Study design and setting
In this cross-sectional study, we conducted an online survey with 1545 parents of elementary school-aged children (6–12 years) who resided in Chiba prefecture, between 25 and 27 November 2020. At this time, a state of emergency due to COVID-19 had not been declared in Chiba prefecture, and therefore, the elementary schools were open (schools were temporarily closed due to COVID-19 only between March and May 2020). The online survey included questions on children’s modes of travel to and from school, neighborhood environments, and socio-demographic information. All participants provided informed consent prior to the commencement of the survey. This study was conducted in accordance with the STROBE guidelines for cross-sectional studies (Additional file 1).
We chose Chiba prefecture (population: 6.28 million, population density 1218 per hectare), which consists of 54 municipalities (37 cities, 16 towns, and one village), as the target study area because it has a mix of urban, suburban, and rural areas in a wide prefecture area [24]. Located on the outskirts of Tokyo, Chiba prefecture has several “new towns” (NT) built by municipalities and the Japan Housing Corporation to accommodate people who migrated to Tokyo during the postwar high-growth period from the late 1950s to the early 1970s. Following precedents in the UK, Japanese NTs were designed and built to ensure the safety and security of working-age families and their children [25]. One of their features is a web of greenways connecting abundant parks and greenspaces. These well-connected greenways are separated from the road network for traffic safety and are expected to function as corridors for pedestrian and bicycle commuters [26]. We hypothesized that the built environmental features of NTs might be associated with children’s school travel modes.
Each of the 54 municipalities has a municipal board of education (local educational authority) in charge of the public schools [27]. It is a common practice for each municipal board of education to establish walking to school as compulsory if the school is located within a certain distance (approximately 2 km) of the student’s home [19]. However, in recent years, elementary schools have been integrated and the school catchment area has widened due to the declining birthrate, which may in turn have led to less walking to/from school [28].
Participants
Participants were parents of elementary school-aged children who lived in Chiba prefecture. An invitation email that contained an outline of the study and a link to the online survey was sent to research panelists who lived in the prefecture and voluntarily registered with a third-party research company. Of the 90,050 research panelists, those who passed a screening question of having elementary school-aged children proceeded to the main survey questions. The survey link was deactivated when reaching a pre-determined target of 1500 respondents.
Measures
School travel modes
Participants were asked, “In the past month, how often did your child use a mode of travel to and from school other than walking?”, with five response options of ‘0 days,’ ‘1 or 2 days,’ 3–5 days,’ 6–10 days,’ and ‘11 days or more.’ The responses were recorded separately for ‘to’ and ‘from’ school. The school travel modes were recoded into three categories: everyday walkers (0 days), frequent walkers (more than half a month: 1 or 2 days, 3–5 days, 6–10 days), and less frequent walkers (less than half a month: 11 days or more).
Neighborhood environment
A neighborhood was defined as the area of a postcode (which is a proxy for home location) provided by the participants. The neighborhood environment for each participant was assessed based on the built environment, social environment, and neighborhood safety [18, 29].
Built environments
Neighborhood built environments were determined based on four measures: NT, walkability, distance to school, and population density. NT was measured to determine whether the nearest elementary school from the participants’ homes was located in NTs. The NTs include the Chiba NT (1930 ha; in Funabashi City, Inzai City, and Shiroi City), Ichihara NT (974 ha; in Chiba City and Ichihara City), Urayasu (1241 ha; in Urayasu city), Kaihin NT (1293 ha; in Chiba city), and Narita NT (482 ha; in Narita city), which are the five major NTs in Chiba prefecture (see Fig. 1) [30]. Neighborhood walkability was measured using four subscales of the Neighborhood Environment Walkability Scale in Youth (NEWS-Y) [31] and one item from the Abbreviated Neighborhood Environment Walkability Scale (ANEWS) [32]. The NEWS-Y subscales included walking facilities (three items), aesthetics (four items), traffic safety (seven items), and crime safety (six items). The ANEWS item was derived from the walking facility (“sidewalks are separated from the road/traffic in our neighborhood by guardrails and curbs”). Response scales (1 = agree, 2 = somewhat agree, 3 = neither agree nor disagree, 4 = somewhat disagree, 5 = disagree) were reverse-coded where appropriate, and the mean scores of each subscale were calculated. The shortest distance from a participant’s neighborhood (i.e., gravity center) to the nearest elementary school was calculated using ArcGIS 10.8 (Esri Japan Corporation, Tokyo, Japan). Information on school locations was obtained using the Zenkoku Gakko Data [33]. Neighborhood population density, obtained from Chiba prefecture [34], was used as a proxy for the availability of commercial facilities and public transport [35].
Social environments
Neighborhood social environments were assessed based on neighborhood cohesion (i.e., to what extent participants trusted their neighbors; nine items) and neighborhood connection (i.e., to what extent participants connected to their neighbors; five items) and was rated on a 5-point Likert scale (1 = agree, 2 = somewhat agree, 3 = neither agree nor disagree, 4 = somewhat disagree, and 5 = disagree) [18, 36]. The scales were reverse-coded where appropriate, and the mean scores of both subscales were calculated.
Safety environment
Neighborhood safety on children’s travel routes to and from school was assessed by (1) whether there were any closed-circuit televisions (CCTVs); (2) whether there was a road section where children walked from school more than 100 m alone; and (3) whether parents and/or local residents voluntarily watched out for children on streets to/from school on a daily basis. Response options were ‘yes,’ ‘no,’ and ‘not sure.’ One general response to the CCTV item was recorded per participant, which applied to both the ‘TO’ and ‘FROM’ school models (see the Statistical Analysis section for further details). Walks over 100 m alone were specifically asked on return from school, but its response was also used for the TO school model given that both situations were likely to be similar. Different responses between to and from schools were recorded for the safety volunteer item.
Parent-perceived influence of COVID-19 on school commuting
At the time of the survey (25–27 November 2020), the number of people infected with coronavirus disease 2019 (COVID-19) in Japan exceeded 2500 per day. In response to COVID-19 and the related physical distancing advice, it was assumed that children might have changed their school travel behavior. Therefore, we asked whether children’s modes of travel to and from school had been influenced by COVID-19, with response options of ‘yes,’ ‘no,’ and ‘not sure’.
After-school activity
Modes of travel from school can be diversified because more children participate in after-school activities. The frequency of going directly from school to public facilities for children (‘Jidou-kan’ where children spend time playing and studying) or after school lessons and clubs outside school was measured, and responses ranged from to 0–5 days per weekday.
Socio-demographic characteristics
Children’s school grades (1st–6th grade; ages 6–12 years), sex (male or female), and the number of cars owned by the household (1–5, 6, or more) were assessed [18, 37, 38].
Statistical analysis
Descriptive statistics, which included numbers and percentages, were calculated for all variables. For continuous variables, t-tests were used to test the differences in the means between children who attended schools in NTs and those who did not. Associations between neighborhood environments (built environment, social environment, and safety) and children’s school travel modes were examined using multinomial logistic regression analysis because of their non-linear relationships [39, 40]. In each outcome model (TO and FROM school), NT and neighborhood walkability were computed separately as explanatory variables due to a collinearity issue between these variables. Thus, four models were tested: TO school (1) NT model, (2) neighborhood walkability model, FROM school (3) NT model, and (4) neighborhood walkability model. Distance to school, population density, parent-perceived influence of COVID-19, after-school activities, and socio-demographic characteristics were used as control variables. Pseudo R-squared for the multinomial logistic regression model (Cox & Snell, Nagelkerke, and McFadden) was calculated to assess the model performance. The significance level was set at P < 0.05. All statistical analyses were conducted using IBM SPSS Statistics 26 (IBM Corp., Armonk, NY, USA).