A GIS Based Approach for Assessing the Association between Air Pollution and Asthma in New York State, USA

Studies on asthma have shown that air pollution can lead to increased asthma prevalence. The aim of this study is to examine the association between air pollution (fine particulate matter (PM2.5), sulfur dioxide (SO2) and ozone (O3)) and human health (asthma emergency department visit rate (AEVR) and asthma discharge rate (ADR)) among residents of New York, USA during the period 2005 to 2007. Annual rates of asthma were calculated from population estimates for 2005, 2006, and 2007 and number of asthma hospital discharge and emergency department visits. Population data for New York were taken from US Bureau of Census, and asthma data were obtained from New York State Department of Health, National Asthma Survey surveillance report. Data on the concentrations of PM2.5, SO2 and ground level ozone were obtained from various air quality monitoring stations distributed in different counties. Annual means of these concentrations were compared to annual variations in asthma prevalence by using Pearson correlation coefficient. We found different associations between the annual mean concentration of PM2.5, SO2 and surface ozone and the annual rates of asthma discharge and asthma emergency visit from 2005 to 2007. A positive correlation coefficient was observed between the annual mean concentration of PM2.5, and SO2 and the annual rates of asthma discharge and asthma emergency department visit from 2005 to 2007. However, the correlation coefficient between annual mean concentrations of ground ozone and the annual rates of asthma discharge and asthma emergency visit was found to be negative from 2005 to 2007. Our study suggests that the association between elevated concentrations of PM2.5 and SO2 and asthma prevalence among residents of New York State in USA is consistent enough to assume concretely a plausible and significant association.

in New York State during three consecutive years from 2005 to 2007 using GIS. The application of GIS will assist toward better understanding of the problem and its potential solution [33,34].

Study Area
In the present work, New York State is selected as the area of study for the analysis and estimation. The study area is shown in Figure 1. New York is a state in the Northeastern region of the United States. The longitude and latitude of the state are 71°47'25"W to 79°45'54"W and 40°29'40"N to 45°0'42"N respectively. It is the third most populous (19,378,102), and the seventh most densely populated (415.3 inhabitants per square mile) state of the 50 United States [35]. New York covers 54,556 square miles and ranks as the 27th largest state by size [36]. In general, New York has a humid continental climate. Weather in New York is heavily influenced by two continental air masses: a warm, humid one from the southwest and a cold, dry one from the northwest. The winters are long and cold in the Plateau Divisions of the state. In the majority of winter seasons, a temperature of −25 °C or lower can be expected in the northern highlands (Northern Plateau) and 15 °C or colder in the southwestern and east-central highlands (Southern Plateau) [37]. The summer climate is cool in the Adirondacks, Catskills and higher elevations of the Southern Plateau.

Materials and Methods
Geographic Information Systems (GIS) is an essential scientific tool for health data processing, analysis of geographical distribution and variation of diseases, mapping, monitoring and management of health epidemics [38]. In the past, GIS has been applied to estimate the spatial concentrations of air pollutants [39] and many epidemiologic studies have adopted GIS to explore the health impact of air pollutants on asthma [40][41][42][43].
Although different counties in New York State might have varying air pollution concentrations, the air pollution data in each county was not available. In order to make an exposure assessment for the whole of New York State, we linked the annual exposure levels by geostatistical method and corresponding asthma visits to estimate the impact on asthma emergency department visits rate and asthma discharge rate. In the present study, GIS is used to estimate the association between air pollution (fine particulate matter (PM 2.5 ), sulfur dioxide (SO 2 ) and ozone (O 3 )) and human health (Asthma emergency department visit rate (AEVR) and asthma discharge rate (ADR)). The methodology is applied in two stages to deduce the association between air pollution and asthma rate in New York State. First, we estimated the pollutant level by constructing a spatial model representing a geographical area using daily average pollutant concentration data. Second, we linked air pollutant concentration to asthma rate within the defined study area. All the three pollutants are among the most studied of environmental hazards and are at significant levels in air to adversely affect human health.

Air Pollution Data
Air quality data collected by US EPA's Air Quality System (AQS) at the various monitoring stations located in different counties of New York State for the three years from 2005 to 2007 were used for the study. The air pollution data used in this study was taken from the United States Environmental Protection Agency (US EPA) air quality system data mart [46].

Spatial Analysis Using GIS
GIS is an appropriate tool to deduce spatial relationships and to facilitate the proper understanding and resolving of related complex issues. GIS can be used to characterize the sources of pollution and analyze its impact on human health as revealed by numbers of asthma reported to hospital or caused by elevated air pollution. By analyzing the spatial pattern of asthma hospitalization and air pollutants, major threat areas can be visualized in the form of maps with the help of GIS. The use of GIS techniques for statistical analysis and modelling (e.g., pollutants, and diseases) is rarely done in previous studies. However, the past studies have not been fully explored and applied in analyzing the association between exposure of air pollution and human health. During last decade, GIS-based pollution mapping using interpolation techniques such as inverse distance weighting, Kriging and land use regression modeling [33] was explored by many researchers for epidemiological studies. The outbreak of asthma has drawn much attention in the past two decades since data all around the world have shown a high rate of asthma morbidity and mortality despite the availability of effective symptomatic treatment.
Measurements of air pollutants were based on data routinely collected at 72 (twenty two for SO 2 , twenty five for PM 2.5 and ozone) USEPA administered monitoring stations distributed in different counties as represented in Figure 4. All point data were entered into a Geographic Information System using Arcview software from Environmental Systems Research Inc. (ESRI). The first stage involved determining the location (latitude and longitude) of air pollution monitoring stations. The spatial location (latitude and longitude) of air pollution monitoring stations were also obtained from the same source of air pollution monitoring website. The concentrations of SO 2 , O 3 , and PM 2.5 are reported as daily maximum 1 hour average concentrations, daily maximum 8 h average concentrations and daily 24 h average concentrations respectively. Annual average concentrations were calculated using the daily average value in a particular year for each of the monitoring station. The spatial locations of each of the selected monitoring stations along with the pollutant concentrations were fed into the GIS system. With the point data from each monitoring station, the Ordinary Kriging (OK) method was used to estimate the spatial distribution of pollutant levels in each county from 2005 to 2007 for each of the three pollutants: SO 2 , O 3 , and PM 2.5 . Analyses were carried out by Ordinary Kriging method using the Geostatistical Analyst Extension of ArcGIS. In Kriging, a smooth surface is estimated from irregularly spaced data points based on the assumptions that the spatial variation in the feature (O 3 , PM 2.5 , and SO 2 ) is homogeneous over the domain depends only on the distance between sites. In general, the Kriging method [39,47] was used as a statistical mapping technique for using data collected at each point location, to predict concentration in each grid cell over a spatial domain. This paper used stable kind of semivariogram model for prediction of pollution concentration. Mean standardized error (MSE) and root mean square standardized (RMSSE) were used to select accuracy of the model fit which estimate the distribution of air pollutants. Partial sill, range and nugget values of the selected error were determined to represent the semivariogram model characteristics.
The parameters of semivariogram model along with the errors outcome of each individual model are listed in Table 2. The cross-validation of the four air pollutants was done manually by ArcGIS Geostatistical extension. The criteria for a good fitting Kriging model used in this study were an average MSE near 0 and RMSS near 1. According to the cross validation results, the variance of PM 2.5 , SO 2 and O 3 was little bit overestimated as the RMSS value was less than 1 [48]. Since, we are more emphasized on the relative concentration level and its association with asthma rate; it doesn't influence much on the final outcome. In this study, ambient O 3 , PM 2.5 , and SO 2 levels for each county within the New York State were estimated because these values could be linked to the residences of asthma rate obtained from recorded and interpolated data. The spatial distribution maps for the O 3 obtained from Kriging method are represented in Figure 5a  For the distribution analysis of asthma rate (asthma discharge rate and emergency department visit rate), we have designed a points shape file by considering a location at the centroid of each county of New York State. The attributes entered to particular centroid point were the asthma rate calculated for same county. Subsequently, the raster images for asthma discharge rate and emergency department visit rate were created by interpolation using the Kriging method. The spatial distributions of asthma discharge rate obtained from Kriging analysis are represented in Figure 8a

Statistical Analyses
The study sought to investigate the spatio-temporal association between air pollutants and asthma rate. To understand the inter-relationships among predictor variables, both GIS based correlation analysis (map correlation analysis) and point data correlation analyses were carried out. We conducted Pearson Correlation analysis among the pollutant variables and asthma rates.

Map Correlation Analysis
Spatial distribution maps of ambient air pollution (SO 2 , O 3 , and PM 2.5 ) and asthma rates (AEVR and ADR) were used for correlation analysis in GIS on year wise basis. The map correlation results for three years (2005, 2006, and 2007) are reported in Table 3. Map correlation results clearly indicate that PM 2.5 and SO 2 are positively correlated with asthma rate whereas ozone has a negative correlation with asthma rate. Results of the map correlation analyses reveal that AEVR and ADR are negatively correlated with the ground level ozone (O 3

Point Correlation Analysis
The asthma rates (asthma discharge rate and asthma emergency department visit rate) were compared with the extracted values of pollutant concentrations (SO 2 , PM 2.5 and O 3 ) for understanding the association between asthma and air pollution. Asthma rate at the specified centroid position of each county was considered the same as that of the corresponding county value. Since the monitoring values of pollutant concentrations at the same location were not available, they were extracted from the interpolated map using GIS. The extracted values of pollutant concentrations (O 3 , PM 2.5 , and SO 2 ) at the centroid point of each of the county are represented in Table 4. The data represented in Table 4 for three years (2005, 2006, and 2007) were used for correlation analyses to determine the correlation coefficients values. The correlation analyses were carried out using Pearson two tailed correlation analysis using SPSS Statistics software version 21. Correlation analyses results are represented in Table 5.
The results represented in Table 5 clearly indicate that the associations or relations of the asthma rates (ADR and AEVR) with pollutant concentrations (PM 2.5 , O 3 , and SO 2 ) are similar to those obtained from map correlation analyses. That is ozone concentration showed a negative correlation with both asthma discharge rate and asthma emergency visit rate while the other two pollutants (SO 2  According to the past research, the majority of researchers considered air pollutants a risk factor for asthma, although the roles of specific air pollutants on various respiratory illnesses remain unclear [49,50]. That is, the general effect of air pollution leads to adverse respiratory events. But, the effect of specific air pollutants on asthma rate is yet to be examined for a plausible conclusion. According to Chan et al. [51] PM 10 might have a positive impact on asthma outpatient and emergency settings. They further suggested that future research is required to validate robust spatiotemporal patterns and trends. A study conducted by Nawahda [32] suggested that the association between elevated concentrations of PM 2.5 and surface ozone and asthma prevalence among school children in Japan is not strong enough to assume concretely a plausible and significant association. On the other hand, a controlled laboratory study conducted by Koenig [52] revealed that ozone aggravates asthma. Thus, more specific studies are needed for understanding the role of specific air pollutants on asthma rate. These studies should also consider the socioeconomic factors and climatic factors for better reflections.

Limitations of the Study
GIS facilitates the research studies for association between air pollution and health more quickly and with less effort but data quality, lack of spatial detail and spatial consistency between data sets impede their utility. The major limitations of the GIS based studies include the availability of uniformly distributed pollution and health data. A number of data problems and data limitations are encountered with the integration of health data in GIS. Primarily, the interpolation analysis gives better pollution level prediction for uniformly distributed spatial data in the study area. But, the pollution monitoring stations in New York State was not uniformly distributed and this may lead to some errors in prediction level.
A major drawback to the health data used in this analysis is that asthma hospitalization records only provide number of asthma cases, and do not reflect the severity of the asthma problem. Again, there is no statewide reporting of asthma and therefore no centralized asthma database. People suffering from asthma may be seen by a private doctor or may not be seen by any health care provider. This type of cases may not be listed in the asthma database. Furthermore, asthma data available does not consider the movement of people from one county to the next. This study used annual average level of pollutants at a particular location as the population's exposure level but the workplace was not located in the same location and this could be bias about an individual's exposure estimation, which could influence the results. In addition, the counties where outpatient visits and emergency visits took place were assumed to be the same districts where people were exposed to pollution. This may not always have been the case. The true exposure time was difficult to estimate due to lack of exposure information.

Conclusions
We found significant associations between exposure to modeled SO 2 and PM 2.5 with asthma prevalence rate in New York State, USA. Asthma prevalence among the residents was closely associated with the exposure of PM 2.5 followed by SO 2 . But there was a negative association between asthma rate and ground level ozone concentration. In conclusion, this preliminary study illustrates the potential use of the GIS-based method to evaluate the effects of air pollution on asthma rate (asthma emergency visit rate and asthma discharge rate). The results of this study provide a better understanding of the correlation of air pollution with asthma patient visits, and demonstrate that SO 2 , and PM 2.5 might have a positive impact each on asthma emergency visit rate (AEVR) and asthma discharge rate (ADR) while O 3 might have a negative impact. Since many other pollutants such as NO 2 , and VOCs may cause adverse health outcomes, future research is required to provide robust spatiotemporal patterns and trends.