Spatial Analysis of Cardiovascular Disease
Incidence and Potential Environmental Factors in the California Bay Area
Nicholas
J. Papanastassiou ’13 and Sarah A. Holmes ‘13
Environmental
Studies Program, Colby College, Waterville, Maine
Abstract
This project
uses GIS to investigate proposed environmental factors that contribute to heart
disease. This is based on several scientific articles (see references) that
have hypothesized positive correlations between areas of increased air and
noise pollution and dioxin emissions with heart disease incidence. We examined
the Bay Area of California, which consists of nine counties. No significant
correlation between our proposed factors and current heart disease incidence
was found.
Introduction
Cardiovascular
disease (CVD) affects 81 million people throughout the United States. There are
many contributing factors to CVD, including smoking, diet and health decisions,
and congenital heart defects. Additionally, studies have shown noise and air
pollution to increase one’s risk of CVD. We asked the question of whether or
not NPL Superfund Sites, Airports, and Roads increase the likelihood of
contracting CVD. We chose to analyze the Bay Area in Northern California,
looking at incidences of CVD per county.
Methods
We looked
at county data for cardiovascular disease incidence, corrected for age
differences, and compared it to an original model to predict heart disease
incidence. We compiled data containing information on our three factors: air
pollution, noise pollution, and dioxin emissions, which we quantified by
looking at road density (weighted by road type), airport density, and Superfund
site density in each county. Road density was measured by total length of road
system / total area of county, airport density was measured by total # of
airports / cumulative population in all cities with airports, and Superfund
site density was measured by # of sites / total area of county. We converted
each range of densities to a 110 scale, in order to compare them.
We then
ran a multivariable leastsquares regression in order to determine which
variables best predicted incidences of CVD. The resulting equation was: CVD = 7.750 – 0.050(AirDens) – 0.006(RoadDens) – 0.137(SuperDens).
We
assigned weights to each variable based on the coefficients from the regression
equation. Using this equation, we produced a predictive output layer of CVD
incidences. This was coded with the same methods as the observed CVD data in
each county, so as to be able to visually compare the two maps.
Results
We found
no significant predictive value of the hypothesized factors in measuring CVD
incidences. A multivariable regression of our dependent variable (CVD
Incidence) and our three independent variables (Road, Airport, and NPL
Superfund Site Density) yielded Pvalues >0.635 for each variable and
tvalues <0.51 for each variable. An Fvalue of 0.89 resulted, as well as
an R^2 value of 0.11(See Table 1, Figure 3).
The index
layer predicts some CVD ranges for certain counties well, but Napa County and
Contra Costa County are two notable exceptions.
Figure
1
Figure
2
Table
1

Coef 
Std. Err. 
tvalue 
P>t 
Airport Density 
.050 
.188 
0.27 
0.800 
Road Density 
.006 
.276 
0.02 
0.984 
Superfund Site Density 
.137 
.271 
0.51 
0.635 
Constant 
7.750 
1.177 
6.58 
0.001 
P>F = 0.8884, R^2 =
0.109
Figure 3
Discussion
Our findings
indicate that the noise and air pollution associated with airports, roads, and
NPL Superfund sites do not significantly contribute to incidences of CVD. Only
10% of the variance in CVD incidences between Bay Area counties is captured by
the variables. For road density, there is a 98% chance that the minimal effect
that road density has on CVD is attributable to chance alone. These findings
are in opposition to the hypotheses of Hoffmann et al., Kopf et al., Mead, and Román et al.
It is
important to note the severe limitations of this study. Firstly, there are many
confounding variables not accounted for in the model that likely affect CVD
incidences (e.g., smoking) as well as the variables that we used in our model
(e.g., wind direction affecting pollutants). Secondly, our data on CVD
incidences were available only to the countywide scale, further limiting the
precision of our analysis.
Thirdly,
our statistical analysis shows that the combined independent variables do not
predict CVD incidences better than would be predicted by chance alone.
Therefore, the equation resulting from the regression and thus the equation
used to calculate our index map was not significant. Fourthly, it was likely
inappropriate to use linear regression at all. We had a very small sample size
that severely limited the accuracy of the regression, and prevented tests of
normalcy from being used. Additionally, the statistical data were not tested
for autocorrelation, which could impact the accuracy of the results.
Therefore,
as a result of these limitations, the index layer created for the projected
incidence of CVD map does not reflect significant predictions. However, it
showed the inability of the road, airport, and superfund density in our model
to accurately predict observed rates of CVD incidences.
Conclusion
There is
no significant relationship between the densities of NPL Superfund Sites,
Airports, and Roads in the Bay Area counties, and increased likelihood of
contracting CVD in those counties.
References:
Hoffmann
B, Moebus S, Dragano N, Möhlenkamp S, Memmesheimer M, Erbel R, Jöckel KH. Residential traffic exposure and coronary
heart disease: results from the Heinz Nixdorf Recall Study. 2009.
Biomarkers.
Kopf P,
Walker M. Overview of developmental heart
defects by dioxins, PCBs, and pesticides. 2009. Journal of Environmental
Science and Health.
Mead M. Noise Pollution: The Sound Behind Heart Effects. 2007. Environmental Health
Perspectives.
Román A, Prieto C, Mancilla F, Astudillo O, Dussaubat A, Miguel
W, Lara M. Association between air
pollution and cardiovascular risk. 2009. Revista médica de Chile.
Acknowledgements: A big thank you to
Professor Nyhus and Manny Gimond
for help with brainstorming ideas, how to most effectively use GIS in our
project, proper methods for carrying out the analysis, and being available and
willing to answer any questions that we might have.
Data was
obtained from ESRI93, Federal Aviation Administration, EPA National Priorities
List, Metropolitan Transportation Commission and California Department of
Health Services: Center for Health Statistics.