README.md
In STBrinkmann/DRIGLUCoSE: DRI-GLUCoSE
DRI-GLUCoSE
As elaborated in our recent analyses (Walker et al. 2019; Scarpone et
al. 2020), nearly all previous studies in the literature use either
census unit boundaries or simple buffer zones to measure an individual’s
built environment (BE) exposures or to characterize their local
socioeconomic status (SES) (Gong et al. 2014; Fuertes et al. 2014).
Therefore, we present a distance-weighted, network-based model for
quantifying the combined effects of local greenspace and SES on diabetes
risk, from which we derive an area-based Diabetes Risk Index of
Greenspace, Land Use and Socioeconomic Environments (DRI-GLUCoSE). The
goal of the DRIGLUCoSE package is to provide a public package
containing functions and code used in the development of the DRI-GLUCoSE
Index(Walker et al. 2022).
Installation
You can install the latest version of DRIGLUCoSE from GitHub with:
remotes::install_git("https://github.com/STBrinkmann/DRIGLUCoSE")
Once installed, the library can be loaded as follows:
library(DRIGLUCoSE)
Methods
One key purpose of this package is, to provide functions for route
networked derived isochrones. For that purpose we have provided a sample
sf object of 2 points in Erlangen, Germany.
data(Erlangen)
Erlangen
## Simple feature collection with 2 features and 2 fields
## Geometry type: POINT
## Dimension: XY
## Bounding box: xmin: 35199.46 ymin: -159433.5 xmax: 36281.59 ymax: -159243.2
## Projected CRS: ETRS89 / LCC Germany (N-E)
## # A tibble: 2 x 3
## tag Speed geom
## <dbl> <dbl> <POINT [m]>
## 1 1 78.5 (35199.46 -159433.5)
## 2 2 79.8 (36281.59 -159243.2)
Census variables
In our analysis we acquired data of the Canadian census dissemination
areas. It has been converted to a shapefile (sf) with one column per
census variable. To demonstrate we use the following randomly generated
data:
set.seed(1234)
census <- sf::st_make_grid(
# Use Sample Data and apply 25 minutes buffer (Speed[m/min] * 25[min])
Erlangen %>% dplyr::mutate(geom = sf::st_buffer(geom, Speed*25)),
cellsize = 100
) %>%
sf::st_as_sf() %>%
dplyr::mutate(census_var_a = sample(1:1000, n(), replace = TRUE),
census_var_b = sample(1000:10000, n(), replace = TRUE),
census_var_c = sample(100000:150000, n(), replace = TRUE)) %>%
dplyr::rename(geom = x)
census
## Simple feature collection with 2142 features and 3 fields
## Geometry type: POLYGON
## Dimension: XY
## Bounding box: xmin: 33236.96 ymin: -161396 xmax: 38336.96 ymax: -157196
## Projected CRS: ETRS89 / LCC Germany (N-E)
## First 10 features:
## census_var_a census_var_b census_var_c geom
## 1 284 2009 114811 POLYGON ((33236.96 -161396,...
## 2 848 3318 139961 POLYGON ((33336.96 -161396,...
## 3 918 3954 112673 POLYGON ((33436.96 -161396,...
## 4 101 3359 137505 POLYGON ((33536.96 -161396,...
## 5 623 3107 109201 POLYGON ((33636.96 -161396,...
## 6 905 2630 131656 POLYGON ((33736.96 -161396,...
## 7 645 6512 135789 POLYGON ((33836.96 -161396,...
## 8 934 9945 115006 POLYGON ((33936.96 -161396,...
## 9 400 3583 111379 POLYGON ((34036.96 -161396,...
## 10 900 7778 137576 POLYGON ((34136.96 -161396,...
Greenspace
In our analysis we acquired LANDSAT images through the United States
Geological Survey’s EarthExplorer platform
(https://earthexplorer.usgs.gov/). The Normalized Difference
Vegetation Index
(NDVI)
is used as a metric to model greenspace exposure. Pre-processing of the
LANDSAT images and NDVI calculation has been conducted using the
LS_L1C function:
DRIGLUCoSE::LS_L1C(l1c_path = "docs/LC08_L1TP_193026_20200423_20200508_01_T1_small/",
out_dir = "docs/LS_PreProcessed",
# Use Sample Data and apply 25 minutes buffer (Speed[m/min] * 25[min])
sf_mask = DRIGLUCoSE::Erlangen %>%
dplyr::mutate(geom = sf::st_buffer(geom, Speed*25)),
cores = 20)
## Project raster
## DN to TOA Reflectance
## class : RasterStack
## dimensions : 122, 151, 18422, 8 (nrow, ncol, ncell, nlayers)
## resolution : 30, 30 (x, y)
## extent : 33493.69, 38023.69, -161164.2, -157504.2 (xmin, xmax, ymin, ymax)
## crs : +proj=lcc +lat_0=51 +lon_0=10.5 +lat_1=48.6666666666667 +lat_2=53.6666666666667 +x_0=0 +y_0=0 +ellps=GRS80 +units=m +no_defs
## names : Blue, Green, Red, NIR, SWIR1, SWIR2, NDWI, NDVI
## min values : 0, 0, 0, 0, 0, 0, -1, -1
## max values : 0.2020575, 0.2322532, 0.3076383, 0.5424371, 0.4233773, 0.3753066, 1.0000000, 1.0000000
Exposure Model
In order to estimate each participant’s potential exposures to
greenspace and local SES, we (i) mapped age- and sex-specific walkable
zones around their residential address, and (ii) applied a negative
logit weighting function, such that the estimated effect of greenspace
or SES decreases as distance from the home increases.
(i) Road network data and isochrones
In order to compute network-based distance metrics, we acquired street
data from OpenStreetMap using the R-package osmdata (Padgham et al.
2017). Road types not suitable for walking were removed (e.g.,
motorways). Network data were topologically corrected and split into
\~20 metre-long segments using the R package nngeo (Michael Dorman
2020).
erlangen.osm <- DRIGLUCoSE::osm_roads(x = Erlangen, dist = 20,
speed = "Speed", cores = 2)
This network data was used to derive walking distance buffers for each
participant, based on walking speed. Starting from each participant’s
place of residence, we computed network-constrained buffers with an
off-road width of 40 meters, running in 2-minute increments from 0 to 20
minutes, using the A*-algorithm (Hart, Nilsson, and Raphael 1968). This
therefore resulted in each participant having ten concentric isochrones,
the sizes of which are a function of individual walking speed and road
network.
erlangen.isodistances <- DRIGLUCoSE::isodistances(x = Erlangen,
road_network = erlangen.osm,
tag = "tag", speed = "Speed",
isochrones_seq = seq(2, 20, 2),
cores = 2)
erlangen.isochrones <- DRIGLUCoSE::isochrones(x = erlangen.isodistances,
buffer = 40, cores = 2)
erlangen.isochrones
## Simple feature collection with 20 features and 2 fields
## Geometry type: MULTIPOLYGON
## Dimension: XY
## Bounding box: xmin: 34034 ymin: -160951 xmax: 37750 ymax: -157782
## Projected CRS: ETRS89 / LCC Germany (N-E)
## # A tibble: 20 x 3
## time tag geom
## * <dbl> <dbl> <MULTIPOLYGON [m]>
## 1 2 1 (((35292.18 -159582.1, 35292.75 -159584.1, 35293.22 -159586.1, 3~
## 2 4 1 (((35314.38 -159725.3, 35313.74 -159727.3, 35312.99 -159729.2, 3~
## 3 6 1 (((35348.37 -159830.5, 35347.9 -159831.2, 35347.44 -159832, 3534~
## 4 8 1 (((35282 -159969.8, 35281.96 -159971.9, 35281.8 -159974, 35281.5~
## 5 10 1 (((35249.78 -160125.6, 35249.57 -160125.5, 35248.86 -160125.1, 3~
## 6 12 1 (((35264.36 -160300.8, 35263.71 -160300.8, 35263.06 -160300.9, 3~
## 7 14 1 (((35379.71 -160416.5, 35380.79 -160418.3, 35381.78 -160420.1, 3~
## 8 16 1 (((35412 -160571, 35411.95 -160573.1, 35411.78 -160575.2, 35411.~
## 9 18 1 (((35430.23 -160426.8, 35430.28 -160426.3, 35430.28 -160426.3, 3~
## 10 20 1 (((35546 -160803.2, 35546.09 -160803.5, 35546.22 -160804.2, 3554~
## 11 2 2 (((36385.71 -159372.9, 36384.21 -159374.3, 36382.64 -159375.7, 3~
## 12 4 2 (((36379.78 -159545.8, 36379.83 -159546.5, 36379.9 -159547.1, 36~
## 13 6 2 (((36368.01 -159688.3, 36366.95 -159690.1, 36365.8 -159691.9, 36~
## 14 8 2 (((36655.01 -159679.5, 36655.61 -159681.6, 36656.11 -159683.6, 3~
## 15 10 2 (((36520.27 -159907.5, 36518.33 -159908.3, 36516.36 -159909, 365~
## 16 12 2 (((36053.07 -160058.5, 36051.72 -160056.9, 36050.46 -160055.3, 3~
## 17 14 2 (((36832.82 -159976.8, 36832.8 -159977, 36832.79 -159977.1, 3683~
## 18 16 2 (((36016.71 -160325, 36016.5 -160325.4, 36016.23 -160325.9, 3601~
## 19 18 2 (((36020.67 -160327.7, 36022.64 -160327, 36024.65 -160326.4, 360~
## 20 20 2 (((36689.36 -160321.8, 36690.36 -160320, 36691.45 -160318.2, 366~
Figure 1 shows isodistances of the two points of the sample data in
Erlangen, Germany.
(ii) Distance-weighting
In order to account for the diminishing effect of SES and greenspace
exposure as distance increases, we fitted a logit function to weight
each incremental isochrone, such that the influence of a variable
decreases with increasing distance from the household, i.e., features
that are farther away have less influence than nearby features, as
illustrated in Figure 2. A logit function was selected as it
heuristically approximates a suitable distance-decay function (Bauer and
Groneberg 2016; Jia, Wang, and Xierali 2019). The distance-weighting is
separated in two parts, first the logit function (1) that is used for
both SES and greenspace variables, and second the proportional weights
function (4) that is only applied on SES variables.
Each isochrone
is assigned a distance weight
,
calculated as the integral of the logistic distance decay function
(2)
with
and
,
in the interval between the mean inner radius
and mean outer radius
of the isochrone (e.g. 2 to 4 minutes isochrones), normalized by the
integral from 0 to the outermost isochrone boundary
(e.g. 20 minutes isochrone). Weighted summary statistics to describe the
greenspace (e.g. mean or minimum NDVI) are thus described as (3)
For SES variables the proportional weights of the census areas within
the isochrone are further defined as (4)
with the proportion of the area of the intersection of the census area
and the isochrone
,
and the area of the isochrone
.
The weighted value of the SES variable
in the census area
is then defined as (5)
Figure 2 visualizes the different submodels used for distance-weighting
SES and greenspace. Fig. 2a shows the unweighted values of a SES
variable and fig. 2b has been calculated using (5), thus representing
the proportional weights of all intersections with the census areas and
isochrones. Greenspace is weighted as shown in fig. 2c using (3).
The distance-weighting for the LANDSAT derived NDVI raster (greenspace
exposure) is handled using LS_band_weightin, and SES distance- and
areal-weighting using census_weighting.
# Calculate sd, median, 5th percentile, 95th percentile and skew of NDVI values
NDVI_weighted <-
DRIGLUCoSE::LS_band_weighting(isochrones = erlangen.isochrones, tag = "tag",
landsat_list = dir("docs/LS_PreProcessed",
pattern = ".grd",
full.names = T) %>%
lapply(raster::brick),
stats = list("sd", "median",
list("percentile", 0.05),
list("percentile", 0.95),
"skew"),
b = 8, m = 0.6, cores = 2)
NDVI_weighted
## # A tibble: 2 x 6
## tag sd median X5_percentile X95_percentile skew
## <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 1 0.205 0.615 0.281 0.914 -0.145
## 2 2 0.104 0.540 0.360 0.714 -0.0122
census_weighted <- DRIGLUCoSE::census_weighting(isochrones = erlangen.isochrones,
tag = "tag", census = census,
b = 8, m = 0.6, cores = 2)
census_weighted
## # A tibble: 2 x 4
## tag census_var_a census_var_b census_var_c
## <dbl> <dbl> <dbl> <dbl>
## 1 1 562. 5323. 130970.
## 2 2 547. 5419. 124610.
Appendix
Figures
DRI-GLUCoSE scores for Vancouver (top) and Hamilton (bottom), ranging
from low risk (purple) to high risk areas (orange).
Tables
Table A.1: Logistic models for all multivariable models with odds ratios
(OR) for diabetes, 95% CI, and p-values.
Semi-Adjusted
Fully-Adjusted
Parameter
OR(bivariate)
OR(WHR-adjusted)
OR(BMI-adjusted)
OR(WHR-adjusted)
OR(BMI-adjusted)
DRI-GLUCoSE Score
0.37 (0.26-0.51,p<0.001)
0.41 (0.31-0.53,p<0.001)
0.46 (0.35-0.61,p<0.001)
0.45 (0.36-0.56,p<0.001)
0.50 (0.37-0.67,p<0.001)
Age (5 year-interval)
1.26 (1.18-1.34,p<0.001)
1.21 (1.15-1.27,p<0.001)
1.26 (1.20-1.33,p<0.001)
1.20 (1.15-1.65,p<0.001)
1.24 (1.18-1.31,p<0.001)
Sex: female
0.63 (0.50-0.78,p<0.001)
0.85 (0.71-1.00,p=0.056)
0.47 (0.39-0.57,p<0.001)
0.95 (0.81-1.11,p=0.498)
0.59 (0.48-0.71,p<0.001)
Obese (WHR)
5.72 (4.37-7.59,p<0.001)
5.39 (4.50-6.46,p<0.001)
5.06 (4.29-5.97,p<0.001)
BMI
1.14 (1.12-1.16,p<0.001)
1.14 (1.12-1.15,p<0.001)
1.13 (1.11-1.15,p<0.001)
Household income range
0.75 (0.70-0.81,p<0.001)
0.87 (0.82-0.92,p<0.001)
0.85 (0.80-0.90,p<0.001)
0.90 (0.86-0.95,p<0.001)
0.85 (0.80-0.91,p<0.001)
Neighbourhood type: urban
0.93 (0.71-1.21,p=0.605)
0.58 (0.47-0.72,p<0.001)
0.62 (0.50-0.77,p<0.001)
0.60 (0.50-0.72,p<0.001)
0.66 (0.52-0.84,p<0.001)
AHEI Score (E^1)
0.78 (0.70-0.88,p<0.001)
0.95 (0.88-1.03,p=0.197)
0.99 (0.89-1.09,p=0.771)
Recreation Met Score: >=525
0.58 (0.46-0.73,p<0.001)
0.82 (0.71-1.03,p=0.006)
0.85 (0.71-1.03,p=0.090)
Current/Former smoker: yes
1.65 (1.31-2.07,p<0.001)
1.55 (1.34-1.80,p<0.001)
1.38 (1.15-1.66,p<0.001)
Alcohol: <1 drink/day
1.17 (0.91-1.51,p=0.217)
1.81 (1.54-2.14,p<0.001)
1.34 (1.09-1.65,p=0.006)
To analyse the effect of socioeconomic status (SES) and greenspace (GS),
we further build multivariable models using the semi-adjusted model with
BMI as obesity measurent and tested different combinations for the index
variable.
Table A.2: Model Performance and odds ratios of the logistic models,
comparing combinations of socioeconomic status (SES) and greenspace (GS)
as index.
Metric
SES + GS
SES
GS
Probability Threshold *
0.47
0.50
0.48
Accuracy
0.75
0.72
0.74
Sensitivity
0.76
0.72
0.75
Specificity
0.65
0.68
0.68
Youden index
0.41
0.40
0.43
OR (95% CI, p-value)
0.46 (0.35-0.61, p < 0.001)
0.57 (0.42-0.76, p < 0.001)
0.42 (0.31-0.57, p < 0.001)
* Probability threshold used for predicting Diabetes. Values
equal or greater than this threshold are mapped as “No”.
Table A.3: Model Performance for all multivariable models.
Semi-Adjusted
Fully-Adjusted
Metric
OR(WHR-adjusted)
OR(BMI-adjusted)
OR(WHR-adjusted)
OR(BMI-adjusted)
Probability Threshold *
0.44
0.47
0.58
0.60
Accuracy
0.71
0.75
0.64
0.64
Sensitivity
0.72
0.76
0.62
0.62
Specificity
0.70
0.65
0.81
0.83
Youden index
0.42
0.41
0.43
0.45
* Probability threshold used for predicting Diabetes. Values
equal or greater than this threshold are mapped as “No”.
Summary Statistics
Table A.4: Baseline characteristics of the study population by diabetes
status.
Parameter
No Diabetes (N=4616)
Diabetes (N=509)
Total (N=5125)
Total included
4616 (90.1%)
509 (9.9%)
5125
City
Hamilton
2307 (87.5%)
331 (12.5%)
2638
Vancouver
2309 (92.8%)
178 (7.2%)
2487
Participant data
BMI
Mean (SD)
27.3 (5.4)
32.1 (6.4)
27.8 (5.7)
Median (Q1; Q3)
26.4 (23.8; 29.8)
30.9 (27.5; 35.7)
26.8 (24.0; 30.4)
Waist to hip ratio
Mean (SD)
85.2 (9.0)
93.6 (8.5)
86.0 (9.3)
Median (Q1; Q3)
85.2 (78.6; 91.6)
94.2 (87.8; 99.6)
86.0 (79.3; 92.6)
Obesity (WHR)
Low & Moderate
2787 (96.0%)
115 (4.0%)
2902
High
1829 (82.3%)
394 (17.7%)
2223
Age (years)
Mean (SD)
10.5 (1.9)
11.4 (1.7)
10.6 (1.9)
Median (Q1; Q3)
10.6 (9.0; 12.0)
11.6 (10.2; 12.6)
10.6 (9.2; 12.2)
Sex
Male
2076 (87.6%)
294 (12.4%)
2370
Female
2540 (92.2%)
215 (7.8%)
2755
Household Income Range
>90k
1793 (93.3%)
129 (6.7%)
1922
45k-65k
759 (89.8%)
86 (10.2%)
845
30k-45k
564 (85.2%)
98 (14.8%)
662
65k-90k
1010 (92.7%)
80 (7.3%)
1090
20k-30k
304 (81.3%)
70 (18.7%)
374
<20k
186 (80.2%)
46 (19.8%)
232
AHEI Score
Mean (SD)
3.8 (1.0)
3.5 (0.9)
3.7 (1.0)
Median (Q1; Q3)
3.8 (3.1; 4.5)
3.5 (2.8; 4.1)
3.7 (3.0; 4.5)
Physical Activity MET Score
under
1721 (87.3%)
250 (12.7%)
1971
over
2065 (92.3%)
172 (7.7%)
2237
Ever smoked
Mean (SD)
0.4 (0.5)
0.6 (0.5)
0.5 (0.5)
Median (Q1; Q3)
0.0 (0.0; 1.0)
1.0 (0.0; 1.0)
0.0 (0.0; 1.0)
Daily Drinker
>=1 drinks/day
1157 (91.0%)
114 (9.0%)
1271
<1 drink/day
2629 (89.5%)
308 (10.5%)
2937
Census variables
Neighbourhood type
suburban/rural
3607 (89.9%)
405 (10.1%)
4012
urban
1009 (90.7%)
104 (9.3%)
1113
Individual mean income (CAD/1000)
Mean (SD)
37.9 (13.9)
34.3 (12.1)
37.5 (13.8)
Median (Q1; Q3)
35.0 (29.0; 42.0)
32.0 (27.0; 38.0)
35.0 (29.0; 42.0)
Household median income (CAD/1000)
Mean (SD)
65.5 (21.5)
60.6 (20.5)
65.0 (21.4)
Median (Q1; Q3)
61.0 (51.0; 77.0)
57.0 (45.0; 71.0)
61.0 (51.0; 76.0)
Prevalence of low income (%)
Mean (SD)
9.4 (6.9)
11.3 (7.6)
9.6 (7.0)
Median (Q1; Q3)
8.1 (4.3; 13.7)
10.1 (5.2; 16.3)
8.2 (4.4; 14.0)
Commute Walking/Bicycle (%)
Mean (SD)
7.5 (6.9)
6.8 (5.8)
7.4 (6.8)
Median (Q1; Q3)
5.1 (3.0; 9.9)
5.3 (3.2; 8.6)
5.2 (3.0; 9.7)
Labour force participation rate (%)
Mean (SD)
66.6 (7.7)
65.3 (7.6)
66.5 (7.7)
Median (Q1; Q3)
66.8 (61.8; 71.7)
64.9 (60.7; 70.5)
66.6 (61.7; 71.6)
Gov’t transfer payments (%)
Mean (SD)
9.9 (5.2)
12.4 (6.3)
10.1 (5.4)
Median (Q1; Q3)
8.8 (5.7; 13.0)
11.5 (7.2; 16.1)
9.0 (5.8; 13.3)
Unemployment rate (%)
Mean (SD)
5.5 (2.5)
6.3 (2.9)
5.6 (2.6)
Median (Q1; Q3)
5.4 (3.8; 7.0)
5.9 (4.4; 7.7)
5.4 (3.8; 7.0)
Lone parent families (%)
Mean (SD)
14.8 (6.5)
17.3 (7.5)
15.0 (6.6)
Median (Q1; Q3)
14.0 (10.9; 18.4)
16.1 (11.9; 21.8)
14.1 (11.0; 18.8)
Education - No degree (%)
Mean (SD)
19.6 (9.2)
23.7 (10.4)
20.0 (9.4)
Median (Q1; Q3)
17.7 (12.8; 24.9)
21.0 (15.8; 31.6)
17.9 (13.1; 25.6)
Private Dwellings - Owned (%)
Mean (SD)
73.9 (17.8)
72.9 (17.6)
73.8 (17.7)
Median (Q1; Q3)
77.1 (62.1; 88.5)
75.6 (61.7; 87.0)
77.0 (62.1; 88.4)
Private Dwellings - Rented (%)
Mean (SD)
25.3 (17.2)
26.4 (17.0)
25.4 (17.2)
Median (Q1; Q3)
22.1 (11.2; 36.9)
24.0 (12.9; 37.7)
22.4 (11.4; 36.9)
Normalized Difference Vegetation Index (NDVI)
NDVI - Median
Mean (SD)
0.341 (0.089)
0.314 (0.084)
0.338 (0.089)
Median (Q1; Q3)
0.336 (0.291; 0.379)
0.313 (0.266; 0.360)
0.333 (0.288; 0.377)
NDVI - Standard Deviation
Mean (SD)
0.090 (0.026)
0.088 (0.024)
0.090 (0.026)
Median (Q1; Q3)
0.086 (0.071; 0.105)
0.084 (0.072; 0.100)
0.086 (0.071; 0.105)
NDVI - Min
Mean (SD)
0.206 (0.097)
0.182 (0.086)
0.204 (0.096)
Median (Q1; Q3)
0.202 (0.135; 0.262)
0.173 (0.118; 0.243)
0.198 (0.134; 0.260)
NDVI - max
Mean (SD)
0.495 (0.093)
0.465 (0.093)
0.492 (0.093)
Median (Q1; Q3)
0.484 (0.435; 0.546)
0.459 (0.401; 0.514)
0.482 (0.432; 0.543)
About
Package contributors
Brinkmann, Sebastian Tobias (Package creator and author) e-mail:
sebastian.brinkmann@fau.de
Große, Tim (Contributor)
Thesis authors
Walker, Blake Byron (1*)
Brinkmann, Sebastian Tobias (1)
Große, Tim (1)
Dominik Kremer (1)
Schuurman Nadine (2)
Hystad Perry (3)
Rangarajan Sumathy (4)
Teo Koon (4)
Yusuf Salim (4)
Lear Scott A. (5)
1: Community Health Environments and Social Terrains (CHEST) Lab,
Institut für Geographie, Friedrich-Alexander-Universität
Erlangen-Nürnberg, Wetterkreuz 15, 91052 Erlangen, Germany
*corresponding author
2: Department of Geography, Simon Fraser University, Burnaby, Canada
3: Spatial Health Lab, College of Public Health and Human Sciences,
Oregon State University, Corvallis, USA
4: Population Health Research Institute, McMaster University, Hamilton,
Canada
5: Faculty of Health Sciences, Simon Fraser University, Burnaby, Canada
Citation
citation("DRIGLUCoSE")
##
## To cite DRIGLUCoSE in publications use:
##
## Walker, B.B., Brinkmann, S.T., Große, T. et al. Neighborhood
## Greenspace and Socioeconomic Risk are Associated with Diabetes Risk
## at the Sub-neighborhood Scale: Results from the Prospective Urban and
## Rural Epidemiology (PURE) Study. J Urban Health 99, 506–518 (2022).
## https://doi.org/10.1007/s11524-022-00630-w
##
## A BibTeX entry for LaTeX users is
##
## @Article{,
## title = {Neighborhood Greenspace and Socioeconomic Risk are Associated with Diabetes Risk at the Sub-neighborhood Scale: Results from the Prospective Urban and Rural Epidemiology (PURE) Study},
## author = {Blake Byron Walker and Sebastian T. Brinkmann and T. Große et al.},
## journal = {J Urban Health},
## year = {2022},
## volume = {99},
## pages = {506–518},
## url = {https://doi.org/10.1007/s11524-022-00630-w},
## }
Bibliography
Bauer, Jan, and David A. Groneberg. 2016. “Measuring Spatial
Accessibility of Health Care Providers - Introduction of a Variable
Distance Decay Function Within the Floating Catchment Area (FCA)
Method.” *PloS One* 11 (7): e0159148.
.
Fuertes, Elaine, Iana Markevych, Andrea von Berg, Carl-Peter Bauer,
Dietrich Berdel, Sibylle Koletzko, Dorothea Sugiri, and Joachim
Heinrich. 2014. “Greenness and Allergies: Evidence of Differential
Associations in Two Areas in Germany.” *Journal of Epidemiology and
Community Health* 68 (8): 787–90.
.
Gong, Yi, John Gallacher, Stephen Palmer, and David Fone. 2014.
“Neighbourhood Green Space, Physical Function and Participation in
Physical Activities Among Elderly Men: The Caerphilly Prospective
Study.” *International Journal of Behavioral Nutrition and Physical
Activity* 11 (1): 40. .
Hart, Peter, Nils Nilsson, and Bertram Raphael. 1968. “A Formal Basis
for the Heuristic Determination of Minimum Cost Paths.” *IEEE
Transactions on Systems Science and Cybernetics* 4 (2): 100–107.
.
Jia, Peng, Fahui Wang, and Imam M. Xierali. 2019. “Differential Effects
of Distance Decay on Hospital Inpatient Visits Among Subpopulations in
Florida, USA.” *Environmental Monitoring and Assessment* 191 (Suppl 2):
381. .
Michael Dorman. 2020. “Nngeo: K-Nearest Neighbor Join for Spatial Data.”
.
Padgham, Mark, Robin Lovelace, Maëlle Salmon, and Bob Rudis. 2017.
“Osmdata.” *Journal of Open Source Software* 2 (14): 305.
.
Scarpone, Christopher, Sebastian T. Brinkmann, Tim Große, Daniel
Sonnenwald, Martin Fuchs, and Blake Byron Walker. 2020. “A Multimethod
Approach for County-Scale Geospatial Analysis of Emerging Infectious
Diseases: A Cross-Sectional Case Study of COVID-19 Incidence in
Germany.” *International Journal of Health Geographics* 19 (1): 32.
.
Walker, Blake Byron, Sebastian Tobias Brinkmann, Tim Große, Dominik
Kremer, Nadine Schuurman, Perry Hystad, Sumathy Rangarajan, Koon Teo,
Salim Yusuf, and Scott A. Lear. 2022. “Neighborhood Greenspace and
Socioeconomic Risk Are Associated with Diabetes Risk at the
Sub-Neighborhood Scale: Results from the Prospective Urban and Rural
Epidemiology (PURE) Study.” *Journal of Urban Health* 99 (3): 506–18.
.
Walker, Blake Byron, Aateka Shashank, Danijela Gasevic, Nadine
Schuurman, Paul Poirier, Koon Teo, Sumathy Rangarajan, Salim Yusuf, and
Scott A. Lear. 2019. “The Local Food Environment and Obesity: Evidence
from Three Cities.” *Obesity (Silver Spring, Md.)* 28 (1): 40–45.
.
STBrinkmann/DRIGLUCoSE documentation built on Oct. 14, 2022, 11:26 a.m.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
DRI-GLUCoSE
As elaborated in our recent analyses (Walker et al. 2019; Scarpone et
al. 2020), nearly all previous studies in the literature use either
census unit boundaries or simple buffer zones to measure an individual’s
built environment (BE) exposures or to characterize their local
socioeconomic status (SES) (Gong et al. 2014; Fuertes et al. 2014).
Therefore, we present a distance-weighted, network-based model for
quantifying the combined effects of local greenspace and SES on diabetes
risk, from which we derive an area-based Diabetes Risk Index of
Greenspace, Land Use and Socioeconomic Environments (DRI-GLUCoSE). The
goal of the DRIGLUCoSE package is to provide a public package
containing functions and code used in the development of the DRI-GLUCoSE
Index(Walker et al. 2022).
Installation
You can install the latest version of DRIGLUCoSE from GitHub with:
remotes::install_git("https://github.com/STBrinkmann/DRIGLUCoSE")
Once installed, the library can be loaded as follows:
library(DRIGLUCoSE)
Methods
One key purpose of this package is, to provide functions for route networked derived isochrones. For that purpose we have provided a sample sf object of 2 points in Erlangen, Germany.
data(Erlangen)
Erlangen
## Simple feature collection with 2 features and 2 fields
## Geometry type: POINT
## Dimension: XY
## Bounding box: xmin: 35199.46 ymin: -159433.5 xmax: 36281.59 ymax: -159243.2
## Projected CRS: ETRS89 / LCC Germany (N-E)
## # A tibble: 2 x 3
## tag Speed geom
## <dbl> <dbl> <POINT [m]>
## 1 1 78.5 (35199.46 -159433.5)
## 2 2 79.8 (36281.59 -159243.2)
Census variables
In our analysis we acquired data of the Canadian census dissemination areas. It has been converted to a shapefile (sf) with one column per census variable. To demonstrate we use the following randomly generated data:
set.seed(1234)
census <- sf::st_make_grid(
# Use Sample Data and apply 25 minutes buffer (Speed[m/min] * 25[min])
Erlangen %>% dplyr::mutate(geom = sf::st_buffer(geom, Speed*25)),
cellsize = 100
) %>%
sf::st_as_sf() %>%
dplyr::mutate(census_var_a = sample(1:1000, n(), replace = TRUE),
census_var_b = sample(1000:10000, n(), replace = TRUE),
census_var_c = sample(100000:150000, n(), replace = TRUE)) %>%
dplyr::rename(geom = x)
census
## Simple feature collection with 2142 features and 3 fields
## Geometry type: POLYGON
## Dimension: XY
## Bounding box: xmin: 33236.96 ymin: -161396 xmax: 38336.96 ymax: -157196
## Projected CRS: ETRS89 / LCC Germany (N-E)
## First 10 features:
## census_var_a census_var_b census_var_c geom
## 1 284 2009 114811 POLYGON ((33236.96 -161396,...
## 2 848 3318 139961 POLYGON ((33336.96 -161396,...
## 3 918 3954 112673 POLYGON ((33436.96 -161396,...
## 4 101 3359 137505 POLYGON ((33536.96 -161396,...
## 5 623 3107 109201 POLYGON ((33636.96 -161396,...
## 6 905 2630 131656 POLYGON ((33736.96 -161396,...
## 7 645 6512 135789 POLYGON ((33836.96 -161396,...
## 8 934 9945 115006 POLYGON ((33936.96 -161396,...
## 9 400 3583 111379 POLYGON ((34036.96 -161396,...
## 10 900 7778 137576 POLYGON ((34136.96 -161396,...
Greenspace
In our analysis we acquired LANDSAT images through the United States
Geological Survey’s EarthExplorer platform
(https://earthexplorer.usgs.gov/). The Normalized Difference
Vegetation Index
(NDVI)
is used as a metric to model greenspace exposure. Pre-processing of the
LANDSAT images and NDVI calculation has been conducted using the
LS_L1C function:
DRIGLUCoSE::LS_L1C(l1c_path = "docs/LC08_L1TP_193026_20200423_20200508_01_T1_small/",
out_dir = "docs/LS_PreProcessed",
# Use Sample Data and apply 25 minutes buffer (Speed[m/min] * 25[min])
sf_mask = DRIGLUCoSE::Erlangen %>%
dplyr::mutate(geom = sf::st_buffer(geom, Speed*25)),
cores = 20)
## Project raster
## DN to TOA Reflectance
## class : RasterStack
## dimensions : 122, 151, 18422, 8 (nrow, ncol, ncell, nlayers)
## resolution : 30, 30 (x, y)
## extent : 33493.69, 38023.69, -161164.2, -157504.2 (xmin, xmax, ymin, ymax)
## crs : +proj=lcc +lat_0=51 +lon_0=10.5 +lat_1=48.6666666666667 +lat_2=53.6666666666667 +x_0=0 +y_0=0 +ellps=GRS80 +units=m +no_defs
## names : Blue, Green, Red, NIR, SWIR1, SWIR2, NDWI, NDVI
## min values : 0, 0, 0, 0, 0, 0, -1, -1
## max values : 0.2020575, 0.2322532, 0.3076383, 0.5424371, 0.4233773, 0.3753066, 1.0000000, 1.0000000
Exposure Model
In order to estimate each participant’s potential exposures to greenspace and local SES, we (i) mapped age- and sex-specific walkable zones around their residential address, and (ii) applied a negative logit weighting function, such that the estimated effect of greenspace or SES decreases as distance from the home increases.
(i) Road network data and isochrones
In order to compute network-based distance metrics, we acquired street
data from OpenStreetMap using the R-package osmdata (Padgham et al.
2017). Road types not suitable for walking were removed (e.g.,
motorways). Network data were topologically corrected and split into
\~20 metre-long segments using the R package nngeo (Michael Dorman
2020).
erlangen.osm <- DRIGLUCoSE::osm_roads(x = Erlangen, dist = 20,
speed = "Speed", cores = 2)
This network data was used to derive walking distance buffers for each participant, based on walking speed. Starting from each participant’s place of residence, we computed network-constrained buffers with an off-road width of 40 meters, running in 2-minute increments from 0 to 20 minutes, using the A*-algorithm (Hart, Nilsson, and Raphael 1968). This therefore resulted in each participant having ten concentric isochrones, the sizes of which are a function of individual walking speed and road network.
erlangen.isodistances <- DRIGLUCoSE::isodistances(x = Erlangen,
road_network = erlangen.osm,
tag = "tag", speed = "Speed",
isochrones_seq = seq(2, 20, 2),
cores = 2)
erlangen.isochrones <- DRIGLUCoSE::isochrones(x = erlangen.isodistances,
buffer = 40, cores = 2)
erlangen.isochrones
## Simple feature collection with 20 features and 2 fields
## Geometry type: MULTIPOLYGON
## Dimension: XY
## Bounding box: xmin: 34034 ymin: -160951 xmax: 37750 ymax: -157782
## Projected CRS: ETRS89 / LCC Germany (N-E)
## # A tibble: 20 x 3
## time tag geom
## * <dbl> <dbl> <MULTIPOLYGON [m]>
## 1 2 1 (((35292.18 -159582.1, 35292.75 -159584.1, 35293.22 -159586.1, 3~
## 2 4 1 (((35314.38 -159725.3, 35313.74 -159727.3, 35312.99 -159729.2, 3~
## 3 6 1 (((35348.37 -159830.5, 35347.9 -159831.2, 35347.44 -159832, 3534~
## 4 8 1 (((35282 -159969.8, 35281.96 -159971.9, 35281.8 -159974, 35281.5~
## 5 10 1 (((35249.78 -160125.6, 35249.57 -160125.5, 35248.86 -160125.1, 3~
## 6 12 1 (((35264.36 -160300.8, 35263.71 -160300.8, 35263.06 -160300.9, 3~
## 7 14 1 (((35379.71 -160416.5, 35380.79 -160418.3, 35381.78 -160420.1, 3~
## 8 16 1 (((35412 -160571, 35411.95 -160573.1, 35411.78 -160575.2, 35411.~
## 9 18 1 (((35430.23 -160426.8, 35430.28 -160426.3, 35430.28 -160426.3, 3~
## 10 20 1 (((35546 -160803.2, 35546.09 -160803.5, 35546.22 -160804.2, 3554~
## 11 2 2 (((36385.71 -159372.9, 36384.21 -159374.3, 36382.64 -159375.7, 3~
## 12 4 2 (((36379.78 -159545.8, 36379.83 -159546.5, 36379.9 -159547.1, 36~
## 13 6 2 (((36368.01 -159688.3, 36366.95 -159690.1, 36365.8 -159691.9, 36~
## 14 8 2 (((36655.01 -159679.5, 36655.61 -159681.6, 36656.11 -159683.6, 3~
## 15 10 2 (((36520.27 -159907.5, 36518.33 -159908.3, 36516.36 -159909, 365~
## 16 12 2 (((36053.07 -160058.5, 36051.72 -160056.9, 36050.46 -160055.3, 3~
## 17 14 2 (((36832.82 -159976.8, 36832.8 -159977, 36832.79 -159977.1, 3683~
## 18 16 2 (((36016.71 -160325, 36016.5 -160325.4, 36016.23 -160325.9, 3601~
## 19 18 2 (((36020.67 -160327.7, 36022.64 -160327, 36024.65 -160326.4, 360~
## 20 20 2 (((36689.36 -160321.8, 36690.36 -160320, 36691.45 -160318.2, 366~
Figure 1 shows isodistances of the two points of the sample data in Erlangen, Germany.
(ii) Distance-weighting
In order to account for the diminishing effect of SES and greenspace exposure as distance increases, we fitted a logit function to weight each incremental isochrone, such that the influence of a variable decreases with increasing distance from the household, i.e., features that are farther away have less influence than nearby features, as illustrated in Figure 2. A logit function was selected as it heuristically approximates a suitable distance-decay function (Bauer and Groneberg 2016; Jia, Wang, and Xierali 2019). The distance-weighting is separated in two parts, first the logit function (1) that is used for both SES and greenspace variables, and second the proportional weights function (4) that is only applied on SES variables.
Each isochrone
is assigned a distance weight
,
calculated as the integral of the logistic distance decay function
(2)
with
and
,
in the interval between the mean inner radius
and mean outer radius
of the isochrone (e.g. 2 to 4 minutes isochrones), normalized by the
integral from 0 to the outermost isochrone boundary
(e.g. 20 minutes isochrone). Weighted summary statistics to describe the
greenspace (e.g. mean or minimum NDVI) are thus described as (3)
For SES variables the proportional weights of the census areas within the isochrone are further defined as (4)
with the proportion of the area of the intersection of the census area
and the isochrone
,
and the area of the isochrone
.
The weighted value of the SES variable
in the census area
is then defined as (5)
Figure 2 visualizes the different submodels used for distance-weighting SES and greenspace. Fig. 2a shows the unweighted values of a SES variable and fig. 2b has been calculated using (5), thus representing the proportional weights of all intersections with the census areas and isochrones. Greenspace is weighted as shown in fig. 2c using (3).
The distance-weighting for the LANDSAT derived NDVI raster (greenspace
exposure) is handled using LS_band_weightin, and SES distance- and
areal-weighting using census_weighting.
# Calculate sd, median, 5th percentile, 95th percentile and skew of NDVI values
NDVI_weighted <-
DRIGLUCoSE::LS_band_weighting(isochrones = erlangen.isochrones, tag = "tag",
landsat_list = dir("docs/LS_PreProcessed",
pattern = ".grd",
full.names = T) %>%
lapply(raster::brick),
stats = list("sd", "median",
list("percentile", 0.05),
list("percentile", 0.95),
"skew"),
b = 8, m = 0.6, cores = 2)
NDVI_weighted
## # A tibble: 2 x 6
## tag sd median X5_percentile X95_percentile skew
## <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 1 0.205 0.615 0.281 0.914 -0.145
## 2 2 0.104 0.540 0.360 0.714 -0.0122
census_weighted <- DRIGLUCoSE::census_weighting(isochrones = erlangen.isochrones,
tag = "tag", census = census,
b = 8, m = 0.6, cores = 2)
census_weighted
## # A tibble: 2 x 4
## tag census_var_a census_var_b census_var_c
## <dbl> <dbl> <dbl> <dbl>
## 1 1 562. 5323. 130970.
## 2 2 547. 5419. 124610.
Appendix
Figures
DRI-GLUCoSE scores for Vancouver (top) and Hamilton (bottom), ranging from low risk (purple) to high risk areas (orange).
Tables
Table A.1: Logistic models for all multivariable models with odds ratios (OR) for diabetes, 95% CI, and p-values.
Semi-Adjusted
Fully-Adjusted
Parameter
OR(bivariate)
OR(WHR-adjusted)
OR(BMI-adjusted)
OR(WHR-adjusted)
OR(BMI-adjusted)
DRI-GLUCoSE Score
0.37 (0.26-0.51,p<0.001)
0.41 (0.31-0.53,p<0.001)
0.46 (0.35-0.61,p<0.001)
0.45 (0.36-0.56,p<0.001)
0.50 (0.37-0.67,p<0.001)
Age (5 year-interval)
1.26 (1.18-1.34,p<0.001)
1.21 (1.15-1.27,p<0.001)
1.26 (1.20-1.33,p<0.001)
1.20 (1.15-1.65,p<0.001)
1.24 (1.18-1.31,p<0.001)
Sex: female
0.63 (0.50-0.78,p<0.001)
0.85 (0.71-1.00,p=0.056)
0.47 (0.39-0.57,p<0.001)
0.95 (0.81-1.11,p=0.498)
0.59 (0.48-0.71,p<0.001)
Obese (WHR)
5.72 (4.37-7.59,p<0.001)
5.39 (4.50-6.46,p<0.001)
5.06 (4.29-5.97,p<0.001)
BMI
1.14 (1.12-1.16,p<0.001)
1.14 (1.12-1.15,p<0.001)
1.13 (1.11-1.15,p<0.001)
Household income range
0.75 (0.70-0.81,p<0.001)
0.87 (0.82-0.92,p<0.001)
0.85 (0.80-0.90,p<0.001)
0.90 (0.86-0.95,p<0.001)
0.85 (0.80-0.91,p<0.001)
Neighbourhood type: urban
0.93 (0.71-1.21,p=0.605)
0.58 (0.47-0.72,p<0.001)
0.62 (0.50-0.77,p<0.001)
0.60 (0.50-0.72,p<0.001)
0.66 (0.52-0.84,p<0.001)
AHEI Score (E^1)
0.78 (0.70-0.88,p<0.001)
0.95 (0.88-1.03,p=0.197)
0.99 (0.89-1.09,p=0.771)
Recreation Met Score: >=525
0.58 (0.46-0.73,p<0.001)
0.82 (0.71-1.03,p=0.006)
0.85 (0.71-1.03,p=0.090)
Current/Former smoker: yes
1.65 (1.31-2.07,p<0.001)
1.55 (1.34-1.80,p<0.001)
1.38 (1.15-1.66,p<0.001)
Alcohol: <1 drink/day
1.17 (0.91-1.51,p=0.217)
1.81 (1.54-2.14,p<0.001)
1.34 (1.09-1.65,p=0.006)
To analyse the effect of socioeconomic status (SES) and greenspace (GS), we further build multivariable models using the semi-adjusted model with BMI as obesity measurent and tested different combinations for the index variable.
Table A.2: Model Performance and odds ratios of the logistic models, comparing combinations of socioeconomic status (SES) and greenspace (GS) as index. Metric SES + GS SES GS Probability Threshold * 0.47 0.50 0.48 Accuracy 0.75 0.72 0.74 Sensitivity 0.76 0.72 0.75 Specificity 0.65 0.68 0.68 Youden index 0.41 0.40 0.43 OR (95% CI, p-value) 0.46 (0.35-0.61, p < 0.001) 0.57 (0.42-0.76, p < 0.001) 0.42 (0.31-0.57, p < 0.001) * Probability threshold used for predicting Diabetes. Values equal or greater than this threshold are mapped as “No”. Table A.3: Model Performance for all multivariable models.
Semi-Adjusted
Fully-Adjusted
Metric
OR(WHR-adjusted)
OR(BMI-adjusted)
OR(WHR-adjusted)
OR(BMI-adjusted)
Probability Threshold *
0.44
0.47
0.58
0.60
Accuracy
0.71
0.75
0.64
0.64
Sensitivity
0.72
0.76
0.62
0.62
Specificity
0.70
0.65
0.81
0.83
Youden index
0.42
0.41
0.43
0.45
* Probability threshold used for predicting Diabetes. Values
equal or greater than this threshold are mapped as “No”.
Summary Statistics
Table A.4: Baseline characteristics of the study population by diabetes status. Parameter No Diabetes (N=4616) Diabetes (N=509) Total (N=5125) Total included 4616 (90.1%) 509 (9.9%) 5125 City Hamilton 2307 (87.5%) 331 (12.5%) 2638 Vancouver 2309 (92.8%) 178 (7.2%) 2487 Participant data BMI Mean (SD) 27.3 (5.4) 32.1 (6.4) 27.8 (5.7) Median (Q1; Q3) 26.4 (23.8; 29.8) 30.9 (27.5; 35.7) 26.8 (24.0; 30.4) Waist to hip ratio Mean (SD) 85.2 (9.0) 93.6 (8.5) 86.0 (9.3) Median (Q1; Q3) 85.2 (78.6; 91.6) 94.2 (87.8; 99.6) 86.0 (79.3; 92.6) Obesity (WHR) Low & Moderate 2787 (96.0%) 115 (4.0%) 2902 High 1829 (82.3%) 394 (17.7%) 2223 Age (years) Mean (SD) 10.5 (1.9) 11.4 (1.7) 10.6 (1.9) Median (Q1; Q3) 10.6 (9.0; 12.0) 11.6 (10.2; 12.6) 10.6 (9.2; 12.2) Sex Male 2076 (87.6%) 294 (12.4%) 2370 Female 2540 (92.2%) 215 (7.8%) 2755 Household Income Range >90k 1793 (93.3%) 129 (6.7%) 1922 45k-65k 759 (89.8%) 86 (10.2%) 845 30k-45k 564 (85.2%) 98 (14.8%) 662 65k-90k 1010 (92.7%) 80 (7.3%) 1090 20k-30k 304 (81.3%) 70 (18.7%) 374 <20k 186 (80.2%) 46 (19.8%) 232 AHEI Score Mean (SD) 3.8 (1.0) 3.5 (0.9) 3.7 (1.0) Median (Q1; Q3) 3.8 (3.1; 4.5) 3.5 (2.8; 4.1) 3.7 (3.0; 4.5) Physical Activity MET Score under 1721 (87.3%) 250 (12.7%) 1971 over 2065 (92.3%) 172 (7.7%) 2237 Ever smoked Mean (SD) 0.4 (0.5) 0.6 (0.5) 0.5 (0.5) Median (Q1; Q3) 0.0 (0.0; 1.0) 1.0 (0.0; 1.0) 0.0 (0.0; 1.0) Daily Drinker >=1 drinks/day 1157 (91.0%) 114 (9.0%) 1271 <1 drink/day 2629 (89.5%) 308 (10.5%) 2937 Census variables Neighbourhood type suburban/rural 3607 (89.9%) 405 (10.1%) 4012 urban 1009 (90.7%) 104 (9.3%) 1113 Individual mean income (CAD/1000) Mean (SD) 37.9 (13.9) 34.3 (12.1) 37.5 (13.8) Median (Q1; Q3) 35.0 (29.0; 42.0) 32.0 (27.0; 38.0) 35.0 (29.0; 42.0) Household median income (CAD/1000) Mean (SD) 65.5 (21.5) 60.6 (20.5) 65.0 (21.4) Median (Q1; Q3) 61.0 (51.0; 77.0) 57.0 (45.0; 71.0) 61.0 (51.0; 76.0) Prevalence of low income (%) Mean (SD) 9.4 (6.9) 11.3 (7.6) 9.6 (7.0) Median (Q1; Q3) 8.1 (4.3; 13.7) 10.1 (5.2; 16.3) 8.2 (4.4; 14.0) Commute Walking/Bicycle (%) Mean (SD) 7.5 (6.9) 6.8 (5.8) 7.4 (6.8) Median (Q1; Q3) 5.1 (3.0; 9.9) 5.3 (3.2; 8.6) 5.2 (3.0; 9.7) Labour force participation rate (%) Mean (SD) 66.6 (7.7) 65.3 (7.6) 66.5 (7.7) Median (Q1; Q3) 66.8 (61.8; 71.7) 64.9 (60.7; 70.5) 66.6 (61.7; 71.6) Gov’t transfer payments (%) Mean (SD) 9.9 (5.2) 12.4 (6.3) 10.1 (5.4) Median (Q1; Q3) 8.8 (5.7; 13.0) 11.5 (7.2; 16.1) 9.0 (5.8; 13.3) Unemployment rate (%) Mean (SD) 5.5 (2.5) 6.3 (2.9) 5.6 (2.6) Median (Q1; Q3) 5.4 (3.8; 7.0) 5.9 (4.4; 7.7) 5.4 (3.8; 7.0) Lone parent families (%) Mean (SD) 14.8 (6.5) 17.3 (7.5) 15.0 (6.6) Median (Q1; Q3) 14.0 (10.9; 18.4) 16.1 (11.9; 21.8) 14.1 (11.0; 18.8) Education - No degree (%) Mean (SD) 19.6 (9.2) 23.7 (10.4) 20.0 (9.4) Median (Q1; Q3) 17.7 (12.8; 24.9) 21.0 (15.8; 31.6) 17.9 (13.1; 25.6) Private Dwellings - Owned (%) Mean (SD) 73.9 (17.8) 72.9 (17.6) 73.8 (17.7) Median (Q1; Q3) 77.1 (62.1; 88.5) 75.6 (61.7; 87.0) 77.0 (62.1; 88.4) Private Dwellings - Rented (%) Mean (SD) 25.3 (17.2) 26.4 (17.0) 25.4 (17.2) Median (Q1; Q3) 22.1 (11.2; 36.9) 24.0 (12.9; 37.7) 22.4 (11.4; 36.9) Normalized Difference Vegetation Index (NDVI) NDVI - Median Mean (SD) 0.341 (0.089) 0.314 (0.084) 0.338 (0.089) Median (Q1; Q3) 0.336 (0.291; 0.379) 0.313 (0.266; 0.360) 0.333 (0.288; 0.377) NDVI - Standard Deviation Mean (SD) 0.090 (0.026) 0.088 (0.024) 0.090 (0.026) Median (Q1; Q3) 0.086 (0.071; 0.105) 0.084 (0.072; 0.100) 0.086 (0.071; 0.105) NDVI - Min Mean (SD) 0.206 (0.097) 0.182 (0.086) 0.204 (0.096) Median (Q1; Q3) 0.202 (0.135; 0.262) 0.173 (0.118; 0.243) 0.198 (0.134; 0.260) NDVI - max Mean (SD) 0.495 (0.093) 0.465 (0.093) 0.492 (0.093) Median (Q1; Q3) 0.484 (0.435; 0.546) 0.459 (0.401; 0.514) 0.482 (0.432; 0.543)About
Package contributors
Brinkmann, Sebastian Tobias (Package creator and author) e-mail: sebastian.brinkmann@fau.de
Große, Tim (Contributor)
Thesis authors
Walker, Blake Byron (1*) Brinkmann, Sebastian Tobias (1) Große, Tim (1) Dominik Kremer (1) Schuurman Nadine (2) Hystad Perry (3) Rangarajan Sumathy (4) Teo Koon (4) Yusuf Salim (4) Lear Scott A. (5)
1: Community Health Environments and Social Terrains (CHEST) Lab, Institut für Geographie, Friedrich-Alexander-Universität Erlangen-Nürnberg, Wetterkreuz 15, 91052 Erlangen, Germany
*corresponding author
2: Department of Geography, Simon Fraser University, Burnaby, Canada
3: Spatial Health Lab, College of Public Health and Human Sciences, Oregon State University, Corvallis, USA
4: Population Health Research Institute, McMaster University, Hamilton, Canada
5: Faculty of Health Sciences, Simon Fraser University, Burnaby, Canada
Citation
citation("DRIGLUCoSE")
##
## To cite DRIGLUCoSE in publications use:
##
## Walker, B.B., Brinkmann, S.T., Große, T. et al. Neighborhood
## Greenspace and Socioeconomic Risk are Associated with Diabetes Risk
## at the Sub-neighborhood Scale: Results from the Prospective Urban and
## Rural Epidemiology (PURE) Study. J Urban Health 99, 506–518 (2022).
## https://doi.org/10.1007/s11524-022-00630-w
##
## A BibTeX entry for LaTeX users is
##
## @Article{,
## title = {Neighborhood Greenspace and Socioeconomic Risk are Associated with Diabetes Risk at the Sub-neighborhood Scale: Results from the Prospective Urban and Rural Epidemiology (PURE) Study},
## author = {Blake Byron Walker and Sebastian T. Brinkmann and T. Große et al.},
## journal = {J Urban Health},
## year = {2022},
## volume = {99},
## pages = {506–518},
## url = {https://doi.org/10.1007/s11524-022-00630-w},
## }
Bibliography
Bauer, Jan, and David A. Groneberg. 2016. “Measuring Spatial
Accessibility of Health Care Providers - Introduction of a Variable
Distance Decay Function Within the Floating Catchment Area (FCA)
Method.” *PloS One* 11 (7): e0159148.
.
Fuertes, Elaine, Iana Markevych, Andrea von Berg, Carl-Peter Bauer,
Dietrich Berdel, Sibylle Koletzko, Dorothea Sugiri, and Joachim
Heinrich. 2014. “Greenness and Allergies: Evidence of Differential
Associations in Two Areas in Germany.” *Journal of Epidemiology and
Community Health* 68 (8): 787–90.
.
Gong, Yi, John Gallacher, Stephen Palmer, and David Fone. 2014.
“Neighbourhood Green Space, Physical Function and Participation in
Physical Activities Among Elderly Men: The Caerphilly Prospective
Study.” *International Journal of Behavioral Nutrition and Physical
Activity* 11 (1): 40. .
Hart, Peter, Nils Nilsson, and Bertram Raphael. 1968. “A Formal Basis
for the Heuristic Determination of Minimum Cost Paths.” *IEEE
Transactions on Systems Science and Cybernetics* 4 (2): 100–107.
.
Jia, Peng, Fahui Wang, and Imam M. Xierali. 2019. “Differential Effects
of Distance Decay on Hospital Inpatient Visits Among Subpopulations in
Florida, USA.” *Environmental Monitoring and Assessment* 191 (Suppl 2):
381. .
Michael Dorman. 2020. “Nngeo: K-Nearest Neighbor Join for Spatial Data.”
.
Padgham, Mark, Robin Lovelace, Maëlle Salmon, and Bob Rudis. 2017.
“Osmdata.” *Journal of Open Source Software* 2 (14): 305.
.
Scarpone, Christopher, Sebastian T. Brinkmann, Tim Große, Daniel
Sonnenwald, Martin Fuchs, and Blake Byron Walker. 2020. “A Multimethod
Approach for County-Scale Geospatial Analysis of Emerging Infectious
Diseases: A Cross-Sectional Case Study of COVID-19 Incidence in
Germany.” *International Journal of Health Geographics* 19 (1): 32.
.
Walker, Blake Byron, Sebastian Tobias Brinkmann, Tim Große, Dominik
Kremer, Nadine Schuurman, Perry Hystad, Sumathy Rangarajan, Koon Teo,
Salim Yusuf, and Scott A. Lear. 2022. “Neighborhood Greenspace and
Socioeconomic Risk Are Associated with Diabetes Risk at the
Sub-Neighborhood Scale: Results from the Prospective Urban and Rural
Epidemiology (PURE) Study.” *Journal of Urban Health* 99 (3): 506–18.
.
Walker, Blake Byron, Aateka Shashank, Danijela Gasevic, Nadine
Schuurman, Paul Poirier, Koon Teo, Sumathy Rangarajan, Salim Yusuf, and
Scott A. Lear. 2019. “The Local Food Environment and Obesity: Evidence
from Three Cities.” *Obesity (Silver Spring, Md.)* 28 (1): 40–45.
.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Embedding an R snippet on your website
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.