The data used in the present study were extrapolated from a larger research project examining the effects of the environment and exercise on psychological health. Part of the project was conducted using an online questionnaire administered to participants in the 150,000 person Harris Poll panel of Great Britain. The research was approved by the University of Essex Research Ethics Committee and participants provided informed consent. Participants were selected at random from the base sample and invited by email to take part in the survey (n = 22,950). Data from the responding sample were collected over a 2 week period in late September 2011. Data collection was closed after 2 weeks as it reached the requested number of respondents.
This process yielded a sample of 2079 working age adults. In the current study, data were available for 1988 working age adults (997 males) ranging from 22 to 65 years (M = 43.19, SD = 11.46), which is the higher than the UK median of 39 years [33]. Only employed individuals were selected for this research in order to control for the impact of active commuting on visiting LGS and PA levels; 69.8 % were in full-time employment, 18.1 % were in part-time employment, and 12.2 % were self-employed.
The UK Meteorological Office [34] reported that in July and August 2011, mean temperatures were 0.5 to 1.0 °C below average across most of the UK. In contrast, during September, 2011 – during data collection – the mean temperatures were around 1.1 °C above average, making it the sixth warmest September in 100 years. Throughout September, most of England experienced below average rainfall; some parts of Northern England and Scotland, however, received over 50 % more rainfall than average [34].
Self-reported health was assessed with a single item which asked “How would you rate your health in the last month?” Participants responded on a Likert scale from “1 = Terrible” to “7 Excellent”. This was included as a covariate in all statistical analyses alongside age and gender.
Objective representation of the local environment was given as % of LGS available near home. This was calculated to ward level (primary unit of electoral geography), using participants’ home postcodes and Geoconvert (an online geography matching and conversion tool) [35]. For % of LGS, ward coded data were then entered into a database, available from CRESH.org.uk, which has previously been described [36]. In brief, the database used general land use across England, supplemented with a second database covering Scotland, Northern Ireland, and Wales and the coordination of information placed on the environment database [37]. The database provided specific % of LGS, including all vegetated areas larger than 5 m2 in area (excluding domestic gardens) for each ward in the UK. Green spaces included ranged from transport verges (narrow strip of land between carriageway and road boundary) and neighbourhood greens, to parks, playing fields and woodlands.
Perceived access to LGS was assessed by asking participants “How easy is it to get to the green space local to your home?” Participants responded from 1 = “Very difficult” to 7 = “Very easy”. Perceived quality of LGS was assessed with a single item that asked “How would you rate the quality of your local accessible green spaces that are close to your home?” Participants responded from 1 = “Terrible” to 7 = “Excellent”.
NR was assessed using two sections of the NR Scale (NRS; [38]). The self and experience factors were extrapolated to form the NRS-14. The self and experience factor were used to reflect both how strongly people identify with the natural environment and the attraction people have to nature. The perspective factor of the NRS was excluded as we were not interested in global issues such as conservation and species survival rates. Participants were asked to report how they felt about 14 phrases that described their relationship with nature. Examples items included, “Even in the middle of the city, I notice nature around me”, and “I am not separate from nature, but part of nature”. Participants responded using a Likert scale format ranging from 1 = “disagree strongly” to 5 = “agree strongly”. Where appropriate, responses were reversed so that higher scores indicated a greater NR. NR was recorded as a mean of 14 items.
Visit frequency to LGS was assessed by asking participants “How often do you visit the green space closest to your home?” This was rated from 1 = “Every day” to 7 = “Never visit my LGS or any other green spaces”. This score was then reversed scored so that a higher frequency of visits was represented by a higher numerical value. Participants also indicated via multiple choice selection how they usually travelled to LGS, and how long it usually took them.
Self-reported PA levels were recorded using a short-form version of the International Physical Activity Questionnaire (IPAQ-SF, [39]). Participants were required to indicate how many days they undertook PA activity for more than 10 min. Subdomains were vigorous, moderate and walking. Furthermore, participants reported how many hours and minutes they usually spent on these activities on one of those days. Additionally, participants reported how many hours and minutes they would usually spend sitting on a week day.
Raw data were converted into weekly PA levels using IPAQ-SF scoring guidelines [40]. The raw data were calculated into a weekly score described as multiples of the resting metabolic rate (METs). As recommended by IPAQ scoring guidelines, some of the raw data was truncated to reduce potential outliers. Above 180 min in all categories is considered to be unlikely, suggesting participants’ misinterpreted the question. In accordance with guidelines [40], therefore, all moderate minutes that were between 180 and 299 were reduced to 180; those above 299 were divided by seven. Also, vigorous minutes over 180 were divided by seven and walking minutes over 180 were reduced to 180. For the data analysis, participants were dichotomised according to whether they achieved at least 600 MET.min per week or not. Those participants who achieved below 600 MET.min per week in total were classified as not meeting the current minimum requirements for a healthy lifestyle (in accordance with [41]) and in the low category using IPAQ scoring guidelines [40].
A number of variables were included in the study as covariates: age, subjective health, gender, road coverage, environmental deprivation, and active travel to both work and LGS. Environmental Deprivation (at ward level) was obtained from a database that is available on CRESH.org [42]. In summary, ward level measurements were calculated for a variety of environmental dimensions that impact upon health (air pollution, climate, UV radiation, industrial facilities, and green space). Each ward was given a score from −2 to +3, with +3 indicating most deprived environments. For this study, scores of environmental deprivation were reversed so that the most deprived areas had the lowest score.
Road Coverage was calculated by cross referencing ward codes against general land use database [43] across EnglandFootnote 1 to give the amount of road coverage in each ward. This was converted to a percentage of the total land area in each ward. For both environmental deprivation and road coverage, participants’ home post codes were converted to wards using Geoconvert.
Active travel to work was assessed by asking participants “How do you usually travel to work? Tick all that apply”. Any participant who ticked walk or cycle were classified as active commuters. Active travel to LGS was assessed by asking participants “How do you usually travel to your local green space? Tick all that apply” Any participants who ticked walk or cycle were classified as active travellers to LGS.
All data analysis was carried out using IBM SPSS Statistics 20. Three regression models were run. First, an ordinal regression model was run to determine whether objective (% LGS) and subjective (perceived access, perceived quality, and NR) measures predicted frequency of visits to LGS. Additional demographic, objective, and subjective variables were included as covariates in the model (see Table 1).
Second, a binary logistic regression was run to determine whether objective (% LGS) and subjective (perceived access, perceived quality, and NR) measures predicted the likelihood of meeting current UK PA guidelines. Additional demographic, objective, and subjective variables were included as covariates in the model (see Table 2).
Finally, another binary logistic regression was run to determine if visit frequency to LGS predicted the likelihood of meeting current UK PA guidelines. Age, gender and health were included as covariates in the third model. Nagelkerke R2 tests were run to assess how much of the variance in the outcomes could be accounted for by the models. Statistical significance was accepted at p < 0.05 throughout the analyses.