Exploring the Effect of the Environment on Physical Activity A Study Examining Walking to Work

Transcription

1 Exploring the Effect of the Environment on Physical Activity A Study Examining Walking to Work Cora L. Craig, MSc, Ross C. Brownson, PhD, Sue E. Cragg, MBA, Andrea L. Dunn, PhD Background: Methods: Results: Conclusions: Research on physical activity and the physical environment is at the correlates stage, so it is premature to attribute causal effects. This paper provides a conceptual approach to understanding how the physical design of neighborhoods may influence behavior by disentangling the potential effects of income, university education, poverty, and degree of urbanization on the relationship between walking to work and neighborhood design characteristics. The study merges Canadian data from 27 neighborhood observations with information on walking to work from the 1996 census. Hierarchical linear modeling was used to create a latent environment score based on 18 neighborhood characteristics (e.g., variety of destinations, visual aesthetics, and traffic). The relationship between the environment score and walking to work was modeled at the second level, controlling for income, university education, poverty, and degree of urbanization. With the exceptions of visual interest and aesthetics, each neighborhood characteristic contributed significantly to the environment score. The environment score was positively associated with walking to work, both with and without adjustment for degree of urbanization. Controlling for university education, income, and poverty did not influence these relationships. The positive association between the environment score and walking to work, controlling for degree of urbanization supports the current movement toward the development of integrated communities for housing, shops, workplaces, schools, and public spaces. Given the need for research to guide environmental interventions, collaboration among public health practitioners, urban planners, and transportation researchers is essential to integrate knowledge across sectors. Medical Subject Headings (MeSH): behavior, community, exercise, physical fitness, transportation, walking, work (Am J Prev Med 2002;23(2S):36 43) 2002 American Journal of Preventive Medicine Introduction From the Canadian Fitness and Lifestyle Research Institute (Craig, Cragg), Ottawa, Ontario, Canada; Saint Louis University School of Public Health (Brownson), St. Louis, Missouri; and The Cooper Institute for Aerobics Research (Dunn), Dallas, Texas Address correspondence and reprint requests to: Cora L. Craig, MSc, Canadian Fitness and Lifestyle Research Institute, Somerset West, Ottawa, Ontario. ccraig@cflri.ca. Participation in leisure-time physical activity is low in industrialized countries, with the majority of respective populations not meeting the recommendations of the U.S. Surgeon General in the United States or specific national targets elsewhere. 1 3 Physical activity is consistently associated with demographic factors and self-reported barriers to participation (e.g., lack of time, motivation, and skill). 4 8 Among adults, participation is associated with cost, safety, and proximity to facilities. 9,10 Having built a solid foundation of cross-sectional relationships between psychosocial variables and physical activity, research is beginning to examine causal pathways between social environmental variables (e.g., social support) and behavioral variables. 11 In contrast, research on physical environment variables (e.g., traffic volume) is at the case-study or correlates stage with relatively little known about the relationship of these environmental factors and physical activity. Therefore, it is premature to cast physical environment variables as mediating variables, as further research may point to a moderating or direct effect or to no effect on physical activity. 12 A cross-sectional study in 1997 found a nonsignificant association between environmental factors and physical activity level in neighborhoods. 13 However, Canadian and U.S. data suggest that access to physical 36 Am J Prev Med 2002;23(2S) /02/$ see front matter 2002 American Journal of Preventive Medicine Published by Elsevier Science Inc. PII S (02)

2 fitness facilities and other urban design factors (traffic, lighting, and sidewalk maintenance) may be important in influencing population rates of physical activity. 3,14,15 Walking is well suited to lifestyle interventions in the community and is widely promoted to sedentary individuals. 1,16,17 Although the review of Sallis et al. 18 showed that research on environmental interventions is limited, qualitative research has found that aesthetics, proximity to shops, and access to free facilities are important factors influencing walking. 19 In addition, safety, convenience, and aesthetics are positively correlated with walking and bicycling, although differing in degree for commuting and recreational activities Population density and mixed-use zoning are positive correlates of walking, as is a mass-transit as opposed to an automobile orientation. 24 Reviewing the extant transportation literature, Khisty 25 proposed seven factors to reduce barriers to walking: increased attractiveness; comfort; convenience; population density; mixed land use; safety; system coherence (directness); and continuity (completeness). Whereas university education and higher income levels are related to participation in physical activity, 3 low-income people are more likely to walk to work and to shops than to travel by automobile. 7 Other crosssectional studies showed that environmental barriers to walking, such as safety, are higher among low-income people. 3,26 A recent intervention demonstrated that constructing a walking trail in a rural area can encourage use among less-active groups (i.e., women and lower-income people). 27 However, relatively little is known about the influence of university education, income, and living in poverty on the relationship between the physical environment and physical activity. Given the paucity of research to date, more correlational studies are needed to guide the development of interventions. We present a case report illustrating how cross-sectional data can be useful in framing issues that need to be researched in longitudinal designs. Its purpose is to provide a conceptual approach to further our understanding of how environmental variables influence physical activity by disentangling the relationships of education, income, and type of neighborhood with walking to work. Specific questions follow: Are there significant population differences in walking to work according to neighborhood design characteristics? Do neighborhood levels of poverty, income, university education, and degree of urbanization influence the relationship between environmental factors and walking to work? Methods The current study was designed to merge data from two Canadian sources, a neighborhood observational study and the 1996 Canadian Census. Key considerations for sampling included maximizing the variation in urban design features, population density, and the social and demographic characteristics of residents. The two data sources are discussed below. Neighborhood Study A wide variety of neighborhood types was essential to investigate the relationship between commuting behavior and neighborhood characteristics. Accordingly, a convenience sample of 27 neighborhoods with known diversity of urban design, social class, and economic status was chosen in Ontario, Quebec, and Alberta. In each province, an urban center ( 900,000 residents), a nearby suburban center, and a small urban center ( 50,000) were selected. Neighborhoods were defined to include one or more census tracts, with the number of tracts determined to match municipal definitions of the neighborhood. Because of high population density, five neighborhoods consisted of multiple tracts (median 3). Diversity of neighborhood characteristics was increased by including a range of income levels across areas yielding at least three neighborhoods in each urban area, at least two in suburban areas, and at least three in small urban areas. The diversity of income, education, and neighborhood features 28 was confirmed using census data CANADIAN CENSUS The 1996 census was collected by Statistics Canada using self-administered questionnaires. 29 All citizens, landed immigrants, and nonpermanent residents were eligible to participate, and just under 98% of those eligible did so. 30 One fifth of the sample received a longer version, including questions on education, income, and usual mode of transportation to work, with the latter including walking to work as a distinct response category. University education was created by combining bachelor s degree and higher education levels. Income was a census-derived variable for annual employment income before taxes, averaged across employed residents. The percentage living in poverty was also a census-derived variable representing the percentage of individuals living in families who spend at least 20% more of their income on food, shelter, and clothing than the Canadian average, adjusted for community and family size. 31 Data Collection Characteristics of the neighborhood environment were rated by an observer on a 10-point Likert scale (Table 1). The four observers attended a 2-day training session that included photographs of neighborhood characteristics to be recorded and a practical session in a nearby neighborhood. Observers compiled ratings of the 27 neighborhoods between late fall 1999 and early spring All observations were taken during the morning and afternoon, and were distributed over both weekdays and weekend days. In a small substudy, the same observers independently coded environmental factors in two or four assigned neighborhoods, which yielded 156 values. A three-level hierarchical linear model 32 estimated inter-rater reliability with observer ratings nested within mea- Am J Prev Med 2002;23(2S) 37

3 Table 1. Environmental items rated on a ten-point scale Items Anchors Coding considerations covered in training Number of destinations None to very many Number of facilities, such as schools, parks, recreation facilities, stores, libraries, industrial buildings, businesses, etc. Variety of destinations Homogeneous to Mix of facilities (see number of destinations) highly mixed Inclusive of pedestrians Not at all to very inclusive Features making neighborhood accessible by all ages and both genders, people/oriented buildings, signage, Exclusive of pedestrians Not at all to very exclusive amenities, etc. Inaccessible to certain ages or genders, not people oriented, factories, few crossing opportunities, lack of curb cuts, signs limiting access, etc. Social dynamics Static to dynamic Potential to see people sitting, standing, and moving about (e.g., seating at facilities; seating in windows, porches, or balconies; garages in front; commercial windows; outdoor cafes, etc.) Walking routes None to very many Existence of accessible walking routes, such as sidewalks and paths, from a pedestrian s perspective Meets pedestrian s needs Not at all to very well Continuity of routes, multiple route choices, crossing lights, ease of navigating topography, traffic, obstacles, etc. Walking system Doesn t exist to well Continuous, uninterrupted walking routes developed Transportation system Discontinuous to seamless Connection to other modes of transportation, benches at transit stops, secure bicycle parking, etc. Complexity of stimulus Simple to complex Amount and variety of visual and auditory stimuli (e.g., architectural detail, building variety, signage, traffic volume and sound, vegetation, etc.) Potential overload of stimulus Not at all to very overwhelming Amount and variety of visual and auditory stimuli (e.g., architectural detail, building variety, signage, traffic volume and sound, vegetation, etc.) Visual interest low to high Type and variety of buildings, architectural interest, human scale, open versus closed facades, property maintenance, water features, green spaces, etc. Visual aesthetics Low to high Color, texture, composition, relative proportions (e.g., sky to concrete) Time and effort required Not at all convenient to very convenient Directness of routes, topography, obstacles, characteristics of intersections, etc. Traffic threats Not at all threatening Amount, speed, and separation from traffic to very threatening Obstacles None to many a Debris, construction, maintenance Safety from crime Not at all comfortable to very comfortable Lighting, front porches, escape routes, number of people around, potential avenues of escape, property maintenance, street type, etc. Potential for crime a Coded in reverse order. Not at all threatening to very threatening Graffiti, vandalism, disrepair, potential for lurking, street lighting, property maintenance sures, which were nested within neighborhoods. Given the nature of ecologic scores, it is perhaps not surprising that the inter-rater correlations ranged from 0.9 to 1.0. Analysis The neighborhood and census data were merged by census tract. In neighborhoods covering multiple census tracts, averages of the census tract data were calculated, weighted by resident population. Hierarchical linear modeling 32 was used to examine the relationship between neighborhood characteristics and walking to work. Conceptualizing neighborhood factors as measures of an underlying environment construct, an ecologic score was estimated at the first level of the model and then its relationship to walking to work was modeled at the second level. 33 (See Mâsse et al. 34 for details of this approach.) The factors were assumed to have equal variances, given the relatively small number of neighborhoods and the lack of contrary theoretical considerations. In each step, percentage walking to work, environmental factors, and neighborhood characteristics were centered around respective grand means. Degree of urbanization was modeled using dummy variables for suburban and small urban, with the effect for urban being reflected in the intercept. Income and university education were positively associated ( 0.70), controlling for degree of urbanization. Owing to the relatively small number of neighborhoods, the effects of university education, income, percentage living in poverty, and degree of urbanization were first tested independently and only those significantly contributing to the variance were included in the final model (Table 2). All data were included, as the exclusion of outliers did not affect the pattern or significance level of results. 38 American Journal of Preventive Medicine, Volume 23, Number 2S

4 Table 2. Summary of model testing Step Model 1 Fit unadjusted model of ecological environment score across environmental factors (Level 1) and neighborhoods (Level 2) 2 Test significance of whether controlling for specific environmental factors at Level 1 reduces variation of ecological environment score across neighborhoods (Level 2) 3 Test association between environment score and percentage walking to work at Level 2, controlling for environment factors at Level 1 4 Independently test associations between environment score and percentage walking to work, adjusting for each social class indicator (university education, income, and poverty) at Level 2 and for environment factors at Level 1 5 Fit final model, testing association between environment score and walking to work, adjusting for degree of urbanization at Level 2 and controlling for environment factors at Level 1 6 Examine impact of potential outliers on pattern of results and tests of significance Results Neighborhoods varied by population, university education, income, and percentage living in poverty. Neighborhood population ranged from 1118 to 53,817 residents, with a mean population of 10,983 and median population of 4953 (Table 3). Education level varied considerably across the areas, with the percentage having university education ranging from 8.9% to 69.2%. Annual income ranged from $15,294 to $49,866, and the percentage of residents living in poverty from 2.9% to 54.6%. Finally, the housing age of neighborhoods varied, including those associated primarily with older established neighborhoods (n 7); those of the postwar years (n 10); and those constructed more recently, typically after the advent of suburban design (n 10). 28 Most employed adults in the study neighborhoods commuted to work by car (Table 3) with relatively few taking public transportation, walking, or bicycling. The same pattern of travel behavior was observed regardless of degree of urbanization, although car travel was more prevalent in suburban neighborhoods. Accordingly, the percentage of the population walking to work was highest in urban areas and lowest in suburban areas. The ecologic environment score modeled the variation in the ratings of environmental factors. The intercept represents the mean score for number of destinations, with the factor coefficients representing difference scores. The acceptability of the multivariate normality assumption was confirmed by examining a plot of the observed versus expected distributions. Inclusion of the environmental factors increased the reliability of the score from 0.62 in the unconditional model to 0.80, increasing the discrimination between neighborhoods. The environmental factor coefficients ranged from 1.82 to Each factor was a significant contributor to the variation of the environment score (p 0.10 for transportation system and p 0.05 for other factors), except for visual interest and aesthetics. The inclusion of environmental factors (destinations, social dynamics, transportation system, and traffic) reduced the variation in the score by 46%. Walking to work was significantly related to the environment score (T-ratio [25] 3.32, p 0.003), with a one-unit increase in the score being associated with a 25-percentage-point increase in the percentage walking to work (Figure 1). The degree of urbanization altered the relationship between the environment score and walking to work (Figure 2). The predicted environment score was lower in both small urban (T-ratio [23] 3.61, p 0.002) and suburban neighborhoods Table 3. Demographic characteristics and travel behavior in the 27 study neighborhoods, 1996 Canadian Census Characteristic All areas All areas Urban Suburban Small urban Mean SD Median Median Median Median Number of residents 10,983 2,818 4,953 15,922 5,754 3,529 Annual employment income of workers 28,175 1,851 24,147 24,952 31,451 21,932 Residents with university education (%) Living in poverty (%) Mode of travel to work, workers aged 15 years (%) Driving a car Passengers in a car Using public transportation Walking Bicycling SD, standard deviation. Am J Prev Med 2002;23(2S) 39

5 Figure 1. Scatter plot of the environment score and percentage walking to work (T-ratio [23] 4.42, p 0.000) than in urban neighborhoods (Table 4). Nonetheless, the environment score was related to the percentage walking to work, controlling for degree of urbanization (T-ratio [23] 2.03, p 0.054). In contrast, income, university education, and percentage living in poverty exhibited nonsignificant relationships with environmental factors and walking to work. Figure 2. Environment score and walking to work, adjusted by degree of urbanization 40 American Journal of Preventive Medicine, Volume 23, Number 2S

6 Table 4. Neighborhood environment: walking to work and degree of urbanization (final model) Fixed Effect Mean SD Model coefficient SE p value For intercept Suburb Small urban Walking to work For gradients Number of destinations Reference Variety of destinations Inclusive of pedestrians Exclusive of pedestrians Social dynamics Walking routes Meet pedestrian s needs Walking system Transportation system Complexity of stimulus Potential overload of stimulus Visual interest Visual aesthetics Time and effort required Traffic threats Obstacles a Safety from crime Potential for crime a Coded in reverse order. Discussion Walking to work was significantly related to the neighborhood environment score and this was moderated by degree of urbanization. The increased likelihood of walking to work with increasing environment scores in suburban as well as urban neighborhoods supports the emergence of neotraditional design the utilization of traditional design concepts in urban development. 35 However, due to the cross-sectional nature of the design, it cannot be determined whether people were more likely to walk to work due to environmental factors or whether people live where they do because of personal circumstances and preferences. To determine the effect of various environmental factors on the behavior of residents, interventions or prospective studies examining the relationship of changes in the environment on changes to physical activity patterns are required. Southworth 36 cautions that simply mimicking the layout and features of traditional neighborhoods may not promote walking in new suburban neighborhoods if such neotraditional designs lack key features of traditional metropolitan areas (e.g., housing for poorer residents, and commercial and industrial spaces). Indeed, the positive association between poverty and walking to work found elsewhere 7 was not observed in this study due to the confounding of poverty with degree of urbanization. The relative absence of lowincome housing in suburbs is noteworthy and may be related to municipal policies and economic factors shaping local urban development. Research is needed to identify these factors and to assess the long-term health and social consequences of such urban design practices among people living in poverty. Berman 37 stated that the layout of suburbs and traditional urban neighborhoods is insufficient to encourage walking over automobile use, and that a number of factors, including the location of retail stores, nearby residences, and pedestrian-oriented commercial buildings, are required. This idea was expanded in the Ahwahnee Principles for planning communities, 35 comprised of 15 community, 4 regional, and 4 implementation principles. Of particular relevance to this study is the call for the development of integrated communities containing housing, shops, workplaces, schools, civic facilities, and open areas in easy walking distance of one another and transit stops; the design of streets and paths that discourage high-speed traffic and result in a fully connected system of intersecting routes to all destinations; and the design of neighborhoods to support diversity in age and socioeconomic status levels of residents and to encourage the presence of people in public space at all times of the day and night. Given the wide range of environmental factors covered by these principles, it is not surprising that most factors in this study contributed significantly to the environment score. The two exceptions were visual interest and, more notably, visual aesthetics. 19,38 This latter exception may be due to the timing of data collection in the late fall through early winter months, which minimized the relative contribution of individual gardens and emphasized permanent features of neighborhood aes- Am J Prev Med 2002;23(2S) 41

7 thetics, such as the relative proportion of concrete and buildings to open areas and sky. The neighborhoods were selected to cover a range of environmental characteristics inherent in urban and suburban neighborhoods. This included a range of size and population densities of areas, which in larger, more heterogeneous areas may not have captured important street-level variation. In addition, there may be increased variability in ratings across neighborhoods since observations were made on weekend days and weekdays as well as throughout the day. Furthermore, controlling for the distance to work might help to explain differences in commuting behavior between urban and suburban areas. Despite these shortcomings, the environment score exhibited appropriate attributes for an item response model. 39 The intercept value of 5.02 compared favorably with the identified scale midpoint of 5. In addition, a reasonably wide range of predicted coefficients were observed, with positive coefficients representing factors conducive to walking and negative coefficients representing factors limiting walking. Furthermore, the score reliability of 0.80 indicated a reasonable ability to discriminate between neighborhoods based on the environment values. Overall, the properties suggest that the creation of such ecologic scales may provide useful tools for further exploring relationships between neighborhood environments and physical activities appropriate to community settings. To date, studies of the influence of the environment on physical activity have been primarily cross-sectional, qualitative, or case-based. Although research is increasing, little progress has been made in furthering the research plan proposed by Sallis et al. 18 Therefore, while controlled intervention studies are desirable, natural experiments of environmental change are appropriate and more feasible. Given the urgent need for research to guide environmental interventions, collaboration among public health practitioners, urban planners, and transportation researchers is essential to rapidly advance knowledge by integrating findings across sectors. We are grateful to Dr. Storm J. Russell for her contribution to the conceptualization of the project and to Dr. Lise Gauvin for guidance in hierarchical linear modeling techniques. This research was supported by the Physical Activity Unit, Health Canada, Government of Canada. References 1. U.S. Department of Health and Human Services. Physical activity and health: a report of the Surgeon General. Atlanta, GA: U.S. Department of Health and Human Services, Centers for Disease Control and Prevention, National Center for Chronic Disease Prevention and Health Promotion, Armstrong T, Bauman A, Davies J. Physical activity patterns of Australian adults. Canberra: Australian Institute of Health and Welfare, Australian Institute of Health and Welfare, Craig CL, Cameron C, Russell SJ, Beaulieu A. Increasing physical activity: creating a supportive recreation and sport system. Ottawa, Ontario: Canadian Fitness and Lifestyle Research Institute, King AC, Blair SN, Bild DE, et al. Determinants of physical activity and interventions in adults. Med Sci Sports Exerc 1992;24(suppl 6): Dishman RK, Sallis JF. Determinants and interventions for physical activity and exercise. In: Bouchard C, Shephard RJ, Stephens T, eds. Physical activity, fitness and health: international proceedings and consensus statement. Champaign, IL: Human Kinetics, 1994: Craig CL, Russell SJ, Cameron C, Beaulieu A. Progress in prevention. Bulletin no. 4. Ottawa, Ontario: Canadian Fitness and Lifestyle Research Institute, Murakami E, Young J. Daily travel by persons with low income. In: Proceedings from the Nationwide Personal Transportation Survey Symposium. Washington, DC: U.S. Department of Transportation, Federal Highway Administration, Office of Highway Policy Information, 1997: Bauman AE, Sallis JF, Dzewaltowski DA, Owen N. Toward a better understanding of the influences on physical activity: the role of determinants, correlates, causal variables, mediators, moderators, and confounders. Am J Prev Med 2002;23(suppl 2): Sallis JF, Hovell MF, Hofstetter CR, Elder JP. Distance between homes and exercise facilities related to frequency of exercise among San Diego residents. Public Health Rep 1990;105: Corti B, Donovan RJ. The relative influence of individual, social and physical environment determinants of physical activity. Soc Sci Med 2002;54: Lewis BA, Marcus BH, Pate RR, Dunn AL. Psychosocial mediators of physical activity behavior among adults and children. Am J Prev Med 2002;23(suppl 2): Baron RM, Kenny DA. The moderator mediator variable distinction in social psychological research: conceptual, strategic, and statistical considerations. J Pers Soc Psychol 1986;51: Sallis JF, Johnson MF, Calfas KJ, Caparosa S, Nichols J. Assessing perceived physical environment variables that may influence physical activity. Res Q Exerc Sport 1997;68: Brownson RC, Baker EA, Houseman RA, Brennan LK, Bacak SJ. Environmental and policy determinants of physical activity in the United States. Am J Public Health 2001:91; Centers for Disease Control and Prevention. Increasing physical activity: a report on recommendations of the Task Force on Community Preventive Services. MMWR Morb Mortal Wkly Rep 2001;50: Siegel P, Brackbill R, Heath G. The epidemiology of walking for exercise: implications for promoting activity among sedentary groups. Am J Public Health. 1995;85: Health Canada, Canadian Society for Exercise Physiology. Canada s physical activity guide to healthy active living. Ottawa, Ontario: Health Canada, Sallis JF, Bauman A, Pratt M. Environmental and policy interventions to promote physical activity. Am J Prev Med 1998;15: Corti B, Donovan RJ, Holman CDJ. Factors influencing the use of physical activity facilities: results from qualitative research. Health Promotion J Aust 1997;7: Antonakos C. Environment and travel preferences of cyclists. Transportation Res Record 1994;1438: Baylis E. Pedestrianizing Bellevue s super block. In: Designing the urban village. Proceedings, 11th International Pedestrian Conference. Bellevue, WA: City of Bellevue, 1990: Hinshaw ML. Transforming a suburb. In: A decade reviewed commitment renewed. Proceedings, 10th International Pedestrian Conference. Boulder, CO: City of Boulder Transportation Division, 1989: Jacobs AB. City streets and their contexts. In: A decade reviewed commitment renewed. Proceedings, 10th International Pedestrian Conference. Boulder, CO: City of Boulder Transportation Division, 1989: Frank LD, Pivo G. Impacts of mixed use and density on utilization of three modes of travel single-occupant vehicle, transit, and walking. Transportation Res Record 1995;1466(April): Khisty CJ. Evaluation of pedestrian facilities: beyond the level-of-service concept. Transportation Res Record 1994;1438: Moore BJ, Glick N, Romanowski B, Quinnley H. Neighborhood safety, child care, and high costs of fruits and vegetables identified as barriers to increased activity and healthy eating linked to overweight and income. FASEB J 1996;10:A Brownson RC, Houseman RA, Brown DR, et al. Promoting physical activity in rural communities: walking trail access, use and effects. Am J Prev Med 2000;18: American Journal of Preventive Medicine, Volume 23, Number 2S

8 28. Parsons Brinckerhoff Quade, Douglas, Inc., Cambridge Systematics, Inc., Calthorpe Associates. Making the land use transportation air quality connection: the pedestrian environment. Vol. 4A. Portland, OR: 1000 Friends of Oregon, Statistics Canada Technical reports: sampling and weighting. Ottawa, Ontario: Statistics Canada, Statistics Canada Technical reports: coverage. Ottawa, Ontario: Statistics Canada, Statistics Canada. Low income cut-offs. Ottawa, Ontario: Statistics Canada, Raudenbush S, Bryk A, Cheong YF, Congdon R. HLM 5 hierarchical linear and non-linear modeling. Linconwood, IL: SSI Scientific Software International, Raudenbush SW, Sampson RJ. Ecometrics: toward a science of assessing ecological settings, with application to the systematic social observation of neighborhoods. Sociol Methodol 1999;29: Mâsse LC, Dassa C, Gauvin L, Giles-Corti B, Motl R. Emerging measurement statistical methods for studying mediators and moderators in physical activity intervention research. Am J Prev Med 2002;23(suppl 2): Corbett J, Velasquez J. The Ahwahnee Principles: toward more livable communities. Sacramento, CA: Center for Livable Communities, Southworth M. Walkable suburbs? An evaluation of neotraditional communities at the urban edge. J Am Planning Assoc 1997;63: Berman MA. The transportation effects of neo-traditional development. J Planning Lit 1996;10: Untermann RK. Street design: reassessing the function, safety and comfort of streets for pedestrians, or designing pedestrian-friendly commercial streets. In: The road less traveled: getting there by other means. Proceedings, 11th International Pedestrian Conference. Boulder, CO: City of Boulder Transportation Division, 1990: Lord FM. Applications of item response theory to practical testing problems. Hillsdale, NJ: Lawrence Earlbaum, Am J Prev Med 2002;23(2S) 43