Study background and evaluation
An opportunity for a natural experiment arose through discussions with a state-based government organisation (Parks Victoria) which manages State parks, reserves, waterways and other public land in Victoria, Australia. Parks Victoria was planning to install a play-scape (a play area designed with the intent of bringing children and accompanying adults back to nature) in a large metropolitan park.
The Recording and EValuating Activity in a Modified Park (REVAMP) study was designed to evaluate the impact of the park modification by using multiple measures to comprehensively assess park visitation and park-based physical activity in the intervention park and in a control park. More detailed information on the study methods have been provided elsewhere [12]. Measures included observational data of park visitors, objective monitoring of path usage within the parks and of vehicles entering on-site carparks, intercept surveys with adult park visitors, and two cross-sectional surveys with local residents.
Baseline assessments were conducted in April–May (Autumn) 2013 (T1), the park improvement occurred between September 2013–February 2014, first follow-up measures were conducted in April–May 2014 (T2) and second follow-up measures were conducted in April–May 2015 (T3). Each data collection took place at the same time of the year to account for potential seasonal effects.
Study setting
The intervention park (329 ha) is located 28 km north-west of Melbourne’s central business district (CBD) in a low SES area. The control park (120 ha) is located 22 km east of Melbourne’s CBD in a high SES area and is approximately 35 km from the intervention park via the road network. It was not possible to find a matching large park in a disadvantaged area with similar features to the intervention park that was not undergoing any refurbishment during the study period. Despite the differences in overall size and SES, at baseline these two parks provided similar infrastructure and settings for being active, such as extensive walking/cycling paths, grassy open space areas and basic playground equipment. In addition, both parks had other supportive amenities to encourage visitation such as toilets, car-parking and a variety of picnic shelters, tables and barbeque areas. More details on the park features have been provided elsewhere [23].
The park refurbishment
The refurbishment at the intervention park involved the installation of an innovative AUD$1.1 million play-scape suitable for children of all abilities that was designed by a landscape architect sourced by Parks Victoria. The new equipment included a large 360 degree swing, traditional swing set, maze, rockers, sandpit, nature play area, climbing equipment, landscaping, and various sculptures and was designed to be accessible for children with disabilities. The play-scape was also designed to encourage visitors to connect with both the natural environment and the significant indigenous cultural heritage of the region with references to local flora, fauna, past farming practices and key indigenous stories throughout the play-scape. Prior to refurbishment, the area where the play-scape was built was an open space area with no features or amenities. The playground at the control park was an older style adventure playground which included play equipment such as: slides, swings, climbing equipment, fireman’s pole, and swing bridges. See Additional file 1 for images of the new play-scape at the intervention park and the playground at the control park.
Measures
Observations of park visitors
Direct observations of park visitors were conducted using a modified version of the System for Observing Play and Recreation in Communities (SOPARC) to obtain counts of the number of people in the park and the activity in which they were engaging [24]. SOPARC is a reliable, objective observation tool for assessing physical activity in community settings that is often used to specifically assess visitation and physical activity in parks [19]. It is based on momentary time sampling and involves undertaking systematic scans (an observation sweep moving from left to right) of each participant within a target area at a particular time. Prior to each time-point, observers were trained to use SOPARC in a classroom workshop and on-site parks visits. Strong inter-rater reliability was obtained following the training; 92% of scans at T1, 96% at T2 and 99% at T3 had at least 80% agreement between the observers and 86% of scans at T1, 94% at T2 and 95% at T3 had 100% agreement.
Research staff conducted observation scans of pre-determined target areas that were identified after discussions with the park rangers to determine the most highly visited areas and included, for example, the playground site, walking/cycling paths, grassy open spaces, shelters and picnic areas. There were 10 target areas in each park at T1. At T2 and T3, the area where the play-scape was installed was split into five target areas to enable more accurate observations of play-scape users. This resulted in 14 target areas at the intervention park at T2 and T3. The days and times of data collection were the same for both parks at each time-point (eight days, including four weekdays and four weekend days). During weekdays, observations were conducted every hour from 7:30 am-4:30 pm (except for one day at T1 when observations concluded at 1:30 pm due to rain), and on weekend days every hour from 8:30 am-4:30 pm. At T1 a total of 730 scans at each park (10 target areas * 73 time-points) were completed. At T2 and T3, 1064 scans (14 target areas * 76 time-points) were completed at the intervention park and 760 scans (10 target areas * 76 time-points) at the control park. This equated to a total of 5108 scans across both parks.
During each scan, research staff recorded each individual in view within their target area according to: their estimated age group (i.e. child (1–12 yrs), teen (13–20 yrs), adult (21–59 yrs), or older adult (60 yrs.+)); sex (male or female); and the activity they were engaged in (lying down or sitting, standing, moderate activity (e.g. walking), or vigorous activity (e.g. jogging, cycling)).
Observations were not conducted on days of forecasted rain; however, unexpected rain showers occurred on some days. The variation in average hourly temperature and rainfall during the observation periods (i.e. 7.30 am-4.30 pm on weekdays) between the intervention and control parks and across the three time-points was minimal (data obtained from the Bureau of Meteorology). The average hourly temperature and rainfall during the observation periods at the intervention and control parks are presented in Additional file 2.
Electronic path monitors and car traffic counters
Both parks contained a network of sealed paths. Electronic path monitors were used to record counts of people walking and cycling on two pre-selected paths on the same days observations were conducted. The monitors were positioned at areas that were used most frequently and were most comparable between the two parks. Further details are provided elsewhere [12]. The path monitors were set up at 7:30 am on weekdays and 8:30 am on weekend days and were removed at 4:30 pm each day. Total counts for the two monitors at each park were calculated for the eight days of data collection at each time-point.
There was one main entrance to each park. At both parks a traffic counter was located at this entrance to record the number of vehicles entering and exiting the parks (hourly counts) on the days when park observations were conducted. Total counts of traffic entering each park from 7 am-5 pm were calculated for the eight days of data collection at each time-point.
Intercept surveys
Face-to-face intercept interviews were completed with English-speaking adult park visitors on days when observations were conducted (for logistic reasons it was not possible to have translators available in the park). The intercept surveys provided an opportunity to gain more detailed information about the park visit than could be obtained from the observations. Trained, clearly identifiable research assistants approached park users in the specified target areas, explained the study and all ethical considerations, and invited participation. At T1, 794 park visitors completed an interview (75.3% of those approached, excluding 201 park visitors who had already been intercepted); at T2, 1158 park visitors completed an interview (71.2% of those approached, excluding 293 park visitors who had already been intercepted at T2); and at T3, 1043 park visitors completed an interview (74.3% of those approached, excluding 371 park visitors who had already been intercepted at T3).
Park visitors were asked their age and sex and how often they had visited the intervention/control park in the past three months (daily, 2–3 times/week, once/week, 2–3 times/month, once/month, <once/month, have not visited in past three months). For the purpose of analysis these response options were collapsed to: ≥once/week; once/month to 2–3 times/month; or <once/month. They also reported their usual activity levels during visits to the park in the past three months (mostly sitting, mostly light activities, mostly moderate activities, or mostly vigorous activities).
Participants were also asked if they had a child(ren) aged 2–15 years, and if so, they were asked to consider the child in the age range who had the next birthday and report that child’s age and sex and how often that child had visited the intervention/control park in the past three months. Response options were the same as those described above for adults.
Resident surveys
Cross-sectional surveys were completed by adult residents at T1 and T3. These surveys provided a population estimate of park visitation rather than relying solely on observation or park intercepts, which only captures visitors and may also capture repeat visitors. Recruitment was via two methods: 1) families with children attending pre-schools, primary and secondary schools located within 3 km of each park; and 2) a postal survey from the local City Council to households located within 5 km of each park. At T1, 9694 surveys were delivered, 37 were returned to sender (no longer resided at that address) and removed from the denominator, and 1487 surveys were returned completed (15.4% response rate; 15.1% intervention park, 15.7% control park). At T3, 9537 surveys were delivered, 44 were returned to sender and removed from the denominator, and 1460 were returned completed (15.4% response rate; 14.1% intervention park, 16.6% control park).
The survey included socio-demographic variables (age, sex, country of birth, number of children, highest level of education, employment status, marital status, and dog ownership). Participants also reported how often they had visited the intervention/control park in the past three months (daily, 2–3 times/week, once/week, 2–3 times/month, once/month, <once/month, had not visited in past three months). These response options were collapsed to: ≥once/week; once/month to 2–3 times/month; or <once/month. They also reported how long they were usually active on each park visit in the past three months (minutes) and their usual activity levels during visits to the park in the past three months (mostly sitting, mostly light activities, mostly moderate activities, or mostly vigorous activities). Time spent (minutes) in transportation and leisure-time physical activity in the last seven days was examined using the long form of the International Physical Activity Questionnaire (IPAQ-L) [25].
Respondents with a child(ren) aged 2–15 years living in the household, were asked to complete proxy-report survey questions on behalf of their child (next birthday method) regarding age and sex, how often their child had visited the intervention/control park in the past three months, how long their child was usually active for on each park visit in the past three months, and their child’s usual activity levels during visits to the park in the past three months. Response options were the same as those described above for adults.
Statistical analysis
Descriptive statistics for overall observation visitor counts, observation visitor counts in the new play-scape at the intervention park and playground at the control park, path monitor counts, traffic counts, intercept surveys and resident surveys for the two parks at each time-point were calculated.
Analyses of the observation and traffic data included three separate count outcomes with hourly counts as the unit of analysis; overall number of visitors observed, number of people observed in MVPA and traffic counts. There were insufficient cases to run inferential analyses for the path monitor data as path monitor counts were recorded as total daily counts. For each of these three outcomes, a multilevel negative binomial regression model was conducted with random intercepts for measurement days (i.e. accounting for clustering of hourly observations within measurement days at each park). Models included main effects of time (T1/T2/T3) and park (intervention/control), as well as a time by park interaction. The time by park interaction was used to assess the effect of park refurbishment. As the time factor had three values, two interaction coefficients were produced for each model. T1 (2013) was set as the reference value for time and the control park set as the reference park. These two coefficients represented differences in outcomes at the intervention park; firstly between T1 and T2, and secondly between T1 and T3, relative to the control park. As the outcome variables were counts, the interaction effects have been reported as Incidence Rate Ratios (IRRs). The models adjusted for the following covariates: hourly temperature; hourly rainfall; and whether it was a weekday or weekend day. All analyses were conducted using Stata/SE 14 (StataCorp, TX).
Logistic regression models were used to test the effect of the park refurbishment on odds of regular visitation (>once/week over the past three months) among adult participants (and their children) who completed the intercept surveys and, separately, among adult participants (and their children) who completed the resident surveys and had visited the intervention/control park in the past three months. Models included main effects for park and time-point, their interaction, and potential confounders of age and sex. Statistical significance of the interaction term was used to determine if the outcome varied between the two parks at T2 (intercept surveys only) or T3 relative to their baseline difference. This method is also known as difference-in-difference analysis [3]. Equivalent logistic regression models were also used to examine effects of the park refurbishment on odds of adult participants (and their children) engaging primarily in MVPA while in the park, among those who reported (on behalf of themselves and their child) that they had visited the intervention/control park in the past three months. Finally, among those who had visited the park in the past three months, linear regression models (with main effects for park and time-point as well as their interaction) were used to examine effects of the park refurbishment on the time in minutes adult respondents (on behalf of themselves and their child) reported they were usually active on each park visit.