physical activity was significantly negatively associated with distance to the coast and positively with recreation facility density (β = .036: 1-mile buffer; P < .05). Conclusions: Although effect sizes were small, we did find important relationships between walkability and walking in older adults, which supports recommendations for community design features to include age friendly elements. More intense physical activity may occur in recreational settings than neighborhood streets. Keywords: urban form, walking, exercise The purpose of this study was to assess the relationship between the built environment and physical activity in older adult women. We explored the impact of different environmental features on different types of physical activity and at different environmental scales to see if we could explain some of the inconsistent findings in the current literature. Physical inactivity is the 4th leading contributor to death in the U.S.1 and older adults are the least active and most sedentary group2 in the United States (US) population. Recent studies using objective measures found that only 2.5% of older adults meet physical activity (PA) guidelines.3 There are strong and well established relationships between PA and mortality, morbidity, physical, mental, emotional and cognitive functioning in older adults.4 Ecological Models of Behavior Change suggest that environmental influences on health should be considered in addition to individual and interpersonal factors.5 Even while conclusive evidence is lacking, multiple US agencies and the World Health Organization (WHO) have recommended built environment changes to increase PA for whole populations. In adults, more walkable environments with short and straight street blocks, grid like street design (versus cul-de-sacs), many destinations, and high residential density, have been strongly associated with walking for transportation and less consistently associated with total physical activity or walking for leisure.5,6 For older adults, authoritative groups including the WHO age-friendly cities, Environment Protection Agency aging initiative, Kerr (
[email protected]), Norman, Millstein, Adams, Morgan, and Allison are with the Dept of Family and Preventive Medicine, University of California, San Diego, San Diego, CA. Langer is with the Dept of Preventive Medicine, Jackson Hole Center for Preventive Medicine, Jackson Hole, WY. 1070
and American Association of Retired Persons have emphasized the importance of environmental changes to support walking in older adults. Theories from environmental gerontology and international classifications of functioning suggest that older adults may be more susceptible to environmental barriers to activity.7–9 For example, Lawton’s person environment fit model7 suggests that as we lose functioning the importance of our “fit” with the environment increases (ie, if our functioning does not allow us to overcome environmental barriers then we are mismatched with our environment and this leads to further loss of functioning as our mobility becomes limited). Some studies have shown an interaction between the built environment and physical functioning in older adults. While the number of studies of the built environment and PA in older adults has increased in the last,8,10 findings have been inconsistent.11–14 In older adults, 3 out of 5 studies that compared urban (more walkable) versus rural neighborhoods found no association with walking or physical activity.15–17 In 6 studies of objectively measured walkability and walking, only 3 showed significant findings in the expected direction.15,18,19 Two studies found that walkability was related to walking for transportation but not to moderate-to-vigorous activity.12,20 Relationships for other built environment features such as parks and recreation centers have also been inconsistent. One possibility for inconsistent findings may be that studies operationalized neighborhoods differently, with differing buffer types and sizes around individuals’ residential addresses. However, the scale at which the built environment most influences older adults remains uncertain.11–13,21 In addition, different PA outcomes were employed, which also has led to differences in results. This paper presents data on 5401 women from the San Diego Women’s Health Initiative Cohort with the aim of examining the association between built environment and physical activity in
Built Environment and Physical Activity 1071
a cohort of postmenopausal women. Women, particularly older women, are an important population to study because of their prevalence and because studies suggest a stronger relationship between built environment features and physical activity in women than men.5,11 While there have been some studies in older women,17,22–24 they have used smaller samples, and have not used GIS-based measures of walkability at the residential address level employed in this study. Although perceptions of the environment are important in older adults, especially for factors like safety, GIS measures of walkability allow us to assess the environment at different scales and using objective measures that are not susceptible to problems of bias and recall. This study aimed to assess built environment features in 3 different sized buffers around participants’ homes to investigate whether current inconsistent findings could be explained by the use of different buffer distances. We also explored different physical activity outcomes, again to try to cast light on possible reasons for different findings in previous studies. Finally, we investigated an interaction effect for physical functioning found in previous studies and supported by theories of aging.
Methods Sample and Study Design Participants were derived from the San Diego clinical center of the Women’s Health Initiative (WHI), a large NIH-funded multicenter study of the predictors of morbidity and mortality among postmenopausal women25 who were followed for an average of 7 years. A complete description of the methodology and findings from the WHI study are available elsewhere.26 Briefly, women aged 50 to 79 years old and living in San Diego County were recruited between 1993 and 1998. At the time of enrollment, potential subjects were excluded if they did not plan to reside in the area for at least 3 years, had medical conditions predictive of survival less than 3 years, or had complicating conditions such as alcoholism or drug dependency. A total of 5626 women were enrolled in the San Diego site of the WHI. The current study used baseline measures for all study arms. The institutional review board at the University of California, San Diego approved the study.
Measures Geocoding Residential Addresses. Built environment measures
were created using geographic information systems (GIS). ArcGIS 9.3 (ESRI; Redlands, CA) software was used to accomplish these tasks. A total of 5401 (96.0%) of 5626 enrolled WHI women had valid residential addresses within San Diego County that could be geocoded to the residential level. Of the 225 addresses that could not be geocoded, the reasons included P.O. Box address only (n = 125), address outside of San Diego County (n = 79) or California (n = 14), incorrect address (n = 5), or address missing (n = 2).
Creating Residential Buffers. Half-mile, 1-mile and 3-mile street network buffers around residences were created to represent individuals’ neighborhoods. The rationale for multiple buffers was that the appropriate neighborhood scale is unknown for older adults, and built environment associations with outcomes may vary by either type of environmental measure14 or participants’ characteristics (eg, age).21 For example adult residents may travel several miles to go to recreation facilities or parks,27 but residents may only walk 20 minutes to shops (ie, 1 mile).
Creating GIS Environment Variables. Spatial data on existing
land use was obtained from the San Diego Association of Governments (SANDAG) for the year 2000. Various land use codes were aggregated into residential, commercial, office, and institutional land use categories. To calculate acreages of specific land uses within each network buffer, we clipped the various land use polygons so that they matched the individual resident buffer polygon boundaries, then calculated new acreages for the clipped land uses, and finally summed the clipped land use acreages by land use type and by buffer.
Neighborhood Walkability Index. The walkability of participant’s
residential environment was calculated using a walkability index developed by Frank at al.28 based on principles of transportation and urban planning that are commonly used in built environment studies. The walkability index employed here includes 3 components—land use mix, residential density, and street network connectivity. Land use mix was calculated as the evenness of the distribution of acreage of residential, commercial, office, and institutional land use. As land use approaches a value of 0, land uses become less evenly distributed (eg, the buffer is 100% residential land use). A value of 1 indicates equal distribution of the 4 land use types in the buffer. Residential density is the number of housing units per acre of residential land use within each network buffer. Housing unit values were obtained and apportioned from the 2000 US Census at the block group level to our buffers and residential land use was obtained from SANDAG’s 2000 Land Use shapefile. Intersection density per acre, using data from SANDAG’s 2003 Road shapefile, was calculated by summing the number of intersections with 3 or more legs in the buffer. The walkability index was calculated as the sum of z-scores for land use mix, residential density, and intersection density. Greater mixed use, higher residential density and greater intersection density have been related to overall physical activity and walking for specific purposes such as transportation.12
Recreation Facilities Density. Access to recreation facilities has been related to physical activity in multiple population groups.29 Locations of recreation facilities in San Diego County were obtained from another study.30 Facility addresses were geocoded to the street level, and the density of recreation facilities was computed by summing the number of facilities within each participant’s buffer and dividing by the buffer acreage. For the half-mile buffer, 73.3% (3961) had no facilities in the buffer. As result, the count of recreation facilities in the half-mile buffer was coded as 0 for no facilities, and 1 for 1–16 facilities. Distance to the Coast. One previous study investigated distance
to the coast and found that it was related to physical activity.31 Coastal beaches in San Diego County provide a safe and pleasant environment for walking. Distance to the coast in feet was calculated along the street network and coded as deciles for analyses. Distance to the Nearest Park. Access and proximity to parks is
related to physical activity in multiple population groups.29 Parks were identified from county lists and supplemented with official city websites, lists from several park and recreation departments, and local maps. Parks were geocoded using addresses or cross-streets approximating a point of park entry. The distance to the nearest park of any size was computed by deriving the shortest distance via the street network. Distance to the nearest park along the street network (regardless of the buffer size) was calculated in feet and categorized into deciles for the analyses.
Physical Activity Measures. At the baseline WHI visit, women
completed the WHI PA questionnaire.32 Questions on recreational
1072 Kerr et al
walking, moderate recreational PA, and strenuous recreational PA had high test-retest reliability (weighted kappa range = 0.50 to 0.60). They were asked how often each week they usually did ‘strenuous or very hard exercise,’ ‘moderate exercise,’ and ‘mild exercise.’ A description and activity examples for each category were provided. For each intensity of exercise, frequency of exercise was assessed with response options of: none, 1, 2, 3, 4, and 5 or more days per week. If they were physically active at least 1 day per week, duration of an exercise session was asked as, ‘how long did you usually exercise at that level at one time’ with response options of: less than 20 min., 20–39 min., 40–59 min., or 1 hour or more. Values for the midpoints of duration were imputed as 10, 30, 50, and 70 minutes, respectively. Metabolic equivalent (MET) values for physical activity intensity and type of walking were assigned as: strenuous activity = 7, moderate activity = 4–5, and mild activity = 3 following other studies that have used WHI physical activity data.32 Women were also asked about the walking they did outside the home for more than 10 minutes without stopping in terms of frequency, duration, and intensity (ie, casual strolling, average or normal, fairly fast, very fast, or don’t know). In the current study, responses of ‘don’t know’ were recoded as ‘casual strolling.’ MET values for walking were assigned as: very fast walking = 5, fairly fast walking = 4, and average normal walking = 3. Three physical activity estimates were calculated as MET/ hrs/wk. Total walking was calculated as MET hours/week for all walking intensities. Moderate to vigorous physical activity (MVPA) was calculated as MET/hrs/wk of strenuous and moderate activity. Total activity (walking and MVPA) was the sum of MET/hrs/wk of activities greater than 3 MET units: strenuous, moderate, mild, average normal walking, fairly fast walking, and very fast walking. Individual Characteristics. Demographic and health status information included age, education, ethnicity, and family income. General health status was determined from a question asking women to rate their general health on a 5-point scale with responses of poor, fair, good, very good, and excellent.
75% identified themselves as White-Non-Hispanic. About 36% of women reported annual household incomes greater than $50,000. The majority of the sample was married (58%) and had an education greater than 12th grade (68%). Most women (88%) perceived their general health to be good to excellent. Compared with the geocoded sample, the 225 women whose addresses were not geocoded were more likely to be White (P = .005), more likely to have a family income greater than $100K (P = .035), and were more likely to have a high school education (P = .041). Geocoded and nongeocoded women were equivalent on age, marital status, and general health status. Table 1 Characteristics of 5401* With Geocoded Residences in the San Diego Cohort of the Women’s Health Initiative N
%
50–59
1667
30.9
60–69
2258
41.8
70–79
1474
27.3 (2 missing)
White Non-Hispanic
4059
75.4
Non-White
1325
24.6 (17 missing)
