4. Discussion
This study implemented, for the first time, a spatial lag grid analysis of Hierarchical Bayesian Model (HBM) assembled experimental AOD-PM
2.5 and baseline PMB fused surfaces and respiratory-cardiovascular ED visits and IP hospitalizations with patient residences uniquely assigned to 12 km
2 CMAQ grids. Because the spatial lag grid analyses were completed using the same concatenated exposure-health outcome data files that were previously used to evaluate temporal lag days [
3], it was possible to identify unique differences and similarities associated with these two case-crossover data analysis procedures. Unique findings include a description and interpretation of differences among the size of homogeneous spatial areas (HOSAs), greater risk of manifesting a respiratory-cardiovascular chronic disease as a result of exposure to higher AOD-PM
2.5 concentration levels in rural areas, and warm-cold season differences that showed associations between ambient temperature and AOD-PM
2.5 concentration levels, and increases in respiratory-cardiovascular chronic disease ED visits for asthma and IP hospitalizations for asthma, MI, and HF. Each of these four topics are detailed below after reviewing correlations and descriptive statistical analyses for baseline PMB and the four experimental AOD-PM
2.5 fused surfaces.
Correlations and three-year fine PM concentration level means suggest that PMCK may be more representative of ambient PM2.5 concentration levels in urban grids with air monitors and in rural grids without air monitors than baseline PMB, or the other three AOD-PM2.5 fused surfaces. The percentage decrease in shared variance between PMB-PMC and PMB-PMCK in grids with and without air monitors was 32.4% and 35.7%, respectively. In grids without air monitors, the baseline PMB concentration levels reflect less emphasis on air monitor readings and greater importance on CMAQ PM2.5 model estimates. In grids without air monitors, PMC uses AOD-PM2.5 readings (with missing values) while PMCK utilizes Kriged AODK-PM2.5 readings (without missing values) to estimate ambient fine PM concentrations. One interpretation could be that in grids without air monitors, AOD-PM2.5 concentration levels are more representative of ambient monitor PM2.5 measurements than the Community Multiscale Air Quality (CMAQ) PM2.5 model estimates.
The first aim was to determine the size of homogeneous spatial areas, HOSAs. HOSA sizes in km for width and height or km2 for area differed as a function of AOD-PM2.5 fused surface, respiratory-cardiovascular ED visits or IP hospitalizations, and monitor grid condition. The largest HOSAs included 5 interconnected grids (720 km2), and the smallest HOSAs were two grids wide (288 km2). In grids with monitors, there was only one HOSA two grids wide for PMCK and ED asthma. In grids without air monitors, there were seven HOSAs five grids wide: three each for PMC or PMCK paired with ED asthma, IP MI, or HF, and one for PMCKQ with ED asthma. The grid sizes of HOSAs in grids without monitors were identical to the grid sizes of HOSAs in the Baltimore study area for PMCKQ, and similar for PMC and PMCK. In the Both grid monitor condition, three HOSAs five grids wide occurred for PMC with ED asthma, and PMCK with ED asthma or IP HF. These maximum grid size HOSA results suggest that the benefit of utilizing the experimental AOD-PM2.5 fused surfaces may be only evident when the analyses are completed in rural grids without air monitors. The other implication is that the same association between an experimental AOD-PM2.5 fused surface and a specific respiratory-cardiovascular ED visit or IP hospitalization outcome can occur up to 60 km from lag grid 0. It is possible that residents in these five interconnected rural grids without air monitors were at equal risk of exposure to higher fine PM concentration levels and in need of medical care for ED asthma or an overnight stay in the hospital for MI or HF. Residents in two interconnected rural grids without air monitors were also at risk of exposure to higher fine PM concentration levels and in need of medical care as an IP because of the onset of uncontrolled asthma attacks.
The second objective concerned the presence of differences in AOD-PM
2.5 concentration levels between urban grids with air monitors and rural grids without air monitors. The contribution of PMC or PMCK concentration levels to ED asthma, IP asthma, MI, and HF produced significantly higher ORs in grids without air monitors than in grids with air monitors at lag grids 0, 1, and 01. Interestingly, the no monitor–monitor odds ratio difference percent (∆OR%) values were positive and larger for PMC and PMCK than for PMCKQ fused surfaces. The ∆OR% values were larger at lag grids 0, 1, and 01 than at lag grid 04. It is possible that the PMC and PMCK concentration levels were more accurate approximations of ambient fine PM values in urban grids with air monitors and in rural grids without air monitors than the PMCKQ concentration levels. Because the number and percentage of true positive cases and true negative controls were similar in urban grids with air monitors and in rural grids without air monitors, we concluded that the concentration–response function between AOD-PM
2.5 concentration levels and respiratory-cardiovascular ED visits and IP hospitalizations could be similar in the entire Baltimore study area, and in urban grids with air monitors and in rural grids without air monitors [
80]. One limitation, however, is that this statement is based on a temporal lag day analysis and does not utilize the spatial lag grid case-crossover design analysis.
Categorical analyses of the Baltimore study area identified, for the first time, a subset of homogeneous spatial grids in rural areas without ambient air monitors that resembled urban grids with ambient air monitors by showing higher poverty percent and increased population density. In rural areas without air monitors, there were eight grids with poverty percent values in the Above category, and seven grids in urban areas with air monitors that had poverty percent values in the Above category. There were 19 rural grids with population density in the Above category and 11 urban grids in the Above category. In addition, there were higher totals and percentages of IP MI (12,016, 62.6%) and IP HF (15,684, 57.0%) in rural areas than in urban areas (7185, 37.4% and 11,834, 43.0%, respectively). The totals and the percentages of Black ED asthma patients (11,844, 25.2%) and IP asthma patients (2358, 17.5%) were higher in urban areas than in rural areas (10,852, 23.1% and 2152, 16.0%, respectively). These results support the conclusion that persons living in rural grids are also at risk of developing respiratory-cardiovascular chronic diseases after exposure to higher fine PM concentration levels, as are persons who reside in urban grids.
Hirshon and associates [
81] evaluated PM
2.5 zinc concentration levels and children’s asthma ED visits and IP hospitalizations. These authors reported that PM
2.5 zinc concentrations contributed to increases in asthma hospital events. This publication did not identify the zinc source, however. One explanation could be that the ambient PM
2.5 zinc levels recorded at the Baltimore PM
2.5 Supersite [
82] may have come, in part, from a nearby Toxic Release Inventory site [
83]. The Baltimore PM
2.5 Supersite location can be remapped onto one of the 99 CMAQ grids utilized in the Baltimore study area. As a result of this remapping, we discovered that the location of the Baltimore PM
2.5 Supersite is in CMAQ grid R6, C6. This same grid also had three federal reference method PM
2.5 air monitors and one U.S. Environmental Protection Agency-identified Toxic Release Inventory facility that emitted ambient zinc fumes and dust. An adjacent Baltimore study area grid, R
6, C
7, had one ambient PM
2.5 air monitor and one Toxic Release Inventory facility that released zinc fumes and dust in the air.
As a follow-up to the Hirshon et al. study [
81], we also looked at the number of Toxic Release Inventory facilities that released zinc fumes or dust in the Baltimore study area between 2004 and 2006. There were five different businesses operating during this three-year timeframe. One company had facilities in two distinct locations, contributing a total of 11 zinc point sources: seven zinc fumes or dust point sources in grids with air monitors (urban grids) and four in grids without air monitors (rural grids). This analysis suggests that the ambient PM
2.5 zinc measured at the Baltimore PM
2.5 Supersite in 2002 could have originated from a nearby Toxic Release Inventory facility that emitted zinc fumes and/or dust. Although the federal reference method PM
2.5 air monitors data we utilized did not include PM
2.5 zinc measurements, it is possible that ambient zinc from Toxic Release Inventory zinc-emitting facilities in selected at-risk grids (with or without air monitors) in the Baltimore study area could have indirectly contributed to children’s asthma ED visits and IP hospitalizations.
Although this study did not evaluate environmental hazards associated with living close to brownfields [
84,
85] or U.S. Environmental Protection Agency Toxic Release Inventory facilities [
83], several published studies have described the environmental contamination from manufacturing efforts in Maryland that have included U.S. Environmental Protection Agency Toxic Release Inventory facilities in the state [
86,
87,
88,
89] and brownfields in south Baltimore [
84,
85]. Perlin, Sexton, and Wong [
87] found that there were 122 Toxic Release Inventory sites in Maryland. About half of the Toxic Release Inventory facilities were in Baltimore city, Howard, Anne Arundel, and Baltimore counties. Maryland residents near a Toxic Release Inventory site were medically underserved [
89]. South Baltimore brownfields have higher respiratory and heart disease mortality rates among White working-class residents than the rest of Baltimore city and state [
84]. Litt, Tran, and Burke [
85] described a variety of environmental hazards in southeast Baltimore that included heavy metals, solvents, and insecticides. Living in these environmentally compromised areas for long durations could increase residents’ adverse responsiveness to lower ambient PM
2.5 concentration levels and their enhanced contribution to respiratory-cardiovascular ED visits or IP hospitalizations [
84,
85,
90,
91].
The third aim was to spatially evaluate warm–cold season differences. For the Baltimore study area, there were significant correlations between ambient temperature and fused surface concentration levels. The percentage of shared variance was highest between ambient temperature and PMCK in all three grid conditions. After controlling for experienced apparent temperature in the CLRs, only the PMCK AOD-PM2.5 fused surface had the highest number of significant ORs, thereby representing a greater risk of warm season effects on the occurrence of respiratory-cardiovascular ED visits and IP hospitalizations at lag grids 0 (all four health outcomes), 1 (ED asthma, IP MI, and HF), 01 (all four health outcomes), and 04 (IP asthma, MI, and HF). The PMCK warm-cold season odds ratio difference percent (∆OR%) values were smallest at lag grid 04 compared to lag grids 0, 1, and 01. Unlike these spatial lag grid warm-cold season results, the temporal lag day outcomes had (1) no significant lag day 04 and (2) higher ∆OR% values for the five fused surfaces and four health outcomes at lag days 1, and 01. These new findings suggest that the warm-cold season differences depend on what AOD-PM2.5 fused surface is used and the implementation of the spatial lag grid analysis or temporal lag day analysis.
Additional results dealing with season differences appear to increase our understanding of the complex relationship between ambient pollution sources and their synergistic interaction with house heating fuel type, urban-rural environments, ambient temperature, and AOD-PM2.5 fused surface characteristics. During the cold season, all correlations between ambient temperature and PMB AOD-PM2.5 concentration levels were significant and negative. This inverse correlation suggests that higher fine PM levels were associated with lower ambient temperatures. For the five fused surfaces, the percent of shared variance (r2%) measures were higher in rural grids without air monitors than in urban grids without air monitors. Three-year ambient temperature means were significantly higher in urban grids with air monitors than in rural grids without air monitors during the warm and cold seasons. The use of oil, kerosine, or similar carbon-based products as a home heating fuel was higher in rural grids without air monitors than in urban grids with air monitors. The use of oil as a house heating fuel could be another unique contributor to the documented elevated AOD-PM2.5 concentration levels in rural grids without air monitors. Finally, spatial autocorrelation analyses showed the absence of a positive relationship only for PMC and PMCK fused surfaces during the cold season.
The last aim was to compile this study’s findings regarding differences between the spatial lag grid case-crossover method used in this report and the temporal lag day case-crossover method previously implemented in an earlier publication [
3]. There were more differences than similarities between lag grids and lag days. First, the longest lag day was two (lag day 01), while the longest lag grid was five (lag grid 04). Second, the longer lag grid (0, 1, 01, 04) and shorter lag day (0, 1, 01) values were also found for the monitor and season analyses. Third, the no monitor-monitor odds ratio difference percent (∆OR%) lag grid values were either the same, higher for lag days, or higher for lag grids based on what fused surface was used to analyze a specific respiratory-cardiovascular chronic disease. Fourth, the warm-cold season ∆OR% lag grid values were either the same (lag value 0) or higher for lag days than lag grids (lag values 1, or 01). Fifth, the lag grid and lag day analyses showed that PMCK identified more significant warm season ORs than either PMC or PMCKQ.
Work on the adverse effects of ultrafine PM on a variety of physiologic measures and health outcomes has continued for three decades, but it is now becoming more relevant given the unequivocal evidence of the detrimental effects of ambient and modeled fine PM on many health outcomes [
28]. Mechanistically, ultrafine PM’s adverse effects should be even more severe than fine PM’s adverse effects on the occurrence of respiratory-cardiovascular chronic diseases [
21,
22,
24,
25,
27,
28,
29,
92]. It is possible that technical issues related to the use of the selected data analytic methods could explain why there have not been more epidemiologic studies reporting on the significant association between ultrafine PM and respiratory-cardiovascular outcomes [
93,
94,
95]. Other possibilities could include the ambient ultrafine PM air monitor location and its distance from the residential addresses of study participants [
19,
26,
93,
94]; the type of study participant selected with pre-morbid conditions that would enhance the adverse effects of ultrafine PM on the occurrence of a respiratory-cardiovascular chronic disease [
20,
26,
27,
29]; and that ultrafine PM’s adverse physiologic effects could compromise other organs besides the lungs and heart, e.g., liver, kidney, brain, and this type of system-wide structural damage and physiologic modification could facilitate the development of respiratory-cardiovascular chronic diseases at a later date [
22,
23,
26,
28].
The suggestion introduced here is that the accuracy of AOD-PM
2.5 concentration levels in estimating ambient monitor PM
2.5 measurements is related to spatial scale [
38,
96,
97,
98,
99]. Remote sensing studies that have used AOD to estimate ambient PM
2.5 measurements have concluded that smaller grids provide greater accuracy than larger grids. We utilized the National Aeronautics and Space Administration 10 km
2 grid when we accessed the AOD unitless readings for the Baltimore and New York City study areas. Since our initial objective was to evaluate the accuracy of the four experimental AOD–PM
2.5 concentration level fused surfaces relative to the performance of the previously developed baseline PMB, we decided to map the concentration level values to CMAQ’s 12 km
2 native grid system before the Baltimore and New York City epidemiologic studies were undertaken. For the Baltimore and New York City study areas, we controlled for scale effects by using the CMAQ 12 km
2 grid system as the smallest spatial area of analysis. In the Baltimore study analyses, all data files with different spatial polygons, Zone Improvement Plan (ZIP) codes for ED visits and IP hospitalizations, and ZIP Code Tabulation Areas (ZCTAs) for population density and poverty measures were mapped to CMAQ grids. In subsequent analyses, the CMAQ grids were categorized into one group including air monitors (urban grids), and another group without air monitors (rural grids). The larger groups, consisting of urban grids with air monitors or rural grids without air monitors, should not be less accurate than the individual 12 km
2 grids since the larger areas represent the inclusion of 12 km
2 grids into larger spatial grid areas. Likewise, the accuracy of the identified homogeneous spatial areas (HOSAs) and heterogenous spatial areas should have the same accuracy as each of the individual Community Multiscale Air Quality (CMAQ) grids.
There are strengths in the analytical methods utilized in this spatial lag grid case-crossover study. To our knowledge, this is the first time the temporal lag day case-crossover design has been modified to evaluate lag grids. The selection of controls that preceded or followed the cases makes the lag grid case-crossover design spatially bidirectional [
66]. The spatial lag grid analytic method permits the identification and assessment of homogeneous spatial areas (HOSAs) and heterogeneous spatial areas. The study results demonstrate, also for the first time, the presence of rural grids without air monitors that resembled urban grids with air monitors in exposing rural residents to increased fine PM concentration levels and seeking medical care for respiratory-cardiovascular chronic disease hospital events. A final contribution was confirmation that PMCK could become a replacement for baseline PMB and its use in epidemiologic studies to evaluate the contribution of AOD–PM
2.5 concentration levels on the future occurrence of respiratory-cardiovascular ED visits and IP hospitalizations in urban grids with ambient air monitors and in rural grids without ambient air monitors.
There are unresolved methodological issues that limit the generalizability of these study results. The 99 CMAQ grids that defined the Baltimore study area included 15 urban grids with air monitors and 84 rural grids without air monitors. All 15 urban grids with monitors had associated respiratory–cardiovascular hospital event data. Only 57 of the 84 rural grids without air monitors had respiratory-cardiovascular hospital event data. Grids without air monitors that lacked health data could have included some of those grids over the Chesapeake Bay. The Chesapeake Bay CMAQ grids, located in the south-east corner of the Baltimore study area, included more water than land mass. Residents live on part of Maryland’s irregular coastline and islands. Underestimates for total patients with respiratory-cardiovascular ED visits and IP hospitalizations could have occurred because some Maryland residents can and do obtain medical treatment out of state, e.g., in Washington, DC; Virginia; or Pennsylvania. To be consistent with the way the linear boundaries for the New York City study area and the Baltimore study area were established for the purpose of developing the AOD–PM
2.5 and PMB fused surfaces, it was necessary to include all 99 CMAQ grids in the Baltimore study area. There were boundary grids that crossed the Maryland state line into neighboring states. It was not possible to identify which grids were included in homogeneous spatial areas (HOSAs). Based on the way the spatial lag grid analytical method was implemented, only HOSA size could be determined. There was no independent confirmation of actual ambient PM
2.5 concentration levels in rural grids without air monitors. There was also (relative) spatial heterogeneity in the 15 urban grids with air monitors because there were only 17 ambient air monitors for an area of (12 km x 12 km x 15 grids) 2160 km
2. If the 17 fine PM ambient air monitors were equally distributed among the 15 CMAQ grids, each monitor grid would occupy 127 km
2. Monitor accuracy for the fine PM measurements is highest at the monitor’s location. Fine PM measurement accuracy decreases as the distance from the monitor increases [
8,
33]. It is possible that ambient PM
2.5 concentration levels in the most distant grid with a lag grid value of 4 was the same at the ambient PM
2.5 concentration level in lag grid 0. It is not clear to what extent health care access impacted the results. Available evidence indicates that this may not have been a bias since, for some respiratory-cardiovascular chronic diseases included in the Baltimore study area, there were more patients with the four respiratory-cardiovascular chronic diseases in rural grids without air monitors than in urban grids with air monitors.
Future research efforts should involve the identification of criteria that will lead to the replacement of the currently used baseline PMB with another AOD-PM
2.5 fused surface, e.g., PMCK. Relevant attributes that could facilitate the selection of an updated AOD-PM
2.5 baseline could include grid resolution below 10 km
2, absence of missing AOD unitless readings, and improved accuracy of PM
2.5 concentration levels in estimating ambient fine PM concentration levels in grids without ambient air monitors. In addition, to increase the reliability and validity of AOD-PM
2.5 fused surface concentration levels in grids without air monitors, it will be necessary to have available, independent, on-the-ground ambient PM
2.5 measurements in grids without ambient air monitors. This more ambitious goal could be reached by using portable and accurate PM
2.5 monitors to supplement fine PM readings available from the U.S. Environmental Protection Agency Air Quality System national network of ambient air monitors [
50]. In addition, ultrafine PM monitoring should be included along with fine PM monitoring within selected communities in urban and rural areas. The overall goal of these proposed improvements is to protect the respiratory-cardiovascular health of residents from the adverse consequences of breathing ambient air with elevated fine PM and ultrafine PM concentration levels wherever they live, in urban or rural areas of Maryland or other states in the U.S.