Spatial relationships among asthma prevalence, health care utilization, and pollution sources in neighborhoods of Buffalo, New York.

Introduction

The adverse respiratory-system effects of pollution substances such as nitrogen dioxide, sulfur dioxide, ozone, diesel exhaust, and diesel exhaust particles (DEPs) are well documented (Dockery, 2001; Haahtela et al., 1991; Koenig, 1999; Loh, Sugerman-Brozan, Wiggins, Noiles, & Archibald, 2002; Pandya, Solomon, Kinner, & Balmes, 2002; Takenaka, Zhang, Diaz-Sanchez, & Saxon, 1995; Wong et al., 1999). A growing body of epidemiological evidence links current increases in morbidity from respiratory illnesses to air pollutants (Briggs et al., 1999; Donaldson, Gilmour, & MacNee, 2000; Edwards, Walters, & Griffiths, 1994; Ishizaki, Koizumi, Ikermori, Ishiyama, & Kushibiki, 1987; Lin, M., Chen, Burnett, Villeneuve, & Krewski, 2002; NILU, 1991; Norris et al., 1999; Schwartz, 1993, 1994; Weiland, Mundt, Ruckmann, & Keil, 1994; Wijst et al., 1993). Data linking air pollution to increases in asthma prevalence, however, are of limited availablity (Koenig, 1999).

[ILLUSTRATION OMITTED]

Asthma is a chronic inflammatory disease characterized by episodes of reversible airflow obstruction and hyper-responsiveness to triggers, including air pollutants. Common symptoms of an attack include coughing, wheezing, shortness of breath, and chest tightness. Recent studies have reported an increase in asthma diagnosis in the past two decades among children and adults in the United States (Centers for Disease Control [CDC], 1998; Crain et al., 1994; National Institute of Allergy and Infectious Diseases [NIAID], 2001). According to NIAID (2001), asthma is a growing health problem in the United States, particularly among inner-city African-American and Latino populations.

Components of traffic-related pollutants, including particulates and ozone, have been shown to induce lung inflammation (Schwartz, 1993, 1994). Little information, however, is available at the subnational level regarding the health impacts of traffic-related pollution along the North American Free Trade Agreement (NAFTA) corridors. A few studies in the Northeastern United States have attempted to analyze the association between traffic-related pollution and asthma (Lwebuga-Mukasa & Dunn-Georgiou, 2002; Lwebuga-Mukasa, Wojcik, Dunn-Georgiou, & Johnson, 2002; Lin, Munsie, Hwanga, Fitzgerald, & Cayo, 2002). A recently completed 10-year study conducted by Lwebuga-Mukasa, Oyana, Thenappan, & Ayirookuzhi (2002) at a U.S.-Canada border crossing reported a positive association between increased commercial traffic volume and increased utilization of asthma health care. When there was a sudden drop in total crossborder traffic, following the World Trade Center tragedy of September 11, 2001, an associated decrease in the utilization of health care for all respiratory diseases occurred in the subsequent several weeks, with recovery to levels comparable to those of 2000 occurring with a rebound in traffic (Lwebuga-Mukasa, Ayirookuzhi, & Hyland, 2002). Furthermore, two cross-sectional house-to-house surveys conducted six years apart found that Buffalo communities in close proximity to the NAFTA corridor have an increased prevalence of asthma not explained by the usual personal or household factors (Lwebuga-Mukasa, Johnson, & Oyana, 2002; Lwebuga-Mukasa, Wojcik, et al., 2002). Geographically supported relationships between sources of pollution and use of health care have not, however, been fully explored. Identification of geographic patterns and distributions may contribute to better spatial and temporal characterization of respiratory diseases and thus contribute to better preventive and mitigation measures.

Residents living in close proximity to the Peace Bridge Complex and major roadways have complained of respiratory illnesses (Peace Bridge Review Commission, 1999). The major roadways are conduits of heavy trucks and buses, which serve the increasingly busy U.S.-Canada NAFTA corridor. The volume of commercial traffic passing along this trade corridor grew at a rate of 9 percent per year between 1991 and 1999 (Peace Bridge Review Commission, 1999). While infrastructure development has been emphasized, little information is available on the health impacts of traffic-related pollution. Previous studies have reported that high levels of vehicle emissions are associated with cardiopulmonary morbidity and mortality (Dockery, 2001; Dockery et al., 1993; Schwartz, 1996).

[FIGURE 1 OMITTED]

The current study was motivated by two key interests: First, the authors wanted to investigate the spatial distributions of asthma cases in relation to major traffic corridors and the Peace Bridge Complex, and assess possible contributions of other U.S. Environmental Protection Agency (U.S. EPA)-identified pollution release sources in the study area. Second, the authors wanted to evaluate utilization of health care in the communities surrounding the city of Buffalo as previously reported (Lwebuga-Mukasa, Oyana, Thenappan, & Ayirookuzhi, 2002) in order to gain a broader perspective of the problem. The study evaluates five questions:

1. Is utilization of health care different at certain different distances from sources of pollution and major roadways?

2. Is utilization of health care different at certain different distances from major hospitals?

3. Are asthma cases clustered in certain locations--for example, in relation to health care facilities or pollution sources?

4. Where is asthma risk significantly elevated in the study area?

5. Does proximity to a pollution site increase the risk of having asthma?

Materials and Methods

Study Area and Population

The study area, as shown in Figure 1, comprised 28 zip codes located in Erie County, New York. The study area included the city of Buffalo and its major first-ring suburbs to the north and northeast (Tonawanda, Amherst, Kenmore, and Cheektowaga). Health care needs are primarily served by six major hospitals (Figure 1). According to the 2000 U.S. Census, the population of the study area was 544,269 (U.S. Census Bureau, 2000). In 1990, the study area had a total population of 609,992 (U.S. Census Bureau, 1990); the difference implies a decrease of 5.32 percent. According to the 2000 U.S. Census, the race/ethnic composition of the study area is 71.5 percent Caucasian, 21 percent African-American, 5 percent Latino/Hipansic, 2 percent Asian, and below 0.5 percent others. Of the 544,269 people, 47.1 percent were males and 52.9 percent were females. Approximately 24 percent of the population were between 1 and 17 years of age, 38 percent were between 18 and 44 years, 21 percent were between 45 and 64 years, and 17 percent were 65 years of age or older.

[FIGURE 2 OMITTED]

Data Categories

Four data categories were analyzed in this study:

1. emergency room visits for asthma, 1999-2001;

2. hospital discharges for asthma, 1991-1996;

3. hospitalization and outpatient visits for asthma, 1995-2000; and

4. suspected sources of pollution sites derived from U.S. EPA databases.

The Kaleida Health System, the largest health system in Western New York (comprising Buffalo General Hospital, Millard Fillmore Hospitals [Gates and Suburban], DeGraff Memorial Hospital, and Children's Hospital of Buffalo), maintains automated relational databases of inpatient and outpatient records, which is updated regularly. The functions of these databases require a high level of accuracy; they are used, for instance, for billing and recovering patients' benefits from insurance companies. In-house data quality standards are based on international classification of diseases and metadata standards. Individual information contained in these databases permits the identification of individual cases that may involve multiple visits, which eliminates the duplication of records. The authors' experience with these databases, however, indicates that incomplete information collected from some patients constitutes about 10 percent of the information (including incorrect birthdays, spelling mistakes, and incorrect patient residential addresses), which can render some records redundant. This characterization applies to both databases (Emergency Room Data, 1999-2001, and Hospitalization and Outpatient Visits for Asthma, 1995-2000) that are discussed later.

A database of hospital discharges for asthma (1991-1996) is maintained by the Statewide Planning and Research Cooperative System (SPARCS) of the New York State Department of Health. The SPARCS database is of high quality and contains hospital discharge data from the whole of New York State, excluding psychiatric and federal hospitals. Geographical information from this database is available only at zip code level. The databases are described in detail below.

Emergency Room Data, 1999-2001

The emergency room (ER) data are based on electronic records of patients kept by Buffalo General Hospital, a division of the Kaleida Health System, between November 1999 and April 2001. The records contain the patients' dates of birth and admission, and their place of residence. Using the ER data, the authors were able to analyze the number of visits per patient, the period when the visits were made, and the distances between the patients' place of residence and the hospital. Distance analysis was performed for case address locations and Buffalo General Hospital. The spatial relationships enabled the characterization of relationships between case address locations, hospitals, hazardous sites, the Peace Bridge Complex, and major roadways. The ER data had certain limitations. They did not provide personal information on living conditions, the reasons patients frequented the emergency room, or the reasons they preferred certain health care facilities to others.

Hospital Discharges for Asthma, 1991-1996

Hospital discharges for asthma were obtained from the Statewide Planning and Research Cooperative System (SPARCS) data files in the New York State Department of Health Bureau of Biometrics (NYSDOHBB) database. The data files contained records for all patients admitted with respiratory diseases in New York State. For the period between 1991 and 1998, only data for the study area were extracted. A subset of the database has been reported on in a separate publication (Lwebuga-Mukasa, Oyana, Thenappan, & Ayirookuzhi, 2002). A more detailed analysis was conducted at zip code level for 28 zip codes within the study area. The current study includes the comprehensive database, reflecting health care utilization for the entire study area. The database lacked personal information on the subjects, however, and did not permit identification of multiple visits by the same individual.

Hospitalization and Outpatient Visits for Asthma, 1995-2000

The hospitalization and outpatient visits data were based on patient records kept by the Kaleida Health System. They covered admissions from January 1995 to August 2000. The records contained date of birth, date of discharge, age, and case address. Using these data, the authors were able to analyze the number of visits during this period for each case address, as well as the distance between the case addresses and the hospital. This analysis was vital for the identification of spatial relationships among case address locations, hospitals, and polluting sources. These data did, however, have some limitations. First, they were based on a single hospital system. Second, it was not possible to determine why patients frequented the hospitals. Third, the data set did not contain information pertaining to the living conditions of the patients.

Suspected Sources of Pollution

Figure 2 shows the location of emission sites in the study area. These sites, which include hazardous-material, multiple-release, and toxic-emission sites, release emissions of several types. The locations were downloaded from the U.S. Environmental Protection Agency (U.S. EPA) Web site and mapped as point data. The U.S. EPA definition of multiple emissions involves all classes and grades of chemical substances. When a site releases substances of most or all of these types, U.S. EPA describes it as a multiple-emission site. Five air release sites were identified in the study area. For the current study, the authors considered 33 multiple-release sites, one toxic site, and the five air release sites. The other sites, which were nontoxic hazardous sites, were eliminated. Most of the sites selected are located on Buffalo's west side.

Data Analysis

The data analysis had four major components:

1. exploratory data analysis;

2 GIS mapping;

3. statistical models, which comprised analysis of variance (ANOVA) and multilevel modeling; and

4. spatial methods to identify patterns and analyze hotspots, which comprised spatial autocorrelation, K-means, nearest-neighbor hierarchical spatial clustering, Besag and Newell's method (Besag & Newell, 1996), Turnbull's method, the Score Test of Lawson and Waller (Lawson, 1989; Waller & Jacquez, 1995), and Bithell's Linear Risk Score Test (Bithell, 1995, 1999).

GIS mapping and spatial analysis were conducted with the following GIS software: ARCGIS 8.1.2 and ArcView 8.1 (ESRI, 2002); CrimeStat Version 1.00 (Ned Levine & Associates, 2000); ClusterSeer (Terra-Seer, Inc., Ann Arbor, Michigan); and Centrus Desktop Version 2.1 (Sagent Technology, Inc., Mountain View, California). Statistical and exploratory data analysis was conducted with Microsoft Excel (Microsoft, Inc., Seattle, Washington) and Statistical Package for Social Sciences (SPSS, Inc., Chicago, Illinois). The ANOVA model and Bonferroni's adjustment were implemented with Microsoft Excel and SPSS. Map production was done in Microsoft Visio and Corel Draw (Corel Corporation, Ltd., Ottawa, Ontario).

ER, hospitalization, and outpatient visits for asthma were mapped with points because cases were provided at individual locations. Hospital discharge data from the SPARC database were mapped as areal units because they were provided at the zip code level. The data on hospitalization and outpatient visits for asthma were aggregated to the zip code level to allow comparisons with hospital discharge data. Population census tract and zip code boundary maps were obtained from the Cornell University Geospatial Data Information Repository (CUGIR) and were included as themes (map layers) with the geocoded case points. Once cases were geocoded, they were sorted by population census tract, town, and zip code.

A multilevel approach (Bryk & Raudenbush, 1992; Goldstein, 1995; Longford, 1993; Snijders & Bosker, 1999) was considered appropriate for analyzing hierarchical or clustered data. K-means clustering and nearestneighbor clustering techniques, which are incorporated in CrimeStat Version 1.1, were used for initial screening of the clusters. The minimum number of points per cluster was set at 30. The threshold distance was set at p-value = .05. Additional cluster detection techniques based on Besag and Newell's method (Besag & Newell, 1991) and linear risk score methods (Bithell, 1995, 1999; Lawson, 1989; Waller, Turnbull, Clark, & Nasca, 1992) were used to analyze local or global spatial clusters in a group level data as well as to detect focused spatial clusters. A specialized disease-modeling software, ClusterSeer Version 1.1.4's functionality, includes all these methods. Spatial autocorrelation was measured by Moran's I indices and Geary's C (Bailey & Gatrell, 1995; Chou, 1992).

Spatial queries were conducted to characterize the relationships between case address locations and suspected sources of pollution at varying distances. Specific distances were utilized according to English and co-authors (1999), Versluis (1994), and Fraigneau, Gonzalez, and Coppalle (1995). These studies indicate that 80-90 percent decay of pollutants, as revealed by most emission dispersion models, occurs between 492 and 656 feet (153-204 meters). Rijnders, Janssen, van Vlient, and Brunekreef (2001) also recommend that variables such as degree of urbanization, traffic density, and distance to a near-by highway or any potential pollution source can be used to estimate exposure to traffic-related air pollution. Following the review of work by these authors, the following distance categories were defined for the multilevel model: 0-153 meters (1-153 m), 153-204 m, and 204-700 m.

Results

Geocoding

Geocoding, the process of assigning latitudes and longitudes to physical addresses, was implemented in Centrus Desktop Version 2.1 (Sagent Technology). An interactive matching process with user control over the process was used to increase the likelihood of achieving a match for an address. Of the 548 physical ER addresses that were provided, 94 percent were matched. A match of 80 percent was obtained for 5,601 records of hospitalization data. The geocoded data were all exported to ARCGIS 8.1.2 for further spatial analysis.

Hospitalization and Outpatient Visits for Asthma, 1995-2000, from the Kaleida Database

In the Kaleida database, 5,601 patients were hospitalized as a result of asthma. There were 13,616 hospital discharges, of which the majority (76.7 percent) were for adults (17 to 64 years of age). The average hospital stay was 2.2 days for children, 2.81 for adults, and 1.96 for seniors. Thirty-two percent of those hospitalized were from the city of Buffalo, 3.2 percent from the town of Amherst, 3.1 percent from Williamsville, and 61.7 percent from other places.

Table 1 presents the number of diagnosed cases of asthma, the total population, and the asthma hospitalization rates per 10,000. Hospitalization rates were normalized to the 1990 population census for each zip code. The highest asthma hospitalization rates were observed in three zip codes, 14201, 14213, and 14228. Zip codes 14026, 14068, 14207, 14209, and 14222 also had slightly elevated rates of health care utilization. Together, these eight zip codes had the highest rates of health care utilization in the entire study area.

Hospital admissions increased between 1995 and 2000, especially in zip codes 14201 and 14213. These zip codes also had the highest rates of hospitalization for asthma per 10,000 population. Hospitalization rates for asthma, however, were stable in zip codes 14211, 14215, and 14221, even though the population decreased slightly during the study period.

Emergency Room Data, 1999-2001, from the Kaleida Database

The Kaleida database indicated 548 emergency room visits. Of these, 14 percent were made between December 1999 and April 2000, and 24 percent were made between April 2000 and November 2000. The peak season for most ER visits was between November and April, when 62 percent of the ER visits were made.

The majority (80.6 percent) of ER visits were made by adults (17 to 64 years of age). Most of the ER visits (63.1 percent) during this period were from the city of Buffalo, with highest use of the ER occurring in zip codes 14201 and 14213. The age distribution of the study population reflects the fact that Millard Fillmore Hospital is predominantly an adult hospital; the Children's Hospital of Buffalo, which is located in the study area, takes care of the majority of pediatric population.

Hospital Discharges for Asthma, 1991-1996 from SPARCS Database

Table 2 presents descriptive statistics of hospital discharges between 1991 and 1996. Hospitalization rates for asthma were derived per population of 10,000 people (10K). The denominators for the calculation of the hospitalization rates were obtained from the 1990 population census for each zip code. The lowest and highest mean asthma hospitalization rates were observed in 1991 and 1996, respectively. The mean asthma hospitalization rates decreased slightly over the years. There was a small difference between the mean and standard deviation suggesting that there is a narrow dispersion in asthma hospitalization rates. Confidence intervals for asthma hospitalization rates were narrow during the study period.

Spatial Analysis of Case Address Locations

The analysis of variance (ANOVA) model was implemented to investigate whether the hospital discharge rates obtained from the SPARCS database differed among the zip codes. The results indicated a marked difference in rates of health care utilization among zip codes. The difference was statistically significant (at [alpha] = .05, F-observed = 27.42, F-critical [27, 140] = 1.566, and p-value < .0000). To investigate which zip codes showed a marked difference, a multiple-comparison procedure was undertaken. Bonferroni's technique was favored because of its strong statistical power. A critical difference of 16.4 hospitalization rates for asthma was observed (at [alpha] = .05, df = 135, and 378 possible combinations from N = 28). This comparison showed zip codes located in the west side of the study area standing out strongly when compared with other zip codes located further away (p [less than or equal to] .05). The majority of the zip codes located in the center of the study area showed some differences when compared with the others. About 66 percent of the remaining zip codes were not significantly different (p [less than or equal to] .05). Zip codes located closest to the busily traveled major roadways had high rates of hospitalization for asthma compared with those located further away.

Figure 3 shows a classification of asthma hospitalization rates per 10K of population calculated from the Kaleida database data on hospitalization and outpatient visits. The map indicates that the highest utilization of health care occurred on Buffalo's west side. Nine zip codes had rates of over 111 asthma hospitalizations per 10K. Zip codes 14228 and 14026 stood out because of a small population size (Table 1). Moreover, most of the people residing in zip code 14228 were college students who originated elsewhere. Buffalo's east and south sides had lower rates of health care utilization, as did zip codes 14225, 14227, and 14224 in the neighboring suburb of Cheektowaga. Figure 4 presents the critical-difference results for the SPARCS database, analyzed with Bonferroni's adjustments. The pattern that emerges from this comprehensive database is similar to the one derived from the Kaleida database (Figure 3). Five zip codes had rates of over 106 asthma hospitalizations per 10K. Sixteen zip codes had a median asthma hospitalization rates ranging from 64 to 105 per 10K. Overall, areas with high rates of health care utilization and hospitalization for asthma were near each other; similarly, areas with low rates asthma hospitalization were near each other.

[FIGURE 3 OMITTED]

[FIGURE 4 OMITTED]

Table 3 presents the results of a test of spatial autocorrelation that uses the maps in Figures 3 and 4. For Figure 3, Moran's I was .956 and Geary's C was .034. Figure 4 gave a Moran's I of .944 and a Geary's C of .025. The results indicate a positive spatial autocorrelation (similar, regionalized, and clustered) among health care utilization rates.

The analysis of spatial relationships between case address locations and hospitals showed 43 percent (1,253 of 2,914) and 75 percent (2,185 of 2,914) of the geo-referenced cases of hospitalization were observed within a distance of 3 km and 6 km respectively from Millard Fillmore Health Systems. Forty-three percent (300 of 546) and 74 percent (404 of 546) of the ER cases were observed within a distance of 3 km and 6 km, respectively from Buffalo General Hospital.

Detection of Spatial Clusters

The detection of clusters was based on case address locations for hospitalization and out-patient visits for asthma and ER data from the Kaleida database. A cluster or clusters were defined as an aggregation of cases grouped together in space and time. The authors wished to determine whether ER and hospitalization cases clustered around polluting sites and whether cases that clustered around polluting sites had higher ER and hospitalization rates.

Samples (for hospitalization cases, N = 2,850, and for ER cases, N = 546) were used for the identification of asthma hotspots. Sixteen asthma hotspots were identified on the basis of K-means and nearest-neighbor hierarchical spatial-clustering approaches. There were, however, only eight independent clusters. Both techniques confirmed the presence of hotspots. Notable numbers of asthma hotspots occurred in areas within the city of Buffalo and in areas adjacent to the Peace Bridge Complex. The authors intentionally did not map these clusters because they considered the clusters as candidate hotspots. Surprisingly, areas adjacent to North Tonawanda had a higher number of asthma hotspots than other geographic sections of the study area.

Thirteen percent of hospitalization cases were observed within 200 meters of suspected sources of pollution. Of these, only 3 percent were observed within 200 meters of identified clusters. Three percent and 4 percent of ER cases were observed within 200 meters of identified clusters and within 200 meters of suspected pollution sources, respectively. The authors expected to find many diagnosed cases of asthma within 200 meters of suspected sources of pollution. But most of the cases occurred outside 200 meters, suggesting that the usual model for the decay of pollutants, which is believed to occur between 153 and 204 meters, may not explain the distribution of cases. These observations confirm the need for a direct air quality monitoring at different distances from polluting sites.

Further analysis of data in relation to identified clusters revealed the following distributions: 61 percent and 25 percent of the georeferenced cases of hospitalization were observed within distances of 700 meters and 3 km, respectively, from the identified clusters; and 74 percent of ER cases were observed within a distance of 3 km from the identified clusters. Three clusters (Marnap Industries [multiple releases] in zip code 14201, Birge Company [air release] in zip code 14211, and Ogrady Winnifred [hazardous release] in zip code 14213) were observed within a distance of 200 meters.

Health care utilization differed by distance from the major roads as follows: 37 percent of the geo-referenced cases of hospitalization were observed within a distance of 1 km from the major roads, 32 percent of ER cases were observed within a distance of 1 km from the major roads, and two of eight identified clusters were observed within a distance of 1 km from the major roadway (Interstate Highway 190).

Application of Other Analytical Approaches to Test Identified Disease Clusters

More rigorous spatial techniques were implemented with ClusterSeer (TerraSeer, 2002) to further analyze pollution sources. The authors wanted to test whether the cases were distributed randomly over the study area and, if not, to evaluate any identified spatial disease clusters for statistical significance. They also wanted to test the null hypothesis that the relative risk of asthma was the same for any census tract, or collection of census tracts, and the remaining census tracts, at [alpha] = .05. The study area had 100 census tracts. It had a total of 6,265 cases, and the population at risk was estimated at 395,293 people on the basis of data from the 2000 U.S. Census. To calculate the average disease frequency (.01585) used as a baseline in this analysis, the authors divided the number of cases by the population size.

Besag and Newell's test was applied to search clusters of size 30 (p-value = .001) and 62 (p-value = .007). The selection of these two clusters was arbitrary since the authors had no prior knowledge of how many cases might constitute a cluster. Table 4 lists regions where the authors identified clusters. Four clusters were detected in census tracts 13, 14, 26, and 32. All of the census tracts are located in the west side of the study area. After adjusting for multiple comparisons, the probability of observing p-value = .001 and p-value = .007 under the null hypothesis of uniform asthma risk within the study area was statistically significant.

To determine whether focus sites were associated with an increased risk of asthma, the authors used the Score Test of Lawson and Waller, and the Bithell Linear Risk Score Test. Table 5 gives the results of these tests. Both tests were statistically significant, indicating that there was an excess of asthma cases near the three sites (Marnap Industries, Birge Company, and Ogrady Winnifred) that the authors had previously detected using the nearest-neighborhood hierarchical and K-means clustering techniques. One focus site (the Peace Bridge Complex), however, was not statistically significant by the Score Test of Lawson and Waller. The results show an increased asthma risk as distance to the three sites decreased.

Table 6 gives the test results from Turnbull's method, based on a population radius of 3,952 people, which was used to investigate if any areas had significantly elevated asthma risk. An excess number of asthma cases was found within a population of 3,952 individuals. The first most likely cluster of asthma was detected in census tract 60 (zip code 14213) at a p-value of .0001 and a test statistic of 923.

Figure 5 shows asthma clusters identified by two methods and the distribution of asthma cases per 1,000 people by census tract. The prevalence of asthma was distributed non-homogeneously. In general, the west, east, and north had medium to very high prevalence of asthma; these areas were separated by a central area with very low to low prevalence. The highest prevalence of asthma was observed on Buffalo's west side. A census tract with high prevalence was surrounded by an area of low prevalence on Buffalo's east side. Clusters identified by two methods (Besag and Newell's method and Turnbull's method) were located predominantly on the west side (zip codes 14201 and 14213). A single cluster identified by Turnbull's method was located in the north (zip code 14150).

Multilevel Modeling of Residence Near Suspected Sources of Pollution

[FIGURE 5 OMITTED]

One sample (N = 1,319) was located in close proximity to the 39 suspected sources of pollution. Figures 6a and 6b show selected sources of emissions (air, toxic, and multiple release sites) and buffer distances. Of the 1,319 cases that were aggregated, 19 percent (252 of 1,319) were 153 meters away from suspected sources of pollution; 17 percent (228 of 1,319) were between 153 and 204 meters from suspected sources of pollution; and 64 percent (839 of 1,319) were between 204 and 700 meters from suspected sources of pollution. Of the 1,319 cases of asthma observed, 2 percent occurred in 1995, 26 percent in 1996, 29 percent in 1999, and 43 percent in 2000.

Table 7a shows the analysis of variance results, and Table 7b shows the breakdown of case address locations, over the years, for proximity to source.

Proximity to source (the main effects) was statistically significant (at [alpha] = .05, F-observed = 5.34, F-critical [2, 6] = 5.14 and p-value < 0.25). Case address locations over the years, nested within proximity to source, did not reach statistical significance (at [alpha] = .05, F-observed = 2.26, and F-critical [3, 6] = 4.76). The analysis of the breakdown of the case address locations over the years showed varied results, none of which reached statistical significance. For the two distance categories 0-153 meters and 153-204 meters, the authors obtained very small F-values, as shown in Table 7b. Case address locations within 204 to 700 meters were found to have the closest F-value (5.75) to F-critical (5.99). Although these results did not reach statistical significance, the authors suspect that the nesting of case address locations within proximity of 204 to 700 meters may explain the high spatial variations in the effect that living near a suspected source of pollution has on asthma prevalence and utilization of health care for asthma.

Discussion

The major findings of this study are as follows:

* areas adjacent to the busily traveled roadways and suspected sources of pollution had high rates of health care utilization;

* a two-factor multilevel model showed that proximity to source was statistically significant (at [alpha] = .05 and p-value < .025);

* Turnbull's method found a significant asthma cluster in the vicinity of the focus sites, as well as local clusters on Buffalo's west side, and Besag and Newell's method found a significant global cluster;

* over 40 percent of the people who utilized health care lived within walking distance of the health care facility they patronized; and

* two-thirds of the case address locations for asthma were within 3 kilometers from identified clusters.

These findings are in agreement with previous findings reported by Briggs and coauthors (1999); Kane and co-authors (1999); Donaldson and co-authors (2000); Lwebuga-Mukasa, Wojcik, and co-authors (2002); Lwebuga-Mukasa and Dunn-Georgiou (2002); Dockery (2001); Lwebuga-Mukasa and Pszonak (2001); Lin, Munsie, and coauthors (2002); and Lin, Chen, and coauthors (2002). It is significant that clusters identified on the basis of health care utilization correspond to those identified by crosssectional house-to-house surveys that were conducted six years apart (Lwebuga-Mukasa, Johnson, et al., 2002; Lwebuga-Mukasa, Wojcik, et al., 2002). Studies by Peterson and Saxon (1996) and Kane and co-authors (1999) also found a high prevalence of asthma in these locations.

Zip codes with high rates of health care utilization for asthma tended to be the same over a 10-year period (1991-2001). Hospitalization and outpatient data for 1995-2000, which came from a single institution, Kaleida Health, compared very well with the 1991-1996 multiple-institution hospital discharge data from the SPARCS database. The 10-year data suggested that high rates of hospitalization for asthma were occurring in areas surrounding the Peace Bridge Complex, the major highways, and suspected sources of pollution. The data also suggest that there is geographic variation in health care utilization and the prevalence of asthma within the study area, in agreement with previous studies (Lwebuga-Mukasa & Pszonak 2001; Lwebuga-Mukasa, Oyana et al., 2002). Most people who utilized health care lived within walking distance of the health care facility they patronized. Only one-quarter of the asthma cases occurred within 700 meters of identified clusters. Two-thirds of the cases, however, occurred in people who resided between 204 and 700 meters of major roadways, and one-third of the cases occurred in people who resided within 1 kilometer of major roadways.

Identification of asthma clusters associated with different sources may provide insights into how mixtures of pollutants may interact and lead to development of asthma in susceptible individuals. The geographic distribution of asthma clusters near focus sites and busily traveled roadways suggests that pollutants not only may be associated with worsening of asthma symptoms but also may play a role in the etiology of asthma. Further analysis of common features of pollutants may help elucidate mechanisms by which exposures are related to the genesis of asthma. These asthma clusters also provide a basis for the development of new hypotheses relating to the spatial distribution of asthma prevalence and morbidity in the study area.

The contribution of socioeconomic characteristics was analyzed in three separate studies reviewed by Lwebuga-Mukasa, Oyana, and co-authors (2002). Although indices of lower socioeconomic status were associated with increased asthma rates at zip code level, the strongest association, among populations with different socioeconomic background and race/ethnicity, was with increase in commercial traffic. Two other studies, based on two house-to-house surveys conducted six years apart (a survey of 214 households in 1996-1997 and a survey of 1,608 households in 2002) found that an increased prevalence of asthma on Buffalo's west side compared with other Buffalo communities was not explained by socioeconomic status, median household income, education level, race/ethnicity, smoking status, presence of household triggers such as pets, or cockroaches.

Another study further reviewed demographic characteristics such as housing conditions, ethnic compositions, and age. The analysis aimed at evaluating whether confounding factors were responsible for unusually high prevalence of asthma (Lwebuga-Mukasa, Wojcik et al., 2002). Zip codes 14201, 14213, 14228, and 14026 were of primary interest, given the statistical association of suspected sources of pollution and asthma prevalence (Lwebuga-Mukasa, Oyana, et al., 2002). An extensive analysis of socioeconomic factors such as income, race/ethnicity, and age of housing over the study period showed that residents living in the vicinity of pollution sources are predominantly of low socioeconomic status, but that these factors do not completely explain the high respiratory burden in these zip codes.

[FIGURE 6 OMITTED]

Although the authors expected to find many diagnosed cases of asthma between 153 and 204 meters from suspected sources of pollution, which would have supported the dispersion-of-air-pollution argument, the data arrived at in a two-factor multilevel model did not fulfill that expectation. Instead, the findings show that most cases of asthma occurred in people who resided between 204 and 700 meters from suspected sources of pollution. This observation does not fit with most of the emission dispersion models reviewed by English and co-authors (1999). It is probable that local ecological factors such as the mixture of indoor and outdoor air quality, the urban-heat-island phenomenon due to warming trends occurring in urban areas, residential overcrowding, and, most important, the land and sea breeze phenomena due to the presence of Lake Erie and Niagara River, could be responsible for the delayed dispersion of pollutants. The urban-heat-island phenomenon, which is evident in statistical studies of surface air temperatures (Myer, 1991; Woolum, 1964), may have a local effect on emission dispersion and human health. The importance of an urban forest on the cooling process cannot be overemphasized, and the comparatively low density of vegetation on Buffalo's west side makes these places even warmer.

High particulate emissions from mobile and fixed sources may be a contributing factor in high health care utilization and the prevalence of asthma, particularly on Buffalo's west side. Three clusters within 200 meters of polluting sources were detected in this area. Measurements of particulate emissions conducted by an independent group of scientists (Widmer & Vermette 2002) found a signal extending more than a kilometer in adjacent communities and in areas downwind of the Peace Bridge traffic corridor. Both fine particulates (with particle size of [less than or equal to]2 microns [[micro]m]) and coarse particulates (with particle size of >2 to 5 [micro]m) were dramatically increased, compared with a site further away from the Peace Bridge Complex; fine particulates increased by 22 percent on average in 19 of the 22 samples taken, and coarse particulates increased by 600 percent on average in 20 of the 22 samples taken. The impact of the detected particulates on human health in the surrounding communities needs further study.

The findings of the study reported here are limited because of the scope and nature of information on health care utilization, as discussed in the description of the data sets. It is also possible that the authors are underestimating the problem in the study area, given that some cases of asthma go unreported. The credibility of the findings is strengthened, however, by consistency with findings from previous cross-sectional house-to-house studies of home environmental factors (Lwebuga-Mukasa, Johnson, et al., 2002; Lwebuga-Mukasa, Wojcik, et al., 2002) conducted in the same area. Those studies reported an increased prevalence of asthma that was not explained by the usual personal or household factors.

Conclusions

This study shows a statistically significant association between proximity of residence to pollution release sources and diagnosed asthma. There is a geographical variation in the utilization of health care and the prevalence of asthma in Buffalo neighborhoods. The data further show that residents who live in close proximity to pollution sources and who are of predominantly low socioeconomic status have a high respiratory-disease burden. Asthma diagnosis in this study population was observed predominantly in adults (18-44 years). Notable numbers of asthma clusters were detected in areas within the city of Buffalo, North Tonawanda, major roadways, and the communities adjoining the Peace Bridge Complex. Further studies of particulate emissions are required for a definitive link to be established between increased risk of asthma and identified sources of pollutants.

TABLE 1 Distribution of Diagnosed Asthma Cases by Selected Zip Code,
from the Kaleida Health Database, 1995-2000

Zip    Count       % of          Total      Rate
Code          Diagnosed Cases  Population  (per K)

14228    165   2.95             1,851      891
14201    697  12.44            16,798      415
14213  1,062  18.96            32,748      324
14026      5   0.09               275      182
14203     17   0.30             1,239      137
14068     73   1.30             5,458      134
14222    150   2.68            12,569      119
14207    278   4.96            23,945      116
14209    102   1.82             9,086      112
14216    190   3.39            24,949       76
14202     21   0.37             2,834       74
14221    369   6.59            54,882       67
14204     74   1.32            11,295       66
14226    190   3.39            32,035       59
14208     63   1.12            14,579       43
14212     77   1.37            19,788       39
14215    167   2.98            46,876       36
14217     85   1.52            25,502       33
14211    127   2.27            38,573       33
14214     70   1.25            21,716       32
14150    166   2.96            53,318       31
14223     80   1.43            26,022       31
14210     36   0.64            18,424       20
14227     44   0.79            25,189       17
14206     41   0.73            25,304       16
14220     34   0.61            28,979       12
14218     22   0.39            22,137       10
14219     13   0.11            13,621       10

Note: N = 5,601.

TABLE 2 Annual Rates of Hospitalization for Asthma Per 10,000
Population, 1991-1996--Statistical Summary from the SPARCS Database

Statistic                    1991      1992      1993      1994

Mean                         49.02     58.48     74.07     75.20
Standard error                6.79      8.14     12.32     10.93
Median                       42.70     48.02     61.53     60.99
Standard deviation           35.92     43.05     65.20     57.85
Variance                  1,290.25  1,853.72  4,250.61  3,346.16
Skewness                      1.74      1.97      2.76      2.46
Confidence level (95.0%)     13.93     16.69     25.28     22.43

Statistic                    1995      1996

Mean                         83.90     89.93
Standard error               12.50     12.23
Median                       60.54     73.47
Standard deviation           66.17     64.73
Variance                  4,378.43  4,189.81
Skewness                      2.33      1.65
Confidence level (95.0%)     25.66     25.10

Source: Hospital discharge data for asthma from the NYSDOHBB SPARCS
Database, 1991-1996.

TABLE 3 Assessment of the Spatial Distribution of Asthma Using Data from
Figures 3 and 4

Source          Technique                      Comments
          Moran's I  Geary's C

Figure 3   0.956      0.0338    Similar, regionalized, smooth, clustered
Figure 4   0.944      0.0249    Similar, regionalized, smooth, clustered

TABLE 4 The Distribution of Asthma Case Clusters by Census Tract
According to Besag and Newell's Method

Clustering  Census   Zip            Local    Upper-tail
Region      Tract   Code           Disease    p-value
                                  Frequency

Cluster Size = 62, [alpha] = .05, and Monte Carlo Distribution (999
simulation runs) = .007
16            26    14211          .0530156   .000003*
22            32    14208          .108       .000000*

Cluster Size = 30, [alpha] = .05, and Monte Carlo Distribution (999
simulation runs) = .001
 8            13    14203, 14204   .0357518   .000452*
 9            14    14204          .433962    .000000*

*Statistically significant.

TABLE 5 The Distribution of Clusters of Asthma by Focus Sites in
Buffalo's Neighborhoods

1. Score Test of Lawson and Waller's Method: Average Disease
   Frequency = .0158490, [alpha] = .05

Focus Site         Test Statistics          p-value          Nominal
                                    (9,999 simulation runs)  p-value

Birge Company         10.5923               .0001*           .00000
Marnap Industries     17.7092               .0001*           .00000
Orgady Winnifred
  Silver              30.8443               .0001*           .00000
Peace Bridge
  Complex              3.92227              .9999            .999956

2. Bithell's Linear Risk Score Method: Average Disease Frequency =
   .0158490, [alpha] = .05, [beta] = 100 and [phiv] = 10, Model = 1 +
   [beta]/(1 + d/[phiv]), Conditional

Focus Site               Test Statistics          p-value
                                          (9,999 simulation runs)

Birge Company               3,229.04              .0001*
Marnap Industries           1,276.96              .0001*
Orgady Winnifred Silver     3,425.56              .0001*
Peace Bridge Complex        2,864.02              .0001*

*Statistically significant.

TABLE 6 Distribution of Asthma Case Clusters by Census Tract, Turnbull's
Method

Cluster             Census Tract  Local Disease     Test     p-value
                     Included      Frequency     Statistics

First Most Likely
  Cluster               60          .233422       922.483     .0001*
Second Most Likely
  Cluster               71          .190248       751.861     .0001*
Third Most Likely
  Cluster               79          .0704014      278.226     .0001*

*Statistically significant.
Population radius under consideration = 3,952.
[alpha] = .05; simulations run = 9,999.

TABLE 7 ANOVA Results for Multilevel Analysis

A. ANOVA Table
Source of Variation         Sum of Squares       df     MS        F

Proximity to source (A)     SSA = 59,872.1667     2  29,936.08  5.34
Case address locations      SSB (A) = 38,667.25   3  12,889.08  2.26
 over the years within
 proximity to source
 [B (A)]
Error (E)                   SSE = 34,281.5        6   5,713.58
Total (TO)                  SSTO = 132,820.9167  11

B. Decomposition of SSB (A)
Source of Variation         SSB ([A.sub.i])      df   MSB          F
                                                      ([A.sub.i])

Case address locations      2,916                 1   2,916        0.51
 over the years within
 0-153 m
Case address locations      2,809                 1   2,809        0.49
 over the years within
 153-204 m
Case address locations     32,942.25              1  32,942.25     5.75
 over the years within
 204-700 m
Total                      38,667.25              3

Acknowledgments: This research was supported in part by Grant R01-CCR220259-01 from the Centers for Disease Control and Prevention and the Troup Fund, Kaleida Health Foundations. All research in this paper was approved by the University of Buffalo (UB) Human Investigation Review Board, which oversees all human investigations at UB in accordance with national and institutional guidelines for the protection of human subjects.

REFERENCES

Bailey, T.C., & Gatrell, A.C. (1995). Interactive spatial data analysis. England: Longman Scientific and Technical.

Besag, J., & Newell, J. (1991). The detection of clusters in rare diseases. Journal of the Royal Statistical Society, Series A, 154, 143-155.

Bithell, J.F. (1995). The choice of test for detecting raised disease risk near a point source. Statistics in Medicine, 14, 2309-2322.

Bithell, J.F. (1999). Disease mapping using the relative risk function estimated from areal data. In A.B. Lawson, A. Biggeri, D. Bohning, E. Lesaffre, J.-F. Viel, and R. Bertollini (Eds.), Discase mapping and risk assessment for public health (pp. 247-55). New York: John Wiley & Sons.

Briggs, D.J., Collins, S., Elliott, P., Fischer, P., Kingham, S., Lebret, E., Pryl, K., Van Reeuwijk, H., Smallbone, K., & Van Der Veen, A. (1997). Mapping urban air pollution using GIS: A regressionbased approach. International Journal of Geographical Information Science, 11(7): 699-718.

Bryk, A.S., & Raudenbush, S.W. (1992). Hierarchical linear models: Applications and data analysis methods. Newbury Park, CA: Sage Publications.

Centers for Disease Control and Prevention. (1998). Surveillance for asthma--United States, 1960-1995. Morbidity and Mortality Weekly Report, 47(SS-1), 1-27.

Crain, E.F., Weiss, K.B., Bijur, P.E., Hersh, M., Westbrook, L., & Stein, R.E.K. (1994). An estimate of the prevalence of asthma and wheezing among inner-city children. Pediatrics, 94, 356-362.

Dockery, D.W. (2001). Epidemiologic evidence of cardiovascular effects of particulate air pollution. Environmental Health Perspectives, 109(4), 483-6.

Dockery, W., Pope, A., Xping, X., Spengler, J., Ware, J., Fay, M., Ferris, B., & Speizer, F. (1993). An association between air pollution and mortality in six U.S. cities. New England Journal Medicine, 329, 1753-1759.

Donaldson, K., Gilmour, I.M., & MacNee, W. (2000). Asthma and PM10. Respiratory Research, 1(1), 12-15.

Edwards, J., Walters, S., & Griffiths, R.K. (1994). Hospital admissions for asthma in preschool children; relationship to major roads in Birmingham, United Kingdom. Archives of Environmental Health, 49, 223-227.

English, P., Neutra, R., Scalf, R., Sullivan, M., Waller, L., & Zhu, L. (1999). Examining associations between childhood asthma and traffic flow using a geographical information system. Environmental Health Perspectives, 107, 761-767.

ESRI, Inc. (2002). ARCGIS 8.1.2. and ARCVIEW 8.1. [Computer software]. Redlands, CA: Environmental Systems Research Institute (ESRI).

Fraigneau, Y.C., Gonzalez, M., & Coppalle, A. (1995). Dispersion and chemical reaction of a pollutant near a motorway. The Science of the Total Environment, 169, 83-91.

Goldstein, H. (1995). Multilevel statistical models. London, Edward Arnold; New York: Halstead Press.

Haahtela, T., Jarvinen, M., Kava, T., Kiviranta, K., Koskinen, S., Lehtonen, K., Nikander, K., Persson, T., Reinikainene, K., Selroos, O. (1991). Comparison of a beta 2 agonist, terbutaline, with an inhaled corticosteroid, budesonide, in newly detected asthma. New England Journal of Medicine, 325, 388-392.

Ishizaki, T., Koizumi, K., Ikermori, R., Ishiyama, Y., & Kushibiki, E. (1987). Studies of the prevalence of Japanese cedar pollinosis among residents in a densely cultivated area. Annals of Allergy, Asthma, and Immunology, 58, 265-70.

Kane, M., Jaen, C.R., Tumiel, L., Gonzalo, M.S., Bearman, M.L., & O'Shea, R. (1999). Unlimited opportunities for environmental interventions with inner-city asthmatics. Journal of Asthma, 36(4), 371-79.

Koenig, J.Q. (1999). Air pollution and asthma. The Journal of Allergy and Clinical Immunology, 104(4 Pt. 1), 717-722.

Lawson, A.B. (1989). Score tests for detection of spatial trend in morbidity data. Dundee, United Kingdom: Dundee Institute of Technology.

Lin, M., Chen, Y., Burnett, R.T., Villeneuve, P.J., & Krewski, D. (2002). The influence of ambient coarse particulate matter on asthma hospitalization in children: Case-crossover and time-series analyses. Environmental Health Perspectives, 110(6), 575-581.

Lin, S., Munsie, J.P., Hwanga, S., Fitzgerald, E., & Cayo, M.R. (2002). Childhood asthma hospitalization and residential exposure to state route traffic. Environmental Research Section A, 88, 73-81.

Loh, P., Sugerman-Brozan, J., Wiggins, S., Noiles, D., & Archibald, C. (2002). From asthma to airbeat: Community-driven monitoring of fine particles and black carbon in Roxbury, Massachusetts. Environmental Health Perspectives, 110(Suppl. 2), 297-301.

Longford, N.T. (1993). Random coefficient models. Oxford: Clarendon Press.

Lwebuga-Mukasa, J.S. & Pszonak, R. (2001). Patterns of asthma hospitalization in Western New York. Journal of Asthma, 38(2), 1-6.

Lwebuga-Mukasa, J.S., Ayirookuzhi, S.J., & Hyland, A. (2002). Traffic volumes and respiratory health care utilization among residents in close proximity to the Peace Bridge before and after September 11, 2001. Journal of Asthma, 40(8), 855-864.

Lwebuga-Mukasa, J.S. & Dunn-Georgiou, E. (2002). A schoolbased asthma intervention program in Buffalo, New York, Schools. Journal of School Health, 72(1), 27-32.

Lwebuga-Mukasa, J.S., Johnson, C., & Oyana, T.J. (2002). Local ecological and atmospheric conditions and healthcare utilization for asthma. Manuscript submitted for publication.

Lwebuga-Mukasa, J.S., Oyana, T.J., Thenappan, A., & Ayirookuzhi, S.J. (in press). Association between volume and health care utilization for asthma among residents at a U.S.-Canada border crossing point. Journal of Asthma.

Lwebuga-Mukasa, J.S., Wojcik, R., Dunn-Georgiou, E., & Johnson, C. (2002). Home environmental factors associated with increased respiratory symptoms in patients with asthma. Journal of Health Care for the Poor and Underserved, 13(2), 214-228.

Microsoft Excel (2002). Microsoft Corporation. http://www.microsoft.com (31 Jan. 2002).

Chou, Y.H. (1992). Management of wildfires with a geographical information system. International Journal of Geographical Information System, 6(2), 123-140.

Myer, W.B. (1991). Urban heat island and urban health: Early American perspective. Professional Geographer, 43(1), 38-48.

National Institute of Allergy and Infectious Diseases, National Institutes of Health. (2001). Focus on asthma report. http://www.niaid.nih.gov/default.htm. (20 Sept. 2001).

Ned Levine & Associates. (2000). CrimeStat Version 1.1.7: A spatial statistics program for analysis of crime incident locations. Washington, DC: The National Institute of Justice.

NILU. (1991). Air pollution and short-term health effects in an industrialized area in Norway. Main Report, Oslo, Norway.

Norris, G., YoungPong, S.N., Koenig, J.Q., Larson, T.V., Sheppard, L., & Stout, J.W. (1999). An association between fine particles and asthma emergency department visits for children in Seattle. Environmental Health Perspectives 107(6), 489-493.

Pandya, R.J., and Solomon, G., Kinner, A., & Balmes, J.R. (2002). Diesel exhaust and asthma: Hypotheses and molecular mechanisms of action. Environmental Health Perspectives, 110(1), 103-112.

Peace Bridge Review Commission. (1999). Peace bridge public review: Interim report, Buffalo, NY: Author.

Peterson, B. & Saxon, A. (1996). Global increases in allergic respiratory disease: The possible role of diesel exhaust particles. Annals of Allergy, Asthma, and Immunology, 77(4), 269-70.

Rijnders, E., Janssen, N.A.H., van Vlient, P.H.N., & Brunekreef, B. (2001). Personal and outdoor nitrogen dioxide concentration in relation to degree of urbanization and traffic density. Environmental Health Perspectives, 109(Suppl. 3), 411-417.

Schwartz, J. (1993). Particulate air pollution and chronic respiratory disease. Environmental Research 62, 7-13.

Schwartz, J. (1994). Air pollution and daily mortality: A review and meta-analysis. Environmental Research, 64, 36-52.

Schwartz, J. (1996). Air pollution and hospital admissions for respiratory disease. Epidemiology, 7, 20-28.

Snijders, T.A.B., & Bosker, R.J. (1999). Multilevel analysis: An introduction to basic and advanced multilevel modeling. London: Sage Publishers.

Takenaka, H., Zhang, K., Diaz-Sanchez, D., Tsien, A., & Saxon, A. (1995). Enhanced human IgE production results from exposure to the aromatic hydrocarbons from diesel exhaust: direct effects on B-cell IgE production. The Journal of Allergy and Clinical Immunology, 95, 103-15.

TerraSeer Environmental Insight. (2002). ClusterSeer[TM] Version 1.1.4: Software for identifying disease clusters. Ann Arbor. MI: Author.

U.S. Census Bureau. (1990). Census of population and housing. Washington, D.C.: U.S. Department of Commerce. http://www.census.gov (15 May 2002).

U.S. Census Bureau. (2000). Census of population and housing. Washington, D.C.: U.S. Department of Commerce. http://www.census.gov (15 May 2002).

Versluis, A.H. (1994). Methodology for predicting vehicle emissions on motorways and their impact on air quality in the Netherlands. The Science of the Total Environment, 146/147, 359-364.

Waller, L.A., & Jacquez, G.M. (1995). Disease models implicit in statistical tests of disease clustering. Epidemiology, 6, 584-590.

Waller, L.A., Turnbull, B.W., Clark, L.C., & Nasca, P. (1992). Chronic disease surveillance and testing of clustering of disease and exposure: Application to leukemia incidence and TCE-contaminated dumpsites in upstate New York. Environmetrics, 3(3), 281-300.

Weiland, S.K., Mundt, K.A., Ruckmann, A., & Keil, U. (1994). Selfreported wheezing and allergic rhinitis in children and traffic density on street of residence. Annals of Epidemiology, 4, 243-247.

Widmer, T. & Vermette, S. (2002, April). Impact of the Peace Bridge traffic on airborne particulates. Poster presentation at the 4th Annual Student Research and Creativity Celebration, Buffalo State College.

Wijst, M., Reitmeir, P., Dold, S., Wulff, A., Nicolai, T., von Loeffelholz-Colberg, E., & von Mutius, E. (1993). Road traffic and adverse effects on respiratory health in children. British Medical Journal, 307(6904), 596-600.

Wong, T.W., Lau, T.S., Yu, T.S., Neller, A., Wong, S.L., Tam, W., & Pang, S.W. (1999). Air pollution and hospital admissions for respiratory and cardiovascular diseases in Hong Kong. Occupational & Environmental Medicine, 56(10), 679-683.

Woolum, C.A. (1964). Notes from a study of the microclimatology of the Washington, D.C. area for the winter and spring seasons. Weatherwise, 17(6), 263-271.

Tonny J. Oyana, M.Sc., Ph.D.

Jamson S. Lwebuga-Mukasa, M.D., Ph.D.

Corresponding Author: Jamson S. Lwebuga-Mukasa, Associate Professor of Medicine, Director, Pulmonary and Critical Care Division, Center for Asthma and Environmental Exposure, Lung Biology Research Program, Department of Internal Medicine, UB School of Medicine and Biomedical Sciences, Kaleida Health Buffalo General Division, 100 High Street, Buffalo, NY 14203. E-mail: [e-mail].