FEATURES, AND SOCIAL CAPITAL IN THE RURAL SETTING OF COLORADO’S SAN LUIS VALLEY by
JULIE ANNA RODRIGUEZ
B.S., Metropolitan State University of Denver, 2013 B.A., University of Florida, 2004
A thesis submitted to the Faculty of the Graduate School of the University of Colorado in partial fulfillment
of the requirements for the degree of Master of Science
Epidemiology Program 2015
This thesis for the Master of Science degree by Julie Anna Rodriguez
has been approved for the Epidemiology Program
by
Julie A. Marshall, Chair Jini Puma, Advisor
Tessa Crume
Rodriguez, Julie Anna (M.S. Epidemiology)
Associations Between Physical Activity, Neighborhood Built Environment Features, and Social Capital in the Rural Setting of Colorado’s San Luis Valley
Thesis directed by Clinical Assistant Professor Jini Puma. ABSTRACT
Background: Chronic disease prevalence in the San Luis Valley of Colorado is
disproportionately greater than the rest of the state. The benefits of physical activity (PA) on the prevention and successful management of chronic disease has been well demonstrated in the literature. Studies demonstrating relationships between the built environment (BE) and PA, and social capital and PA have largely focused on urban centers, with rural area studies being few and in some cases contradictory. This study will investigate the relationship between physical activity and objective measures of the built environment on a respondent’s street segment, self-reported access to environments for exercise in a respondent’s neighborhood, and social capital.
Methods: The 2010 San Luis Valley Community Health Survey (n=1187) is a
comprehensive cross-sectional survey about health and behaviors which included a street audit of the respondent’s street segment for BE features. Self-reported demographics, PA, and perceived access to environments for exercise were taken from the survey and linked to their respective street audits. Self-reported perceived access to BE features for exercise were compared against similar items from the street audits to determine whether they were in agreement. Multiple linear regression and logistic regression were performed to determine the relationship between the respective PA outcomes of total minutes of moderate and vigorous PA per week, and meeting PA recommendations with BE and social capital variables.
Results: Higher levels of PA was associated with integration of natural land (p<0.01), access to parks and trails in neighborhoods (p=0.04), and social capital (p=0.02). A higher likelihood of meeting CDC physical activity guidelines was associated with access to parks and trails in neighborhoods (p=0.04), and access to public exercise facilities (p=0.02). Lower levels of PA and a lower likelihood of meeting PA guidelines was associated with street characteristics for walking and biking (p<0.01), and recreational facilities (p<0.01). These results indicate there is a relationship between BE and PA in rural areas. Better assessment of what BE features are relevant to rural residents and a better definition of “neighborhood” is needed to further elaborate on the relationship between BE and PA in a rural setting.
The form and content of this abstract are approved. I recommend its publication. Approved: Jini Puma
ACKNOWLEDGEMENTS
I would like to acknowledge that the data for this study is a product of the work of researchers and volunteers at the Rocky Mountain Research Prevention Center in the San Luis Valley. I am grateful for being permitted to utilize the data for this study. In addition, I would like to acknowledge Sharon Scarbro for all of her work on the data before the start of this study and for all of her gracious assistance during the course of this study. I am grateful for her kind support.
TABLE OF CONTENTS CHAPTER
I. INTRODUCTION ... 1
Literature on the Built Environment and Physical Activity in Rural Communities ... 3
Literature on Social Capital and Physical Activity in Rural Communities ... 5
Study Aims ... 7
II. METHODS ... 8
Setting ... 8
Study Design ... 9
Definition of Outcome Variables ... 10
Definition of Built Environment and Social Capital ... 12
Confounders ... 16
Data Analysis ... 16
III. RESULTS ... 19
Descriptive Statistics ... 19
Analysis of Covariates ... 22
Comparison between Street Audits and Self-Report ... 23
Results for the Street Audit Variables ... 25
Results for the Survey Variables ... 28
IV. DISCUSSION ... 32
Implications for Future Research ... 37 REFERENCES ... 39
LIST OF TABLES TABLE
1: Age-Adjusted Prevalence (%) of Chronic Diseases ... 8
2: Physical Activity Survey Questions ... 11
3: Built Environment Variables from Street Audit ... 14
4: Built Environment Variables from the Telephone Survey ... 15
5: Survey Demographics ... 20
6: Descriptive statistics for measures of physical activity ... 20
7: Descriptive statistics for measures of built environment and social capital ... 21
8: Association of Covariates with total minutes of PA ... 22
9: Association of Covariates with meeting PA recommendations ... 23
10: Results for Linear and Logistic Regressions for Street Audit Measures ... 28
LIST OF FIGURES FIGURE
1: Natural History of Chronic Disease Diagram ... 2
2: DAG for Association between Built Environment and Physical Activity ... 17
3: DAG for Association between Social Capital and Physical Activity ... 18
4: Agreement between survey and street audit measures of Parks and Trails ... 24
5: Agreement between survey and street audit measures of exercise facilities ... 24
6: Agreement between survey and street audit measures of walkability ... 25
7: Plot and interpretation of linear model of Natural Land Integration ... 26
8: Plot and interpretation of linear model of Street Characteristics for Walking and Biking ... 27
9: Plot and interpretation of linear model of Recreational Facilities ... 27
10: Plot and interpretation of linear model of Access to Parks and Trails ... 29
11: Plot and interpretation of linear model of Social Capital ... 29
CHAPTER I INTRODUCTION
According to the Centers for Disease Control (CDC), about half of all adults have at least one chronic disease as of 2012, and these diseases caused 7 out of 10 deaths in 2010.1 The high prevalence of chronic disease not only impacts those with the diseases, but it also has an impact on the nation as a whole through increased health care spending. The CDC reports that 86% of all of the health care dollars in the US are spent on the treatment of chronic diseases.2 Thus it is clear that the increasing prevalence of chronic disease in the US is an issue of great public health concern not only in terms of morbidity and mortality, but also in terms of economy. Many chronic conditions including diabetes, coronary heart disease (CHD), hypertension, myocardial infarction and obesity share several common risk factors: one of which is low physical activity levels or a sedentary lifestyle.3,4,1 Regular physical activity has been shown to reduce the risk of cardiovascular disease, diabetes, obesity, some cancers, high blood pressure and
osteoporosis.3,4,1
Given the importance of physical activity, the US Department of Health and Human Services released Physical Activity Guidelines for Americans which recommends that adults should engage in at least 150 minutes of moderate-intensity physical activity, 75 minutes of vigorous-intensity physical activity, or a combination of moderate and vigorous (where vigorous minutes are multiplied by two) totaling at least 150 minutes per week.1 Given the strong beneficial link between physical activity and reduced mortality due to chronic diseases, the US Department of Health and Human Services also included physical activity as a target area in the Healthy People 2020 initiative with a goal to increase the proportion of adults meeting the minimum recommendations from 43.5% to 47.9%.5 It is clear that increasing physical activity levels in the population is a public health goal and as such, research to determine environmental
and social factors associated with higher levels of physical activity is an important step in designing potentially effective interventions.
Figure 1 shows the natural history of a number of chronic diseases in which physical inactivity is a known risk factor. While many exposures lead to a multitude of physiologic and pathologic changes which in turn lead to a number of different diseases, this model only lists a small number of examples to illustrate the relative importance of physical activity in the development or management of some chronic diseases. Exposure to risk factors such as poor diet, physical inactivity, and smoking can impact every level of disease and changes in a person’s exposure can prevent disease from developing, or improve prognosis and quality of life for individuals with disease. The continued impact of exposure to risk factors at every stage of disease status is depicted in the diagram as arrows leading from the risk factors to all parts of the disease process. This study will focus on the environmental and social factors which might impact physical activity and thus they are illustrated in the diagram as well with an arrow going into physical inactivity.
Many rural communities in the US suffer from a disproportionately high prevalence of chronic disease and poor health behaviors.43 Understanding the potential impact of the built environment and social capital on physical activity can help identify specific features that might promote increased physical activity in rural communities. Such research has been conducted extensively in urban settings, however rural areas are underrepresented in the literature and some existing studies have conflicting results.6 More research on the impact that social and environmental factors have on physical activity in rural settings is necessary to help determine if relationships exist.
Literature on the Built Environment and Physical Activity in Rural Communities The built environment is the physical form of communities and includes all built and natural features, taking into account how land is used, transportation infrastructure,
architecture, landscapes, and various amenities and facilities such as parks and commercial structures.7 A growing body of research is showing associations between certain features of the built environment and higher levels of physical activity for residents in proximity to them including mixed land use, presence of light traffic, presence of stores, proximity to trails, bike facilities, sidewalks, access to parks and indoor gyms, high walkability, and less urban sprawl.8–12 However, research regarding the roll of the built environment as it relates to physical activity specifically in rural areas is limited and inconclusive.6,13 In 2010, Frost et al. published a review of studies on the built environment in rural areas and found that while some reported positive relationships between physical activity and neighborhood aesthetics, safety from crime,
recreational facilities, trails, parks, and walkable destinations, others contradicted those findings and many more had null results. To improve our understanding of the relationship between the built environment and physical activity in rural areas, further study which includes objective
measures of street characteristics for walking, recreational facilities, aesthetics, and other objective measures of the built environment is needed.
Street characteristics conducive for walking (e.g. sidewalks and street lighting) and destinations (e.g. services and retail) within walking distance are considered to be aspects of the built environment which are important for physical activity. However, the presence of sidewalks has had mixed, and often null, results in rural settings.14–19 In a study of women in rural Alabama by Sanderson, et al, the absence of sidewalks was associated with increased physical activity.20 A qualitative study by Kegler, et al, summarized results from focus groups in which sidewalks were not important in respondents’ decisions to walk.21 In a study of individuals with diabetes in the rural Midwest, having many places to walk was positively associated with physical activity based on self-reported access to walkable destinations.22 Scanlin, et al developed a more objective measure of walkable destinations in their rural walking tool which measures walkable destinations such as food, retail, and service destinations by direct observation by a trailed auditor.19 However, a significant relationship between walkable destinations and physical activity was not detected in their study. The evidence for sidewalks and walkable destinations is still inconclusive in the rural setting and largely based on self-reported data on the built
environment. More research using direct observation of the built environment as an objective measure is necessary to determine if there is a relationship and the direction of that
relationship.
Usage of recreational and exercise facilities have also shown positive associations with increased physical activity in rural communities, but this does not explore whether proximity or access to these facilities is also related to physical activity.14,17,22 Again, there is a need for an objective measure. Aesthetics (i.e. attractive and pleasant built features on a street) have been shown to have a positive association with physical activity in a number of studies.22–24
Sometimes included in aesthetics, but conceptually separate since not man-made, is the presence of natural features such as bodies of water, and wooded areas. A report published in the American Journal of Public Health by Dannenberg, et al, asserts that there is great value in how land development is planned and the manner in which the built environment preserves and integrates natural land features, allowing for residents to have access to natural land
environments.25 Access to natural land features has been shown to have a positive association with physical activity in rural and urban settings, thus examining that association objectively in a rural setting can strengthen this conclusion.25–27 Therefore, recreational facilities, aesthetics, and natural environments all need to be further defined, objectively measured, and studied in rural settings.
The reliance on self-reported aspects of the built environment is a limitation of many studies, some of which have found discordance between perceived and observational measures which indicates there may potentially be self-reporting bias.28–30 Gebel, et al hypothesizes that persons who are more active in their environment perceive it more favorably than those who are not, even within the same neighborhood.29 The built environmental features researchers choose to measure in rural settings may not be the most relevant as pointed out by Brownson, et al in his 2009 review of the state of the science in measurements of the built environment.7 Therefore, it is important to test new methods and to conduct studies which attempt to objectively measure the built environment in a rural setting into order to advance knowledge and improve practice in this important area of research.
Literature on Social Capital and Physical Activity in Rural Communities
Social capital is defined in a variety of ways in the literature, but all measures generally describe the social environment and the degree of connectedness people have with others in their community. The impact that one’s social environment has on their health behaviors was
recognized by the CDC as an important determinant of health in the 1996 Surgeon General Report on Physical Activity and Health.31 Since that time, several aspects of a person’s social environment have been studied for their potential impact on health behaviors, such as social capital and participation in community groups.
Many studies have found positive associations between social capital and a variety of health behaviors including physical activity.32–37 Addy, et al found a positive association between thinking your neighbors can be trusted and being physically active in a rural southeastern county.14 Davison, et al found a positive association between parent’s levels of social capital and their children’s physical activity level in a rural setting in upstate New York.33 However, Mohnen, et al compared the effect of social capital in rural and urban neighborhoods in the Netherlands and found that increases in social capital had a stronger relationship with physical activity in urban areas than in rural areas.34 Urban residents had lower overall social capital than rural residents but the impact of small increases in social capital was strong in urban residents and null in rural residents.34 Another null result was found in a study of diabetics in the rural
Midwest which found no association between social environmental factors and regular physical activity.22 Thus, further research is needed to evaluate the impact of social capital on rural residents’ physical activity levels.
In addition to the studies on social capital, some studies have looked at participation in community organizations as an indicator of social capital. Participation in community groups has been found to have a positive relationship with physical activity, or lower levels of physical inactivity, in a number of studies in urban settings.32,35 However, literature on community participation as a possible indicator of physical activity in rural settings is limited. One study of women in the rural mid-west found an association between attendance at religious services and
higher reported physical activity than those who did not attend services.16 This too is an area worth investigating further.
Finally, it is important to note that the term “social capital” is used to define many different aspects of the social environment such as trust, reciprocity, cohesion, collective efficacy, familial support, and community involvement.38–40 The lack of a defined metric makes comparisons among studies difficult. McNeill, et al. proposed five different dimensions of the social environment, one of which is “social cohesion and social capital,” defined as “the extent of connectedness and solidarity among groups combined with the willingness to intervene for the common good.”37,41 This definition is most similar to the measure used in this study and has not been thoroughly evaluated in a rural settings.
Study Aims
Utilizing data collected by the Rocky Mountain Prevention Research Center in the 2010 San Luis Valley Community Health Survey (SLVCHS) and subsequent environmental audit of respondent’s street segments, this study aimed to demonstrate the hypotheses that the built environment and social capital have positive associations with physical activity in the San Luis Valley of Colorado. The specific built environmental features studied include: access to businesses and services; street characteristics hypothesized to promote walking and biking; presence of recreational facilities; aesthetics; social environment; and self-reported access to environmental features for exercise. The aspects of social capital studied include: participation in community groups, and a composite social capital score. The two study outcomes were self-reported total minutes spent on moderate and vigorous physical activity per week which was evaluated as a continuous variable (minutes per week) and a dichotomous measure of whether the respondent was meeting the physical activity recommendations set forth by the US
CHAPTER II METHODS
Setting
The San Luis Valley (SLV) is a rural area of Colorado composed of six counties, three of which are considered “frontier” on the basis of having fewer than six people per square mile. It has higher prevalence of obesity, diabetes, and coronary heart disease when compared with the rest of the state of Colorado.3 Table 1 compares the prevalence of chronic disease in the SLV using data from the SLVCHS to prevalence estimates for Colorado and the US as a whole from the CDC BRFSS data in 2010. In addition, the SLVCHS showed that only 63.5% of residents in the SLV met the recommendation for weekly moderate to vigorous physical activity whereas the state-level prevalence was 70.9% according to the CDC 2010 State Indicator Report on Physical Activity.42 Adults meeting recommendations were defined by the 2008 US Department of Health and Human Services Physical Activity Guidelines for Americans which recommends at least 150 minutes of moderate-intensity physical activity, 75 minutes of vigorous-intensity physical activity, or a combination of moderate and vigorous-intensity physical activity (where vigorous activity minutes are multiplied by two) totaling at least 150 minutes per week.1,42
Table 1: Age-Adjusted Prevalence (%) of Chronic Diseases
Chronic Disease SLV Colorado* USA*
Diabetes 9 6 9
CHD 4 3 4
Myocardial Infarction 4 3 4
Obesity 25 21 28
San Luis Valley Community Health Survey 2010 Summary Report. Rocky Mountain Prevention Research Center. March 2012. *Colorado and USA data from CDC 2010 BRFSS data.
Study Design
This cross-sectional study utilized data from the 2010 San Luis Valley Community Health Survey conducted by the Rocky Mountain Prevention Research Center (RMPRC) on
comprehensive health and lifestyle data including demographics, disease status, diet, physical activity, and other health related domains. The survey was conducted through the use of Community Based Participatory Research (CBPR) methods that incorporate community member input in all phases of the research process.44
Survey respondents were sampled using a stratified, multi-stage cluster design that identified a sample of residents within the six counties which compose the San Luis Valley. The two stage sampling design included stratification by county, and by population densities of persons 18 years of age and older as defined by the 2009 US Census with “low” density being areas with fewer than 50 people per square mile and “high” density being those areas with greater than 50 people per square mile. Census blocks or clusters of census blocks were used to create Primary Sampling Units (PSU) with 33 to 66 adults per cluster. The sample of households within PSUs was selected such that each household within a county had an equal probability of selection. Next, by utilizing maps including aerial photographs, street layers, and house
numbers, a route was drawn which started from the randomly selected household and passed by every structure that looked like a house in the most efficient manner. Finally, community liaisons walked the route recording occupied homes and adding new structures that may have not been on the maps. Occupied households were then systematically selected by picking every nth home, where n was the total number of occupied households in the PSU divided by 12, such that 12 homes were selected. This sampling was done with the expectation that a 50% response rate would mean a minimum of 6 respondents would be in each Primary Sampling Unit.
to select one adult aged 18 or older to participate in the survey. In a household with multiple adults, the adult selected for the survey had the next closest birthdate to the interviewer’s visit. Interviews were completed by telephone and in person at the respondent’s home if they so requested between June and November, 2010. The overall response rate was 66% with 1187 interviews completed. Table 5 in the results section details the demographics of the population surveyed.
For every completed survey, an audit of the respondent’s street segment (defined as a section of street between two intersections or 0.25 miles or less) was conducted using a version of the Analytic Audit Tool from Active Living Research developed by Brownson, et al, adapted for the San Luis Valley by excluding items that were not applicable to this rural setting such as public transit stops, beaches, and marinas.45 The audit was completed by trained auditors who worked in pairs between the hours of 9am and 4pm on weekdays only to limit potential bias from differences between weekday versus weekend street characteristics such as traffic volume and social behaviors. The audit recorded features of the built environment such as residential, and commercial land use, the quality of sidewalks and streets, aesthetics, the social
environment, and presence of recreational equipment, service amenities, and natural features.45 Definition of Outcome Variables
Two outcomes were tested: a continuous measure of the total minutes engaged in moderate and vigorous physical activity per week, and a dichotomous measure of meeting weekly physical activity recommendations.
The first outcome of interest was the self-reported total minutes spent on moderate and vigorous physical activities in a week. Survey questions that were used to measure physical activity were taken from the Behavioral Risk Factor Surveillance Survey (BRFSS) administered by the CDC (Table 2).46 This continuous outcome variable had a large positive skew and also had
198 observations reporting 0 minutes per week of total physical activity. The large proportion of zeros made log transformation less than ideal because of the inability to log transform a zero value. Thus, a square root transformation was selected to reduce the skew of the outcome variable in preparation for linear regression models. This is similar to what was done in a study conducted by Kirby, et al. in 2006 in which the same continuous outcome variable of total minutes spent on moderate and vigorous physical activity per week was positively skewed and contained a large proportion of zero values.23 All analyses of the continuous outcome in this study use the square root transformed total minutes of physical activity measure as the outcome variable.
Table 2: Physical Activity Survey Questions
Question Recorded Response
Now, thinking about the moderate activities you do [fill in “when you are not working” if employed or self-employed] in a usual week, do you do moderate activities for at least 10 minutes at a time, such as brisk walking, bicycling, vacuuming, gardening, or anything else that causes small increases in breathing or heart rate?
“Yes” or “No” recorded. (If “No”, next 2 questions are skipped) How many days per week do you do these moderate activities for at
least 10 minutes at a time?
Number of days recorded. On days when you do moderate activities for at least 10 minutes at a
time, how much total time per day do you spend doing these activities?
Number of hours and number of minutes recorded. Now, thinking about the vigorous activities you do *fill in “when you
are not working” if employed or self-employed] in a usual week, do you do vigorous activities for at least 10 minutes at a time, such as running, aerobics, heavy yard work, or anything else that causes large increases in breathing or heart rate?
“Yes” or “No” recorded. (If “No”, next 2 questions are skipped) How many days per week do you do these vigorous activities for at
least 10 minutes at a time?
Number of days recorded. On days when you do vigorous activities for at least 10 minutes at a
time, how much total time per day do you spend doing these activities?
Number of hours and number of minutes recorded.
The second outcome was dichotomous and indicated whether or not a respondent was meeting weekly physical activity recommendations based on their self-reported physical activity
data. The metric used to define who was meeting recommendations was what the US
Department of Health and Human Services published in the 2008 Physical Activity Guidelines for Americans which recommends that adults should engage in at least 150 minutes of moderate-intensity physical activity, 75 minutes of vigorous-moderate-intensity physical activity, or a combination of moderate and vigorous (where vigorous activity is multiplied by two) totaling at least 150 minutes per week.1
Definition of Built Environment and Social Capital
Measures of the built environment were taken from the Analytic Audit Tool developed by Brownson, et al, and from questions on the SLVCHS.45 Measures of social capital were taken from the SLVCHS. The following sections detail the built environment variables from the street audit, built environment variables from the SLVCHS, and the social capital variables from the SLVCHS.
Measures of the built environment that were derived from the street audits were established according to domains from Brownson, et al (2004) and conceptual domains that were of specific interest in this study.45 They included businesses and services; integration of natural land; street characteristics for walking and biking; recreational facilities; aesthetics; and the social environment. Missing from these domains is Brownson’s “Land Use” domain which we decided to depart from based on poor internal reliability tests with SLV data (Cronbach’s α=0.11), and the fact that it included items which were disparate (such as residences, commercial structures, and natural land features) which did not suit our study purpose to investigate specific built environmental features. Thus, items from the Land Use domain were used to develop two separate composites – businesses and services, and integration of natural land. Cronbach’s alpha (α) was calculated on each construct to test its internal reliability. However, it is important to note that while this is a common test of internal reliability, there is
no reason to expect similar environmental features to co-occur. Thus, greater weight was placed on the conceptual basis for the constructs rather than solely on their alpha scores. Integration of natural land (α=0.66), street characteristics (α=0.75), recreational facilities (α=0.81), and social environment (α=0.67) showed reasonably good internal reliability. Businesses and services (α=0.34) had a poor alpha score however, this was expected given the range (e.g. laundry, museums, etc…) and large number of items represented which could have resulted in a lower alpha. Aesthetics (α=0.33) was also a weak construct however Brownson notes in his 2004 analysis that this construct showed lower internal reliability which he hypothesized might have been due to the fact that it is heavily reliant on the auditor’s perception of the environment and less static features such as litter.45 Table 3 lists the composite variables and the sub-domains used in this study.
Measures of the built environment that were taken from the telephone survey were from three questions regarding the respondent’s access to certain features and facilities in their neighborhood. Table 4 lists the exact questions asked which assessed perceived access to the built environmental features of parks and trails, public exercise facilities, and safe sidewalks for walking. These variables were compared against similar items from the street audit to assess whether there was agreement between the street audits and survey responses. In addition, they were also independently evaluated in a model for both the continuous and the dichotomous outcomes for physical activity.
Table 3: Built Environment Variables from Street Audit Composite Sub-Domains Businesses & Services (Chronbach’s α=0.34)
Commercial Land Use: Defined as the sum of commercial
establishments (e.g. restaurants, shops, etc…) present on the street segment.
Public/Government Services Land Use: Defined as the sum of government establishments (e.g. libraries, post offices, etc…) present on the street segment.
Integration of Natural Land (Chronbach’s
α=0.66)
Sum of natural features near built street segments which include bodies of water (e.g. lakes, rivers, streams), mountains, canyons, and other natural space (e.g. wooded area, swamp, meadow). Street
Characteristics (Cronbach’s α=0.75)
Walkability: A combination of a subjective assessment of 0 – 3 for walkability plus the presence of sidewalks on one or both sides of the street. Higher values mean greater walkability and presence of sidewalks.
Bike-ability: A combination of a subjective assessment of 0 – 3 for walkability minus whether the road is dirt or gravel. Higher values mean greater bike-ability and a paved road.
Presence of crossing aids for pedestrians and bicyclists to cross the street safely (e.g. crosswalks, stop lights, stop signs).
Presence of street lighting at night. Recreation Facilities
(Cronbach’s α=0.81)
Public Recreational Equipment: Defined as the sum of playgrounds and sports equipment (such as slides, swings, goal posts, etc…) present on the street segment.
Service Amenities: Defined as the sum of service amenities (such as restrooms, picnic tables, trash bins, etc…) present on the street segment.
Recreational Facilities Land Use: Defined as the sum of recreational facilities (such as parks, playgrounds, fitness centers, etc…) present on the street segment.
Availability of Trails: Defined as the presence of a path or trail for multi-use – biking/walking – on one or both sides of the street. Aesthetics
(Cronbach’s α=0.33)
Aesthetics: Defined as the sum score of attractive features (such as décor, shade trees, benches, etc…) minus the sum score of
unattractive features (such as pollution, noise, and disorder). Scoring recorded as “None”, “A little”, “Some”, and “A lot”. Social Environment
(Cronbach’s α=0.67)
Social Environment: Defined as the sum score of people visible on the street and whether they are engaging in activity or social behaviors minus the presence of people being hostile and the presence of unrestrained animals. Scoring recorded as “None”, “A few”, “Some”, and “A lot”.
Table 4: Built Environment Variables from the Telephone Survey Questions:
Are there any parks or trails in your neighborhood where you can walk, run, or bike? Do you have access to public exercise facilities such as walking or running tracks, basketball or tennis courts, swimming pools, sports fields, etc., in your neighborhood?
Are there sidewalks or shoulders of the road in your neighborhood sufficient to safely walk, run or bike?
Measures of social capital were derived from the survey and included social capital and participation in community organizations. Social capital was assessed for each respondent through the use of fourteen questions which were included on the telephone survey. These questions were designed to capture levels of reciprocity, trust, civic participation, collective efficacy, and social cohesion into one continuous score of social capital. Collective efficacy is the ability of a neighborhood to self-regulate and work together through informal social controls as defined by Sampson, et al in 2007.38 In addition, Sampson also defines the concept of social cohesion as the amount of trust community members feel towards their neighbors.38,47 The questions originate from different validated instruments by Burdine, Sampson, and Kawachi and were utilized as a continuous variable for social capital with lower values indicating little social capital and higher values indicating greater social capital.38,39,48 The construct had a Cronbach’s alpha of 0.85 which indicated strong internal reliability. In addition to this overall assessment, membership in community organizations was summed through a series of questions asking whether the respondent belonged to business or civic groups, religious organizations, charity or volunteer organizations, ethnic or racial organizations, neighborhood associations, PTA or other school related groups, political organizations, social clubs, and any other groups the respondent wished to report that weren’t asked. Participation in community organizations was included as a continuous variable.
Confounders
Variables analyzed as potential covariates and confounders included income, gender, age, and ethnicity. These were selected after a literature review of covariates and associations from other studies. The literature supports income as a potential confounder for its
documented association with physical activity, built environment, and social capital.37,40,49 Age is well known to be associated with physical activity levels and is often included in analyses where physical activity is the outcome to account for its independent relationship.50 Gender is also documented as being associated with physical activity.51 Ethnicity is often investigated for an association, but not consistently utilized. It is not hypothesized to be an independent predictor of physical activity in this setting, but it will be investigated.
Data Analysis
Statistical analysis for this study was completed with SAS 9.4 software. Considering the stratified cluster design of the study, SAS survey commands were utilized to account for
stratification and clustering in the study design. These commands include PROC SURVEYMEANS, PROC SURVEYFREQ, PROC SURVEYREG, and PROC SURVEYLOGISTIC.
In preparation for the analysis for this study, completed telephone surveys were linked to corresponding street audits to capture a more comprehensive picture of self-reported survey data and the respective street characteristics for each respondent. The descriptive statistics of mean, standard deviation, frequency, and percentages were calculated for all variables used in the analysis. Then, covariates were analyzed for univariate associations with the outcome variables. Next, agreement between self-reported perceived access to built environmental features and their related measures on the street audit were compared using Cronbach’s kappa. Finally, linear regression models between measures of the built environment and social capital, and total time spent on moderate and vigorous physical activity were run, controlling for the
appropriate covariates. Similar logistic regression models were run against the dichotomous measure of meeting weekly physical activity recommendations and controlling for the appropriate covariates.
Directed Acyclic Graphs (DAG) were constructed to better visualize the variables of interest and how they all related to each other for planning of the final analysis. Figure 2 shows the DAG for the hypothesized relationship between built environment and physical activity. In this diagram, the relationship between the built environment and physical activity might be confounded by income based on evidence from other studies the literature.37,49 Age and gender are included as precision variables for their association with physical activity and their lack of association with the built environment or social capital. Figure 3 shows the DAG for the relationship between social capital and physical activity. Again, income is a confounder of the association between social capital and physical activity, as well as the association between participation in community organizations and physical activity. As in the previous diagram, age and gender are included as precision variables for their association with physical activity. Ethnicity is not included in these diagrams because univariate analysis and literature review determined it was not significantly associated with physical activity. The analysis is detailed in the results section and in Tables 8 and 9.
Figure 2: DAG for Association between Built Environment and Physical Activity
CHAPTER III RESULTS
The following sections detail the results of all analyses conducted for this study. They include the descriptive statistics of all variables, univariate associations between the covariates and both measures of physical activity, and assessment of agreement between street audits and self-reported access to similar built environmental features. In the final section, the results for the linear and logistic models of the measures of the built environment from the street audit are detailed, followed by the linear and logistic models for the measures of built environment and social capital from the SLVCHS.
Descriptive Statistics
Descriptive statistics were run on participant characteristics including total numbers, percent frequencies, the mean of total minutes engaged in moderate and vigorous physical activity per week, and the percent of those meeting physical activity recommendations (Table 5). Descriptive statistics on the physical activity outcome measure of total minutes of moderate and vigorous physical activity includes the range, mean, median, and standard deviation. The dichotomous outcome of meeting weekly physical activity recommendations includes total numbers and percent frequencies (Table 6). Measures of the built environment and social capital include the range, mean, median, standard deviation, and interquartile range (Table 7). All descriptive statistics were run prior to analysis.
Table 5: Survey Demographics Variable n % Mean total weekly PA in minutes % Meets PA Recommendations total 1187 100% 501.55 63.5% Gender Female 691 58.21% 441.03 60.41% Male 496 41.79% 586.32 67.84% Income < $25,000 538 48.86% 474.47 57.77% $25,000 -< $50,000 346 31.43% 519.96 66.96% $50,000 -< $75,000 124 11.26% 530.38 77.87% $75,000 + 93 8.45% 538.87 75.27% Ethnicity Non-Hispanic White 604 50.88% 531.70 66.05% Hispanic 553 46.59% 462.40 60.37% Other 30 2.53% 617.50 70.00% Age 18 – 24 83 6.99% 569.15 67.07% 25 – 34 150 12.64% 573.26 72.30% 35 – 44 141 11.88% 491.04 67.63% 45 – 54 237 19.97% 413.13 61.80% 55 – 64 268 22.58% 553.99 66.04% 65+ 308 25.95% 475.30 55.33%
Age Mean SD Range
(Continuous) 52.37 17.32 18 – 93
Table 6: Descriptive statistics for measures of physical activity
Range Mean Median Standard
Deviation Continuous total Minutes of moderate and vigorous PA 0 – 9480 501.55 240.00 799.55 Meets PA Recommendations Yes No n=746 63.50% n=428 36.46%
Table 7: Descriptive statistics for measures of built environment and social capital
Variable Range Mean Median Standard
Deviation Interquartile Range (Q3 - Q1) Businesses & Services 0 – 6 0.28 0.00 0.71 0.00 Integration of Natural Land 0 – 3 0.46 0.00 0.83 1.00 Street Characteristics 2 – 11 5.88 6.00 1.97 3.00 Recreation Facilities 0 – 13 0.30 0.00 1.41 0.00 Aesthetics 3 – 12 8.29 8.00 1.01 1.00 Social Environment 4 – 19 7.72 7.00 2.51 3.00 Social Capital 13 – 62 47.53 49.00 9.58 13.00 Participation in Community Organization 0 – 7 1.73 1.00 1.49 2.00
Are there any parks or trails in your neighborhood where you can walk, run, or bike?
0 – 1 0.74 1.00 0.44 1.00
Do you have access to public exercise facilities such as walking or running tracks, basketball or tennis courts, swimming pools, sports fields, etc., in your neighborhood?
0 – 1 0.51 1.00 0.50 1.00
Are there sidewalks or shoulders of the road in your neighborhood sufficient to safely walk, run or bike?
Analysis of Covariates
The covariates of gender, age, income and ethnicity were tested for associations with the physical activity out measures of total minutes of moderate and vigorous physical activity and meeting physical activity recommendations. Table 8 lists the covariates and their respective beta coefficients or F-values, and the p-values for associations with the outcome of total
minutes of moderate and vigorous physical activity per week. Gender (β=2.54, p<0.01) had a significant association indicating a positive relationship between being male and total minutes of physical activity. The continuous age variable (β=-0.06, p=0.02) had a negative association with total physical activity indicating that advanced age was associated with decreased physical activity. The categorical income variable (F(3, 1090)=3.84, p=0.01) also had a statistically significant
positive association with total physical activity indicating that higher income was associated with increased physical activity. The categorical ethnicity variable (F(2, 1171)=2.78, p=0.06) did not have
a significant association and thus, was not included in the models. Table 8: Association of Covariates with total minutes of PA
Covariate β 95% Confidence
Intervals p-value
Gender 2.54 0.91, 4.17 <0.01
Age (continuous) -0.06 -0.10, -0.01 0.02
Categorical
Covariate F-value p-value
Income F(3, 1090)=3.84 0.01
Ethnicity F(2, 1171)=2.78 0.06
Table 9 lists the same covariates and their associations with the dichotomous outcome of meeting physical activity recommendations. Gender (β=0.33, p=0.01), age (β= -0.01, p<0.01), and income (χ2(3)=28.12, p<0.01) were all statistically significant thus, included in the final logistic
Table 9: Association of Covariates with meeting PA recommendations Covariate β 95% Confidence Intervals p-value Gender 0.33 0.09, 0.56 0.01 Age (continuous) -0.01 -0.02, -0.01 <0.01 Categorical
Covariate Chi-Square p-value
Income χ2(3)=28.12 <0.01
Ethnicity χ2(2)=4.86 0.09
Comparison between Street Audits and Self-Report
To determine whether survey questions were in some agreement with street audit data, Cronbach’s kappa tests were used to compare similar items. The first question, “Are there any parks or trails in your neighborhood where you can walk, run, or bike?” included two items from the street segment audit (the sum of parks and trails) and showed a significant divergence from each other with a kappa=0.03 (Figure 4).
The second question, “Do you have access to public exercise facilities such as walking or running tracks, basketball or tennis courts, swimming pools, sports fields, etc., in your
neighborhood?” most closely matched the Recreational Facilities Land Use sub-domain from the street audit tool and showed a significant divergence of kappa=0.03 (Figure 5).
The third question regarding sidewalks, “Are there sidewalks or shoulders of the road in your neighborhood sufficient to safely walk, run or bike?” also showed a significant divergence of kappa=0.21 (Figure 6) when tested against the measure of walkability from the street audit. These significant differences show that there was little agreement between the respondent’s perception of their environment and that of the street audit.
Figure 4: Agreement between survey and street audit measures of Parks and Trails
Figure 6: Agreement between survey and street audit measures of walkability
Results for the Street Audit Variables
Results for analysis of the objectively collected street audit variables included both linear regression for the continuous outcome of total minutes of moderate and vigorous physical activity, and logistic regression for the dichotomous outcome of meeting physical activity recommendations.
First, linear regression models for the continuous outcome of the square-root of the total minutes of moderate and vigorous physical activity per week were run using PROC SURVEYREG, controlling for age, gender and income, and taking the sampling method of stratification by county and clustering by PSU into account. Integration of natural land (β= 1.46, p<0.01) had a statistically significant positive association with the square-root of total minutes of moderate to vigorous physical activity. Back-transformation is complex and changes for every value of the measure of the built environment however, this is equivalent to approximately 47 additional minutes of physical activity for every single natural feature present (Figure 7). Street characteristics for walking and biking (β= -0.57 p=0.01) had a significantly negative relationship which represented approximately 16 fewer minutes of physical activity for every additional
point in favor of features hypothesized to be conducive to walking and biking such as sidewalks, crossing aids, and street lights (Figure 8). Recreational facilities (β= -0.53, p=0.01) also had significantly negative relationship which represented approximately 10 fewer minutes of physical activity for every additional recreational feature or service amenity (Figure 9). No other statistically significant associations were found for the linear regression models for the
measures of the built environment from the street audits.
The graphs in Figures 7, 8, and 9 represent the true models for integration of natural land, street characteristics for walking and biking, and recreational facilities including the covariates for age, gender and income. Due to the square-root transformation of the outcome of total minutes of moderate and vigorous physical activity per week, back-transformation yields a quadratic equation. Since this equation will change for every value of the particular measure of the built environment and for each covariate, these models are fixed at the center of the distribution of the measure of the built environment, assuming a male (dotted line) or female (solid line) person aged 52 years and in the lowest income bracket.
Male Female
Graph of individuals aged 52 in lowest income bracket
Plotting the model we see that it looks roughly linear. Solving for the tangent line of the curve y=(14.67+1.46x)2 at the center of the range of values (x=1.5) for Integration of Natural Land, we find a slope of 47.22. Therefore, every 1 additional natural land feature on a street segment is
associated with approximately 47 additional minutes of physical activity per week. Figure 7: Plot and interpretation of linear model of Natural Land Integration
Male Female
Graph of individuals aged 52 in lowest income bracket
Solving for the tangent line of the curve y=(18.49-0.57x)2 at the center of the range of values (x=4.5) for Street Characteristics, we find a slope of -15.95. Therefore, every 1 additional point of improvement (sidewalk, shoulder, crosswalk) is
associated with approximately 16 fewer minutes of physical activity per week.
Figure 8: Plot and interpretation of linear model of Street Characteristics for Walking and Biking
Male Female
Graph of individuals aged 52 in lowest income bracket
Solving for the tangent line of the curve y=(15.5-0.53x)2 at the center of the range of values (x=6.5) for Recreational Facilities we find a slope of -9.54. Therefore, every 1 additional recreational item on a street segment is
associated with approximately 10 fewer minutes of physical activity per week.
Figure 9: Plot and interpretation of linear model of Recreational Facilities
Second, logistic regression models for the dichotomous outcome of meeting weekly physical activity recommendations were run using PROC SURVEYLOGISTIC, controlling for age, gender and income, and taking the sampling method of stratification by county and clustering by PSU into account. Street characteristics for walking and biking (OR=0.91, p<0.01), and recreational facilities (OR=0.90, p<0.01) both had statistically significant negative associations with meeting weekly physical activity recommendations. No other statistically significant
associations were found for the logistic models of the street audit measures. Table 10 details the results for both physical activity outcomes for each measure of the built environment. Table 10: Results for Linear and Logistic Regressions for Street Audit Measures
Model Linear Regression for
Continuous PA outcome
Logistic Regression for Dichotomous PA outcome Businesses & Services β= -0.58 (95% CI: -1.59, 0.43)
p= 0.26 OR= 0.87 (CI: 0.73, 1.03) p= 0.10 Integration of Natural Land* β= 1.46 (95% CI: 0.46, 2.47) p< 0.01 OR= 1.18 (CI: 0.99, 1.41) p= 0.07 Street characteristics for walking and biking*
β= -0.57 (95% CI: -0.96, -0.17) p= 0.01
OR= 0.91 (CI: 0.86, 0.97) p< 0.01
Recreational Facilities* β= -0.53 (95% CI: -0.86, -0.20) p< 0.01 OR= 0.90 (CI: 0.86, 0.94) p< 0.01 Aesthetics β= -0.23 (95% CI: -1.10, 0.64) p= 0.64 OR= 0.99 (CI: 0.86, 1.15) p= 0.92
Social Environment β= -0.08 (95% CI: -0.38, 0.21) p= 0.58
OR= 0.98 (CI: 0.93, 1.03) p= 0.37
*Statistically significant in at least one model All models controlling for age, gender, and income
Results for the Survey Variables
First, linear regression models for the continuous outcome of the square-root of the total minutes of moderate and vigorous physical activity per week were run using PROC SURVEYREG, controlling for age, gender and income, and taking the sampling method of stratification by county and clustering by PSU into account. Self-reported access to parks and trails in the neighborhood (β= 2.31, p= 0.04) had a significantly positive association with the square-root of total minutes of physical activity. After back-transformation, this represents approximately 60 additional minutes of physical activity for individuals who report having access to parks and trails in their neighborhood (Figure 10). Social capital (β= 0.12, p= 0.02) also had a significantly positive association which represented approximately 8 additional minutes of physical activity for every additional point in their social capital score (Figure 11). No other
statistically significant associations were found for the linear regression models for the self-report SLVCHS measures of the built environment and social capital.
Figures 10, and 11 represent the true models for self-reported access to parks and trails, and social capital including the covariates for age, gender and income, after back-transformation of the square-root outcome. These models are fixed at the center of the distribution of the measure of the built environment or social capital, assume a male (dotted line) or female (solid line) person aged 52 years and in the lowest income bracket.
Male Female
Graph of individuals aged 52 in lowest income bracket
Solving for the tangent line of the curve y=(13.39+2.31x)2 at the center of the range of values (x=0.5) for Access to Parks and Trails we find a slope of 59.35. Therefore, Having access to a park or trail in your neighborhood is associated with approximately 60 additional minutes of physical activity per week.
Figure 10: Plot and interpretation of linear model of Access to Parks and Trails
Male Female
Graph of individuals aged 52 in lowest income bracket
Solving for the tangent line of the curve y=(10.31-0.12x)2 at the center of the range of values (x=24.5) for Social Capital we find a slope of 8.35. Therefore, every 1 point gain in social capital score is associated with approximately 8 additional minutes of physical activity per week.
Second, logistic Regression models for the dichotomous outcome of meeting physical activity recommendations were run using PROC SURVEYLOGISTIC, controlling for age, gender and income, and taking the sampling method of stratification by county and clustering by PSU into account. Self-reported access to parks and trails in the neighborhood (OR=1.39, p=0.04), and self-reported access to public exercise facilities (OR=1.4, p=0.02) both had statistically significant positive associations with meeting physical activity recommendations. No other statistically significant associations were found for the logistic models of self-report SLVCHS measures of the built environment and social capital. Table 11 details the results for both physical activity outcomes for each measure of the built environment and social capital. Table 11: Results for Linear and Logistic Regression for Self Report of Environmental Features
Model Linear Regression for
Continuous PA outcome
Logistic Regression for Dichotomous PA outcome Parks/Trails in Neighborhood* β= 2.31 (95% CI: 0.38, 4.24) p= 0.04 OR= 1.39 (CI: 1.01, 1.91) p= 0.04 Public Exercise Facilities* β= 1.62 (95% CI: 0.02, 3.22) p= 0.08 OR= 1.4 (CI: 1.06, 1.78) p= 0.02 Safe sidewalks available β= -0.97 (95% CI: -2.25, 0.31) p= 0.31 OR= 1.10 (CI: 0.82, 1.47) p= 0.53
Social Capital* β= 0.12 (95% CI: 0.04, 0.20) p= 0.02 OR= 1.02 (CI: 1.00, 1.03) p= 0.03 Participation in Community Organizations β= 0.52 (95% CI: -0.06, 1.09) p= 0.06 OR= 1.08 (CI: 0.99, 1.17) p= 0.07
*Statistically significant in at least one model All models controlling for age, gender, and income
Figure 12 depicts the forest plot of all odds ratios and their respective 95% confidence intervals for all measures of the built environment and social capital from the street audit and the SLVCHS.
CHAPTER IV DISCUSSION
The built environment and social capital are associated with physical activity both in terms of total minutes spent physically active and in terms of meeting physical activity
recommendations. Three measures of the built environment were significantly associated with both total minutes of physical activity and with meeting weekly physical activity
recommendations: street characteristics for walking and biking, recreational facilities, and self-reported access to parks and trails. Street characteristics for walking and biking, and
recreational facilities were associated with lower levels of physical activity. Self-reported access to parks and trails in neighborhood was associated with higher levels of physical activity. Two additional measures were associated with higher total minutes of physical activity – integration of natural land, and social capital. Finally, self-reported access to public exercise facilities was associated with a higher likelihood of meeting weekly physical activity recommendations.
Street characteristics for walking and biking was a composite measure from the street audit and was composed of the auditor’s perception of walkability and bikability, and the presence of sidewalks, shoulders, street lighting, and crossing aids. A negative association was found with total weekly minutes of physical activity and with the likelihood of meeting weekly physical activity guidelines. While this is contrary to what was hypothesized, the literature regarding the impact of sidewalks and street lighting in rural settings has been inconclusive to date.6 In a study by Wilcox, et al of rural older African American and white women, the absence of sidewalks was associated with higher levels of physical activity. In addition to this negative result, most studies in the rural setting have had null findings.14–19,22,24 Most striking was a recent focus group study by Kegler, et al from Emory University in which participants in rural Georgia reported sidewalks were unimportant in their decision to exercise and stated reasons such as
being able to exercise on their property, light traffic so they could walk on the roads, and the ability to find ways to walk regardless of sidewalks.21 In another focus group study by Gangeness of rural women from two separate communities, respondents reported their sidewalks to be unsafe to walk on due to being in disrepair, an opinion shared by a city administrator.52
Therefore, the negative association we found raises questions about whether the emphasis on sidewalks in rural physical activity research is an artifact of the rational for their importance in urban settings, or perhaps an issue of disrepair which was not teased out from the audit information into a separate measure. Other possibilities are that wide roads, and low traffic volume could render sidewalks moot, especially in outlying areas such as parts of the frontier counties with fewer than 6 people per square mile.
The positive association between self-reported access to parks and trails in the
neighborhood and physical activity is consistent with what was hypothesized and supported by a study in the rural setting by Brownson, et all in 2000.53 However, many of the studies in rural communities where access to recreational facilities and parks comes up as positively associated with physical activity do not assess the impact of access or proximity to facilities, rather they draw associations between the usage of such facilities for exercise and increased levels of physical activity.14,17,22 Assessing the relationship between proximity to recreational features and physical activity is a weak area in the literature set in rural communities so this fills an important void.
In contrast, the measure of the presence of recreational facilities from the street audit was negatively associated with both total weekly minutes of physical activity and with meeting weekly physical activity recommendations. The recreational facilities construct was a broad measure of the sum of all recreational facilities including parks, trails, sports fields, and service amenities such as trash bins, toilets, and picnic benches. The negative association with both
measures of physical activity observed in our findings was not reported by other studies that used similar constructs, however relationships in other studies were null.15,16,18,24 In addition, recreational facilities on the street segments were rare and so this may have limited our ability to thoroughly investigate this relationship. Still, we do not feel this diminishes the positive association we found with reported access to parks and trails especially since the self-reported measure was in disagreement with the sum of the individual items for parks and trails off of our street audit which implies that perhaps our respondent’s concept of “neighborhood” extends beyond their street segment. Thus, it is important to not only observe the environment but to also record people’s perceptions to better understand behavior revolving around physical activity.
Integration of natural land and social capital both had positive associations with total weekly minutes of physical activity. In terms of the integration with natural land, this finding is in agreement with several studies in the literature indicating the positive association between access to natural land and increased physical activity levels.12,23,27,21 While, there are not many studies looking at this specific construct in the rural setting, studies by Kibry, et al and Björk, et al found results which indicated the positive impact natural environments have on physical activity in rural communities.
The positive finding between social capital and total physical activity is also consistent with the literature with many other studies showing this association in both rural and urban settings.32,35 However, the β coefficient is quite small which raises questions of the clinical significance of this finding and perhaps explains why there was not an association with meeting physical activity recommendations.
Finally, self-reported access to public exercise facilities in the neighborhood had a significant positive association with meeting weekly physical activity recommendations. This is
another positive result which further adds to the much needed associations between access to exercise facilities and increased physical activity. However, this measure also was in
disagreement with the street audit measure for access to exercise facilities.
Other studies have found disagreement between perceived and objective measures of the built environment. Most notable, Gebel, et al conducted a study of concordance between perceived and objective measures of walkability and found that persons who perceived their areas to be more walkable also walked more compared with those who did not, irrespective of the objective measures.29 This points to the importance of perception and that perhaps those who are more inclined to engage in exercise perceive their environment as conducive to this or may be more aware of access to facilities due to their more actively seeking out opportunities for physical activity in their environment.
In addition, the disagreement could further emphasize the need to redefine what rural residents consider to be their “neighborhood” and perhaps look beyond the street segment. It is possible that the concept of “neighborhood” which is asked in the self-report question gets at a broader geography which might be more meaningful to rural residents. In the rural focus group study conducted by Kegler, et al, residents in rural Georgia reported their “neighborhoods” to be anywhere from a hundredth of a square mile to 16 square miles but most reported it to be within about a half mile radius from their home.21 This is an important consideration when deciding what to measure and how far out to go to accurately assess the environment’s impact on physical activity levels.
To conclude, we found several interesting and exciting relationships in this study that add to the much needed evidence for built environment and social capital in rural
Strengths and Limitations
Strengths to this study include the Community Based Participatory Research (CBPR) approach utilized in the survey design, administration, and data collection. It is possible that greater survey compliance could be achieved than was expected due to the efforts of
community member volunteers. In addition, the data quality may be strong since community liaisons and volunteers were invested members of the community. In particular, the community liaisons allowed for in-person recruitment for the telephone survey which could have had a positive impacted the overall response rate. Another strength of this study is the detailed data collection and especially the use of street audits to assess the true built environmental
conditions on respondent’s street segments. Most studies use self-reported data for the built environment which raises questions regarding potential bias such as the issues of perception Gebel, et al found.29 However, allowing for some self-reported data to be collected in the survey also allowed comparisons to be made between objectively collected data versus self-report to elucidate possible contradictions.
There were several limitations in this study as well. One of which is that we relied on self-reported physical activity data, which can also be subject to bias. In addition, comments gathered by survey interviewers indicate that many respondents felt the survey was long and this could have led to possible distraction among some of the respondents and perhaps account for some of the incomplete and missing data. In particular, the social capital section was the last part of the survey and had more missing variables than other questions that were earlier on in the survey. Finally, our street audits were only conducted in front of the respondent’s home and thus did not capture all of the built environmental features in the area that respondents might consider to be within their neighborhood. Despite having access to GIS data, it was not utilized in this analysis and should be investigated further.
Implications for Future Research
The true relationships that may exist between the built environment and physical activity levels in rural settings are far from established. The question of the importance of sidewalks can be further elaborated in an analysis which teases out the condition of sidewalks and make separate variables for sidewalks in good condition and for those in poor condition to assess their impact on total physical activity or walking. While that data was available in this study, it was not utilized as such and will be looked into in future analyses. In addition, more studies should attempt to assess agreement between objectively measured built environmental features with self-reported features. For example, Brownson, et al reports in a 2009 review of the state of the science of built environment and physical activity that there is a lack of validated self-report measures of parks and trails.7
Finally, the large number of studies with null results and contradictory findings are beginning to indicate it might be time to re-think what researchers believe, and urban studies have shown, is important for physical activity in exchange for asking rural residents what they believe is important to them for physical activity. Urban and rural settings not only differ in terms of their environments but also in terms of the people who occupy these geographies and the things they value. Focus group studies much like those conducted by Kegler and Gangeness should be carried out on a larger scale in a variety of rural settings to allow researchers to reassess what independent variables of the built environment are important to rural residents in different regions as well as what they consider to be their “neighborhood.”21,52 Rural
environments are as varied as the geographies and industries across of the United States. From swamps in the south, to deserts in the west, and agriculture rich plains in between, what makes the environment inviting for outdoor physical activity might be very different in different
settings. Perhaps commercial destinations aren’t as important as access to natural features, distance to a neighbor’s home, or size of a resident’s property lot and its features.
In conclusion, more research is necessary to elucidate unique environmental features that are important in different rural settings and then to test those features for their
relationship to physical activity. Future studies should begin by talking with community
members to get a feel for how people engage with their environment and where they engage in physical activity. Next discover what features they find inviting and or necessary for outdoor physical activity. Finally, conduct a study with both self-report and objective measures of physical activity and combine these with a comprehensive assessment of the environment and surrounding areas complete with GIS maps which locate all features that were considered to be important to rural residents.
