Association between Perceived Neighborhood Environment and Walking among Adults in 4 Cities in Japan

Background Recent research highlights the importance of environment as a determinant of physical activity; however, evidence among Japanese is sparse. The aim of this study was to examine the association between perceived neighborhood environment and neighborhood walking for multiple purposes among Japanese. Methods We conducted a population-based, cross-sectional study of 1461 Japanese adults (age: 48.2 ± 14.1 years, men: 44.8%). Neighborhood environment and walking were assessed by a validated questionnaire. The odds ratio of active walkers was calculated in relation to environmental characteristics after adjustment for age, sex, and other potential confounders. Results Participants were more likely to walk when they perceived that there was high residential density (odds ratio, 1.47; 95% confidence interval, 1.11–1.96), fair land use mix–diversity (1.37, 1.04–1.81), good walking/cycling facilities (1.56, 1.19–2.04), and attractive aesthetics (1.49, 1.14–1.95). Environmental factors associated with walking differed with respect to the purpose for walking. The environmental characteristics associated with walking for daily errands and with walking for commuting were similar, and included residential density and land use mix. Walking for leisure was associated with walking/cycling facilities, aesthetics, and traffic safety. Stratified analyses showed some sex-specific associations. Among women, there was an unexpected inverse association of leisure walking with both residential density and land use mix–diversity. Conclusions The association between neighborhood environment and walking differed by walking purpose. The results were generally consistent with those of studies conducted in Western countries, except for the association of high residential density and good land use mix–diversity with less leisure walking in women. These results suggest possible targets for environmental interventions to promote walking.


INTRODUCTION
Regular physical activity reduces the risk of mortality, and the incidence of cardiovascular diseases, diabetes, and some cancers. [1][2][3] However, a large part of the population is not physically active in Japan and in many other countries. 4,5 Thus, physical activity promotion is a public health priority. 6 Data on physical activity determinants and correlates are needed as a basis for developing effective interventions. Many studies have focused on individual demographics and psychobehavioral factors. 7 However, recent progress in research suggests that certain environmental characteristics, such as residential density, access to destinations, walking facilities, aesthetics, safety, and access to exercise facilities are related to physical activity. [7][8][9][10][11][12][13] Interventions that target individuals have only a minimal impact on the physical activity levels of whole populations 14,15 ; however, changes to the environment are believed to have a long-term and substantial impact. 16 Although there is accumulating evidence on the association between physical activity and environment, the relevant studies have been mostly limited to Western countries, in particular the United States and Australia 12 ; only a few have been undertaken in Japan. [17][18][19] Evidence from study settings-including Japan-where the environment, culture, and physical activity patterns differ from those of Western countries, is thus valuable. Indeed, evidence from Japan could support or refute the generalizability of previous studies conducted in Western countries, and/or add new findings regarding associations between environment and physical activity. Also, data from Japanese are needed for the development of physical activity interventions in Japan.
We previously reported associations of environment with physical activity, using a convenience sample of Japanese adults. 18 In that previous study, environmental characteristics were associated with physical activity, but the findings were limited by the use of simple measures that could not differentiate the purposes for walking. In the present crosssectional study, we used a random community sample from 4 Japanese cities and measured walking as the outcome. Because environmental correlates are specific to the type and purpose of physical activity, 11,20 the aim of this study was to examine environmental correlates of neighborhood walking and its components, including walking for daily errands, walking for leisure, and commuting on foot.

Participants and data collection
This cross-sectional study was conducted from February 2007 through January 2008. A total of 4000 residents aged 20 to 69 years and living in 4 Japanese cities (Koganei, Tsukuba, Shizuoka, Kagoshima) were randomly selected from the registry of residential addresses and stratified by sex, age (20-29, 30-39, 40-49, 50-59, and 60-69 years), and city of residence, so that the sample included 2000 subjects of each sex, 800 subjects of each age category, and 1000 subjects from each city. As a result, the addresses of 100 subjects of a specific sex, a specific age category, and a specific city were obtained. Four cities were chosen so as to include various environmental conditions. Koganei is in the Tokyo metropolitan area and Tsukuba is a university town located 50 km northeast of Tokyo. Shizuoka and Kagoshima are located in central and western Japan, respectively, and are the capital cities of prefectures that include both urban and relatively rural areas. For data collection, a questionnaire was sent to and collected from participants via postal mail. To increase the response rate, invitation letters that described the content of the study were sent to all 4000 subjects 2 weeks before the survey. During the survey period, a call center was established to answer the questions of the subjects. Nonrespondents were mailed 2 additional requests to join the survey. If a participant submitted an incomplete survey, we asked that the survey be completed again. Ultimately, of the 4000 subjects identified, 1508 (37.7%) responded to the survey. After data cleaning, valid data were obtained from 1461 participants (final response rate: 36.5%). All participants signed an informed consent document before answering the questionnaire, and the study received prior approval from the Tokyo Medical University Ethics Committee.

Assessment of perceived neighborhood environment
On the self-administered questionnaire, the Neighborhood Environment Walkability Scale-Abbreviated Japanese Version (NEWS-AJ) was used as the environmental measure. [21][22][23] The NEWS questionnaire was originally developed in the United States to evaluate several neighborhood environmental factors believed to be related to physical activity undertaken for multiple purposes. It has been used in various countries. [24][25][26] The NEWS-AJ consists of 54 questions that assess 8 neighborhood environmental factors: (1) residential density, (2) land use mix-diversity, (3) land use mix-access, (4) street connectivity, (5) walking and cycling facilities, (6) aesthetics, (7) traffic safety, and (8) crime safety. Several of these factors are related to the concept of walkability, which is the ability to walk from one's home to nearby destinations. "Neighborhood" in this questionnaire meant the area within a 15-minute walk from a participant's residence. A sample of the questions used is shown in the Appendix. Scores on the 8 subscales were calculated by using a standardized scoring manual. 27 Higher scores indicate a more favorable environment for walking. The score for residential density was calculated as the sum of the weighted score of 5 items. 27 Land use mix-diversity was based on the reported walking distance to a list of 23 possible destinations, including shops, services, and recreation facilities. As for the other variables, scores were estimated as the mean of scale items that used a 4-point rating scale (1 = strongly disagree, 4 = strongly agree), including reverse coding of selected items. The psychometric properties of the questionnaire and the process by which it was translated into Japanese were reported in a previous study. 23 The test-retest reliabilities of the 8 subscales were from r = 0.76 to r = 0.96.

Assessment of walking
For the assessment of physical activity, a self-administered questionnaire was used. The questionnaire asked participants about their walking frequency (days/week), and average walking duration each day (min/day), with respect to 6 purposes: walking for daily errands, walking for leisure, commuting on foot to work, commuting on foot to school, walking during work, and walking for other purposes. The questionnaire instructed participants to consider all walks that involved at least 5 minutes of continuous activity. Walking time (min/week) was calculated as the product of walking frequency and duration. In this study, 4 variables were examined: (1) neighborhood walking (sum of the duration of 4 types of walking, walking for daily errands, walking for leisure, commuting on foot to work, and commuting on foot to school, min/week), and 3 specific types of walking, namely, (2) walking for daily errands (min/week), (3) walking for leisure (min/week), and (4) commuting on foot to work (min/ week). We examined these 3 specific types of walking because they were expected to occur in the participant's neighborhood.
Although commuting to school was also expected to occur in the neighborhood, we excluded this variable from the specific analyses because the present sample included only 31 participants (2.1%) who walked to school. The Spearman correlation coefficient between total walking time (the sum of 6 types of walking time) calculated from the questionnaire and step counts per day, as assessed by accelerometer in a part of the present study sample (n = 783), was 0.30 (P < 0.001).

Sociodemographic and other variables
The sex and age of each participant were obtained from the registry of residential addresses of each city. Information on employment status, years of education, height, weight, and self-rated health was obtained by self-report. Body mass index (BMI) was calculated from self-reported weight and height. Self-rated health was measured with a single item that asked participants to rate their health: participants chose the most suitable answer from a 5-point scale-excellent, very good, good, fair, and poor-for the statement, "In general, would you say that your health is...?".

Statistical analyses
To examine the association between the neighborhood environment as the independent variable and walking as the dependent variable, odds ratios for active walkers were calculated using logistic regression models. For the analysis, the scores for the 8 environmental variables were converted into tertiles (high/middle/low for residential density and good/fair/poor for the other 7 variables). For each of the 4 walking variables, participants were classified into 2 groups. For neighborhood walking, participants were divided into 2 groups by using the median: ≤90 min/week or >90 min/week. Regarding walking for daily errands, walking for leisure, and commuting on foot to work, the proportions of participants who reported walking for these purposes were less than 50%. Thus, participants were divided into 2 groups for each of these purposes: those who walked for a given purpose and those who did not. In the analyses of commuting on foot to work, we used data only from employed participants (n = 1083). To calculate odds ratios, the environmental factors expected to be associated with lower levels of walking were used as references ("low" for residential density and "poor" for the other 7 variables), ie, an odds ratio higher than 1.00 indicates the association of an activity-supportive environmental characteristic with active walking. Odds ratios were adjusted by age, sex, location of residence, employment status, educational level, BMI, and self-rated health. Statistical significance was considered to be present when P < 0.05. All analyses were conducted by using SPSS version 15.0 for Windows (SPSS Inc., Tokyo, Japan). Table 1 shows the characteristics of the participants. In the overall sample, 44.8% were men. The mean age ± standard deviation (SD) was 48.2 ± 14.1 years. The sample included participants of Tsukuba (25.1%), Koganei (26.9%), Shizuoka (26.1%), and Kagoshima (21.9%). The proportion of overweight participants (BMI ≥25 kg/m 2 ) was 26.5% of men and 12.4% of women. The proportions of participants who reported neighborhood walking, walking for daily errands, walking for leisure, and commuting on foot to work were 71.1%, 42.7%, 35.0%, and 29.0%, respectively. Table 2 shows the mean scores and SDs for the 8 environmental variables. The tertiles of these variables are also indicated, and participants were categorized into 3 groups. Table 3 shows the odds ratios for active walkers by environmental factor in the overall sample. Four environmental variables (high residential density, fair land use mix-diversity, good walking/cycling facilities, and good aesthetics) were significantly associated with neighborhood walking. Participants were more likely to walk when they perceived that there was high residential density (odds ratio, 1.47; 95% confidence interval, 1.11-1.96), fair land use mix-diversity (1.37, 1.04-1.81), good walking/cycling facilities (1.56, 1.19-2.04), and good aesthetics (1.49, 1.14-1.95). Regarding walking for particular purposes, there were specific associations between environment and walking. Active walking for daily errands was associated with 6 categories in 4 environmental variables: high residential density, good and fair land use mix-diversity, good and fair land use mix-access, and good street connectivity. In contrast, the environmental factors that were significantly associated with walking for leisure were different, and included good walking/cycling facilities, good and fair aesthetics, and good and fair traffic safety. The results regarding commuting on foot to work were similar to those for walking for daily errands: 3 environmental variables were significant-high residential density, good land use mix-diversity, and good land use mix-access.

RESULTS
Analyses stratified by sex (men, Table 4; women, Table 5) revealed some differences between men and women. Walking for daily errands and commuting on foot to work were associated with a higher number of environmental variables in women than in men. In men, there was no significant association between environment and commuting on foot to work. In the analyses of walking for leisure, the associations between environment and walking also differed by sex. Among men, those who perceived good and fair walking/ cycling facilities, good aesthetics, and good traffic safety tended to walk for leisure; among women, high residential density, good land use mix-diversity, and good and fair aesthetics were significantly associated with this type of walking. An interesting unexpected result was that women who reported high residential density and good land use mix-diversity walked less for leisure.

DISCUSSION
In the present study, the perceived environmental features of a neighborhood were associated with walking in that neighborhood. In addition, the environmental variables associated with walking differed with regard to the purpose for walking, which was consistent with previous studies. 10,11 Walking for transportation (ie, errands and commuting to work) was associated with neighborhood walkability, as defined by high residential density, mixed land use, and good street connectivity. Walking for leisure was associated with the quality of pedestrian facilities, neighborhood aesthetics, and traffic safety.
Because sex differences in the associations between environment and physical activity have not been widely studied, those observed in the present study are of particular interest. Sex-specific analyses revealed significant associations between environment and commuting on foot to work only in women. The reasons for this are unclear. One possible reason is that women are more likely to work within walking   Commuting on foot to work was examined only among the 559 participants who were employed.
Perceived Neighborhood Environment and Walking among Japanese 282 distance. The association between environment and walking for daily errands was also stronger and more consistent in women than in men, most likely because women play a greater role in managing households, and have more opportunities to walk for errands such as shopping, than do men. Because of this, neighborhood features may have been more important for this type of walking in women than in men. There were some unexpected findings in women. High residential density and good land use mix-diversity were both associated with less leisure walking among women. These results have 2 implications. One possibility is that high residential density and good land use mix-diversity, which were consistently related to walking for transportation in previous studies, 11 might create a less desirable environment for leisure walking. Leisure walking is generally faster and more continuous than transport walking. Very high residential density and a good land use mix could generate excess car and pedestrian traffic, thereby interfering with leisure walking. These results were not observed in studies conducted in the United States and Australia, probably because residential density is usually lower and land use mix is less diverse in these countries. We find it interesting that a particular environmental feature could promote 1 type of walking while inhibiting another. This finding also confirms the importance of examining purpose-specific walking in environmental studies. The second implication of the abovementioned findings is that styles of leisure walking might differ by sex. For example, women walking for leisure might seek out relaxing places and avoid high-density areas and mixed-use environments in order to escape people and distractions, while men may prefer more densely populated neighborhoods and convenient places for leisure walking, perhaps because they are not adversely affected by these environmental characteristics.
In a meta-analysis of 16 studies, Duncan reported that 4 environmental factors-physical activity facilities, sidewalks, shops and services (a variable similar to land use mix-diversity in the present study), and traffic safety-were associated with physical activity. 28 Owen reviewed 18 studies that examined environmental correlates of walking and observed that aesthetic attributes, facilities for walking (sidewalks, trails), accessibility of destinations (similar to land use mix-diversity in this study), perception of traffic, and busy roads were associated with walking for particular purposes. 10 This review also found that environmental factors associated with walking for exercise/leisure were different from those associated with walking for transport. Saelens and Handy showed that the findings from previous studies were confirmed in more recent investigations. 11 Although the present study is the first to find that high residential density and mixed land use could interfere with leisure walking among women, our results were generally consistent with those of earlier studies. Thus, results regarding the environmental correlates of walking and the specific environmental associations with different purposes for walking are generalizable to the Japanese population. This is an important finding because the physical and cultural environments in Japan differ from those of the Western countries in which previous studies were conducted. Among Japanese adults, living in walkable communities, as defined by high residential density, good land use mix, and good street connectivity, is an important factor in walking for transport, while walking facilities (eg, sidewalks), aesthetics, and traffic safety are important factors in walking for leisure. These are robust findings across countries.
The results regarding crime safety have been inconsistent. In Duncan's meta-analysis, no significant association was observed between crime safety and physical activity. 28 However, some previous studies reported associations between crime safety and physical activity, 29,30 and differences between sexes in these associations. Specifically, crime safety was associated with physical activity among women. We, too, examined sex-specific associations between perception of crime safety and walking; however, no significant association was identified for either sex. In Japan, variations in the perception of crime safety may be insufficient to demonstrate associations, as the country is generally perceived to be safe. Studies in a wider range of environments might more clearly illuminate the relationship between crime and physical activity.
There are several limitations in this study. First, the study was cross-sectional, so we are unable to address the direction of causality. Longitudinal or intervention studies are therefore needed in future research. Second, both environmental and walking measures were based on selfreports. We acknowledge the possibility of a discrepancy between perception and reality, even though the measures have been validated. [21][22][23] Third, the response rate was somewhat low, which might have resulted in selection bias. If we assume that these participants tended to have healthier lifestyles and greater motivation and skills to overcome environmental barriers to walking, as compared with the general population, then they may walk regularly even in a poor environment. If so, this study would underestimate the association of environmental factors with walking behavior. Studies with a higher response rate and less selection bias will enhance rigor in this field of research. Fourth, participants lived in central and western Japan, not in the colder northern region of the country. Climate may be an independent determinant of walking or an effect modifier of the associations between environment and walking. To ascertain the generalizability of the findings, studies encompassing a wider range of environments are needed.
In spite of these limitations, the present study offers new evidence on physical activity and environment in Japan, and helps to fill a large gap in the data from non-Western countries. The results revealed specific environment-walking relationships and contributed to understanding the environmental correlates of our most common physical activity-walking.

Conclusion
The association of neighborhood environment with walking differed by the purpose for walking. The results of the present study were generally consistent with those of studies conducted in Western countries. However, there were some differences, eg, high residential density and good land use mix were associated with less leisure walking among Japanese women. The findings suggest possible targets for interventions that aim to promote walking.
There are many places to go within easy walking distance of my home.
Street connectivity 3 1-4 The distance between intersections in my neighborhood is usually short (100 yards or less; the length of a football field or less). There are many alternative routes for getting from place to place in my neighborhood. (I don't have to go the same way every time.) Walking/cycling facilities 4 1 -4 There are sidewalks on most of the streets in my neighborhood.
There is a grass/dirt strip that separates the streets from the sidewalks in my neighborhood.

Aesthetics 4 1-4
There are many attractive natural sights in my neighborhood (such as landscaping, views). There are attractive buildings/homes in my neighborhood.
Traffic safety 4 1-4 There is so much traffic along nearby streets that it makes it difficult or unpleasant to walk in my neighborhood.
The speed of traffic on most nearby streets is usually slow (30 mph or less).
Crime safety 5 1-4 My neighborhood streets are well lit at night. Walkers and bikers on the streets in my neighborhood can be easily seen by people in their homes.