Study design and participants
A cross-sectional study was undertaken for 11,404 adults aged 25 and over who (i) were residents of the Perth metropolitan area, (ii) consented to data linkage and (iii) completed the Western Australian Health and Wellbeing Survey between 2003 and 2009. This monthly computer-assisted telephone interview was administered by the Western Australian Department of Health and responses were obtained for a stratified random sample of the state population (N = 1,959,088; 2006 Census). There were 15,502 adult residents in Perth who completed the survey between 2003 and 2009. Of those participants, 11,404 (74%) participants granted permission for data linkage.
Outcome variables
Self-report of prior medically diagnosed heart disease and stroke was obtained from the Health and Wellbeing Survey. Hospital records were obtained from the Western Australian Department of Health. Coronary heart disease and stroke were identified from these records as a primary diagnosis coded I20 – I25 and I60 – I68 according to the International Classification of Diseases (ICD 10). Participants were considered to have been hospitalized for heart disease if the admission occurred within a three year window centered on the year that the participant completed the Health and Wellbeing Survey.
Greenness
Greenness was measured using the Normalized Difference Vegetation Index (NDVI) derived from annually updated Landsat TM remote sensing imagery taken during summer. NDVI provides an indication of the presence and condition of green vegetation with values typically ranging from −1 to +1. Values of −1 generally represent water, while values of zero (−0.1 to 0.1) correspond to bare surfaces such as rock, sand, rooftops and roads. Higher values (0.2 to 0.4) represent grassland or bush land and values of +1 represent healthy green vegetation [11]. Water features were first removed before the NDVI was calculated.
Neighborhoods around participants’ homes were defined using 1600 m (network distance) service areas. The rationale for selecting a 1600 m service area was based on the assumption that physical activity is the most likely pathway by which neighborhood greenness is associated with cardiovascular disease risk. A 1600 m service area represents how far a participant could walk from home at a moderate to vigorous intensity pace, within 15 minutes, which equates to half the recommended level of daily physical activity for adults[12]. That is, the daily recommended level of physical activity of 30 minutes would be attained for a return trip. Moreover, a 1600 m service area has been shown to be a critical distance for examining the relationship between parks and walking. Sugiyama et al. found that although proximity to parks was generally associated with walking, it was the presence of parks within 1600 m that was most associated with sufficient walking (>150 minutes/week)[9].
The mean and standard deviation of NDVI values were calculated for the 1600 m service areas. The simulation shown in Figure 1 demonstrates that the mean NDVI describes the absolute level of greenness, while the standard deviation of NDVI captures the heterogeneity in the distribution of greenness. In reality, the green and non-green areas (pixels) are not randomly distributed as they cluster, such as in parks or along roads. Figures 2 and 3 illustrate the spatial variability in greenness for two study participants and their respective service areas. The standard deviation of NDVI values within the service area can be interpreted as an alternative expression of land-use mix of anthropogenic and natural features. The service area in Figure 2 indicates lower levels of greenness than the service area in Figure 3 but has a higher level of variability. In general, high levels of variability in greenness will occur when the neighborhood contains both high NDVI (green) areas and low NDVI (non-green areas). In this example, the higher level of variability is due to both the strong prevalence of vegetation (high NDVI) and the presence of commercial land-use (low NDVI).
The mean and standard deviation of NDVI values were analyzed as both continuous linear variables (scaled by their interquartile range) and categorical variables (tertiles). The rationale for using tertiles is provided in the Supplementary Material (Additional file 1). Briefly, cut-points at tertiles provided a compromise between capturing the pattern of the association and ensuring sufficient data within each category. Tertiles also provide an equal amount of data in each category and allow interpretation relative to “low”, “medium” and “high” values.
Statistical analysis
Adjustments were made using representations of a range of well-established factors [4] obtained from the Health and Wellbeing Survey: sociodemographics (age, sex, possession of a healthcare card, education, household income), biological factors (non-gestational diabetes, BMI, hypertension, high cholesterol), behavioral factors (daily serves of fruit and vegetables, risky drinking in the last month (>6 standard drinks for men, >4 standard drinks for women), and smoking (never versus ever smoked)), and a proxy for air quality.
A broad class of anthropogenic air pollutants have been causally linked with endothelial dysfunction and vasoconstriction, increased blood pressure, systemic inflammatory responses, oxidative stress, and the progression of atherosclerosis [13]. Traffic-related air pollution specifically has been recently associated with fatal and non-fatal coronary heart disease events [14]. In a synthesis of studies conducted by the Health Effects Institute it was concluded that the exposure zone for traffic-related air pollution is between 300 m and 500 m from a major road, beyond which concentrations decrease to background levels [15]. Adjustment was made for the total length of main roads within a 400 m service area as a proxy for exposure to ambient air pollution. Main roads were defined as all roads traversed by ≥6000 vehicles/day. The metropolitan road network was obtained from Landgate (the state government source of land information and geographic data) at four time points (2005, 2006, 2008, 2009). The year of road network data was matched to participant’s year of interview for the Health and Wellbeing Survey.
Socio-demographic factors, biological factors, behavioral factors and environmental factors were cumulatively included as adjustment variables in multiple regression models. Logistic regression was applied using R 2.13 and road exposure was calculated using ArcMap 9.2.
Ethics
Approval was obtained from the Human Research Ethics Committees of the Western Australian Department of Health and The University of Western Australia (#2010/1). This research conforms to the ethical principles for medical research of the Declaration of Helsinki.