particulate air pollution and daily mortality in detroit

ENVIRONMENTAL RESEARCH 56, 204-213 (1991)

Particulate Air Pollution and Daily Mortality in Detroit

JOEL SCHWARTZ

United States Environmental Protection Agency, PM 221, 401 M Street, SW, Washington, DC 20460

Received March 20, 1991

Particulate air pollution has been associated with increased mortality during episodes of high pollution concentrations. The relationship at lower concentrations has been more con- troversial, as has the relative role of particles and sulfur dioxide. Replication has been difficult because suspended particle concentrations are usually measured only every sixth day in the U.S. This study used concurrent measurements of total suspended particulates (TSP) and airport visibility from every sixth day sampling for 10 years to fit a predictive model for TSP. Predicted daily TSP concentrations were then correlated with daily mortality counts in Poisson regression models controlling for season, weather, time trends, overdispersion, and serial correlation. A significant correlation (P < 0.0001) was found between predicted TSP and daily mortality. This correlation was independent of sulfur dioxide, but not vice versa. The magnitude of the effect was very similar to results recently reported from Steubenville, Ohio (using actual TSP measurements), with each 100 p~g/m 3 increase in TSP resulting in a 6% increase in mortality. Graphical analysis indicated a dose-response relationship with no evidence of a threshold down to concentrations below half of the National Ambient Air Quality Standards for particulate matter. © 1991 Academic Press, Inc.

INTRODUCTION

Episodes of high concentrations of smog were associated with substantial increases in daily mortality in Donora, Pennsylvania in 1948 (US EPA, 1982), in London, England in 1952 (Her Majesty's Public Health Service, 1954), and in the Meuse Valley in Belgium in 1930 (Firket, 1936). The rapid rise and fall in pollution concentrations, and a similar pattern in the daily mortality counts, left little doubt that high concentrations of smog are associated with excess mortality. The 24-hr mean particulate and SO2 concentrations in London during the episode were in excess of 1000 p~g/m 3, and often several thousand micrograms per cubic meter. The concentrations in Donora and the Meuse Valley were likely similar. Retro- spective analyses of similar episodes in previous centuries have also shown excess mortality (Brimblecombe, 1987; Her Majesty's Public Health Service, 1954), and indeed the growth of coal consumption has been anecdotally linked to ill health at least since 1257, when Queen Eleanor fled Nottingham castle after coal burning was introduced (Brimblecombe, 1987). A recent air pollution episode in West Germany (Wichmann et al., 1989) found less dramatic increases in mortality associated with increased particulates and SO2 levels at lower, but still high, concentrations.

Studies of day to day fluctuations in daily mortality and air pollution in London have shown associations across a broad range of pollution concentrations with no evidence of a threshold (Schwartz and Marcus, 1990; Mazumdar et al., 1982). The relationship with particulates was independent of SO2 but not vice versa. A rela-

0013-9351/91 $3.00 Copyright © 1991 by Academic Press, Inc. All rights of reproduction in any form reserved.

204

POLLUTION AND MORTALITY IN DETROIT 205

tionship between annual average smoke concentration and annual mortality from bronchitis in London was also noted by Reid (1956). The London data used an optical measure of particulate concentration rather than a direct gravimetric measure. This measure (British smokeshade index) was nonlinearly related to total suspended particulates (TSP) in a site-specific manner (Holland et al., 1979), and calibration was not available continuously across the study period. Coefficient of haze, another optical measure of particle concentration, has also been associated with daily fluctuations in mortality in New York (Schimmel and Murawski, 1976), Philadelphia (Wyzga 1978), and most recently at much lower pollution levels in Santa Clara, California (Fairley, 1990). The lack of a clear translation between these measures and particulate concentrations measured by current methods makes comparisons with current levels difficult. It also makes it difficult to rep- licate the finding of an association with particulates stronger than with SO2, since different particle measures are used in each study.

Studies using direct gravimetric measurements of particle concentrations in the U.S. have been limited by the every sixth day sampling regime for particulate monitors. One recent study in Steubenville, Ohio (Schwartz and Dockery, 1990, 1991a), where monitoring was available every day, has found a significant association between daily mortality and TSP that was independent of SO2, but not vice versa. The relationship appeared to continue to well below current ambient air quality standards. Replication of such results is clearly critical.

While particulates are generally only monitored every sixth day in the U.S., visible range data is available several times per day at all major airports. Light scattering by particles is a primary factor limiting visibility. Not surprisingly, visibility extinction coefficients have been shown to be highly correlated with particulate matter in the Eastern U.S. (Ozkaynak et al., 1986). This study uses data from over 500 days when both particulate concentrations and extinction coefficients were available in Detroit, Michigan to develop a site-specific predictive model for TSP. Daily concentrations of predicted TSP are then assessed as predictors of daily mortality. This specification allows a direct comparison of the TSP coefficient with the coefficient found in Steubenville and of the relative importance of TSP and SO2. In addition, both 1-hr peak ozone concentrations and the daily mean ozone concentrations were examined for their relationship to daily mortality.

DATA AND METHODS

Daily deaths of residents of the city of Detroit were extracted from the detail mortality tapes of the National Center for Health Statistics for the calendar years 1973-1982. Deaths from accidental causes (ICD 9 /> 800) and deaths occurring outside of the city were excluded from the analysis. These daily counts were then matched to the 24-hr average (midnight to midnight) of TSP and SO2 from all of the population-oriented monitors within the city. The average of the 1-hr peaks and daily means of the city ozone monitors were also computed. There were an average of 13.8 TSP monitors reporting on a 1 in 6 day schedule. There was also an average of 11.8 SO2 monitors and 2.5 ozone monitors in normal operation in Detroit each day. Data on daily mean temperature and daily mean humidity were

206 JOEL SCHWARTZ

obtained from the weather station at the Detroit airport. Humidity-corrected visibility extinction coefficients (bext) w e r e calculated from the average of the visible range measurements taken at 10 AM, 1 PM, and 4 PM each day at the airport. On days with rain or fog, bex t w a s not computed.

Mortality counts are counts of rare events and usually modeled as arising from a Poisson process. Poisson regressions were used to model the daily counts. The analytic strategy was to first fit the best possible model using nonpollution variables and then test whether pollution made any significant additional contribution to explaining variations in mortality. This approach limits the interpretation one can put on the specific weather model that is identified, but more conservatively assesses the association with pollution.

The population of Detroit declined over the period 1973-1982, and the age- adjusted total mortality rate may have changed as well. These changes may induce time trends in the mortality data. In addition, random variations in, e.g., the intensity of influenza epidemics may induce year to year variations in mortality. To control for these, dummy variables for each year of the study were used in the Poisson regression models. In addition, a continuous time trend term was used.

The impact of weather on mortality can be nonlinear and may also vary with season. The initial analyses of weather dependence considered 24-hr mean temperature and dew point temperature, a dummy variable for hot days, a dummy variable for humid days, a dummy variable for cold days, and a dummy variable for hot and humid days. In addition, seasonal dummy variables were considered, as well as interactions between the seasons and the weather factors above. These interactions test for, e.g., a different relationship between temperature and mortality in summer than in winter. Weather variables on the concurrent day and on the previous day were considered. The definitions of hot, humid, and cold were initially set at the 95th percentile (5th percentile for cold days) and then varied to find the model with the greatest explanatory power. This limits the conclusions that can be drawn from the significance of the weather variables, but allows the weather and seasonal factors to explain as much of the variation of mortality as possible before testing the significance of air pollution. To make the air pollution test more conservative, weather terms were kept in the regression model as long as they were at least marginally significant (t > 1).

The Poisson regression model was estimated using the generalized estimating equations of Liang and Zeger (1986) to account for the possibility of overdispersion and serial correlation. Overdispersion refers to the increase in the variance of the distribution of mortality counts that can occur when the underlying population is not homogeneous with respect to risk of death. Serial correlation refers to the situation where mortality counts on 2 days close to each other in time are correlated. This can result from infectious disease epidemics, for example. As in a classic Poisson regression, the model assumes

Log[E(Yi)] = Xi[3,

where Xi is the matrix of covariates on day i, Yi the mortality counts on day i, and E denotes expected value. The covariance matrix is assumed to be of the form


0LA I/2RA 1/2,

where A~/ = E(Y~)~ij, the classic Poisson covariance, et is the overdispersion parameter, and R is an autoregressive matrix. The order of R is estimated empir- ically from the data. To be conservative, we included autoregressive terms if the estimated size of the autoregressive parameter exceeded its estimated standard error.

TSP was predicted for each day in the study from a linear regression model that was fit using data from the every sixth day sampling of TSP during the period. The relationship between TSP and bex t may vary seasonally and may also vary with temperature and humidity beyond the purely physical correction factors included in the standard temperature and humidity correction. Therefore separate intercept terms by season, temperature, dew point temperature, bext, and interactions between bex t and the other terms were considered in the regression model. Dummy variables for each year of study were also considered. Maximum R 2 stepwise regression was used to select the significant model with the greatest predictive power.

Once a model for predicting TSP concentrations was obtained, predicted TSP, SO2, and their values on the previous day were considered individually in the previously established Poisson regression model for daily mortality. If both pollutants were significant individually, a two-pollutant model was considered.

RESULTS

Table 1 shows the distribution of mortality counts, weather variables, and pollution concentrations in Detroit during the study period. The mortality distribution is skewed, as expected. The linear regression to predict TSP from bex t had a coefficient of multiple correlation of 0.78. The results of that model are shown in Table 2.

In Poisson regressions controlling for year and time trend, previous day's temperature, winter temperature, spring temperature, hot days, prior days being hot, and humid days were predictive of variations in daily mortality. The optimal definition of a hot day was a 24-hr mean temperature equal or exceeding 74°F, for a humid day it was a 24-hr mean dew point temperature equal or exceeding 63°F.

When air pollution was considered in the above model, the previous day's predicted TSP was highly significant as a predictor of mortality ([3 = 0.000546 - 0.000145, P < 0.001). Previous day's SO2 was also predictive of daily mortality ([3 = 0.863 -+ .323, P < 0.01), although less so than predicted TSP. In contrast,

TABLE 1 DISTRIBUTION OF MORTALITY, WEATHER, AND AIR POLLUTION IN DETROIT, 1973--1982

Variable 5% 10% 25% 50% 75% 90% 95% Mean

Deaths/day 39 41 46 52 58 64 68 53 Temperature (°F) 15 22 33 50 66 73 75 49 Dew point (°F) 9 14 25 39 55 63 66 39 Predicted TSP (p~g/m 3) 46 54 69 84 102 122 137 87 SO z (ppb) 3.7 4.5 6.4 10.1 15.5 21.7 26.3 12

208 JOEL S C H W A R T Z

T A B L E 2 REGRESSION PREDICTING DAILY TSP IN DETROIT, 1973-1982

Variable Coefficient Std error P-value

b~x t 26.4 5.3 <0.0001 Temperature 1.01 0.19 <0.0001 Winter 31.7 9.75 0.001 bext* Winter - 11.42 4.2 0.007 bext* S u m m e r - 3 . 1 1 1.23 0.01 bext* Dew point - 1.43 0.11 <0.0001 be~t* Tempera tu re 1.15 0.13 <0.0001

ozone was highly insignificant as a predictor of daily mortality. Both ozone and TSP concentrations were higher in the warm weather and lowest in the winter. In contrast, mortality peaked in the winter. This can bias the pollution results toward zero. This problem is most serious for ozone, because its winter concentrations are so low. Hence the ozone regression was repeated after excluding the winter months. Ozone remained highly insignificant as a predictor of daily mortality. Concurrent day's pollution was less significant (TSP) or insignificant (SO2) as a predictor of daily mortality, and the means of the two days' pollution concentrations were less significant for both pollutants than the previous day's pollution. When both previous day's SO2 and previous day's predicted TSP were considered in a model together, predicted TSP remained significantly associated with daily mortality ([3 = 0.000514 - 0.000161, P < 0.005) with no significant change in the magnitude of the slope. In contrast, SOz became highly insignificant (13 = 0.230 +- 0.489, P > 0.60), with about a two-thirds reduction in its effect size. Hence variations in predicted TSP that were independent of variations in SO2 remained associated with variations in daily mortality with a similar slope, while variations in SO2 that were independent of particulates were essentially uncorrelated with mortality. Table 3 shows the final Poisson regression model using prior day's predicted TSP.

To illustrate the dose-response nature of the relationship between particulates and daily mortality, the continuous particle measure was replaced in the Poisson

T A B L E 3 POISSON REGRESSION MODEL FOR DAILY MORTALITY IN DETROIT, 1973--1982

Variable Coefficient Std RR b (95% CI)

Hot 0.0465 .0153 1.05 (1.08-1.02) Hot" 0.0515 0.0135 1.05 (1.08--1.02) Humid 0.0113 0.0138 1.01 (1.04--0.98) Winter temperature 0.00020 0.0004 NC Temperature a - 0.0024 0.0003 NC Spring Tempera tu re 0.0003 0.0002 NC TSP a 0.000546 0.000145 1.06 (1.09-1.03)

Note. NC, not computed. Previous day's value.

b Relat ive risk for TSP is for a 100 i~g/m 3 increment.


regression model with dummy variables for quintiles of predicted TSP concentration. Figure 1 shows the relative risk of mortality versus predicted TSP concentration, with the risk in the lowest quintile taken as one. Excess mortality risk is clearly evident at half the ambient air quality standard (when expressed as TSP) and lower. To check the robustness of the TSP relationship several additional models were estimated. The first model included a quadratic time trend term in addition to the linear term and the second model excluded days with predicted TSP below 46 ixg/m 3 and days with predicted TSP concentrations above 137 ixg/m 3. This ensured that extreme values were not having excessive influence on the results, and that they clearly continued at concentrations below the ambient air quality standards. The third model added an interaction term between temperature and humidity. Predicted TSP remained highly significant in all three models, and the coefficient changed by less than -+ 10% from the original model shown in Table 3.

DISCUSSION

Relationship between TSP and bex t

As expected, the relationship between visibility and TSP differed in the summer, when oxidant hazes also serve to limit visibility. Despite the exclusion of days with rain or fog, the same visible range corresponded to a lower TSP level on humid days, but a higher TSP level on hot days. Some of the seasonal and

1.03

1.02

1.01

1.00

0.99

60 80 100 120 140

Predicted TSP

FIG. 1. The predicted relative risk of mortality versus the mean predicted TSP concentrations for quintiles of predicted TSP, after controlling for each year of study, continuous time trends, and the covariates in Table 3.

210 JOEL SCHWARTZ

weather-dependent variation in the relationship between TSP and extinction coefficient likely reflects the variation in the relationship between fine particles and TSP.

Mortality Models

While the analytic strategy limits the conclusions one can draw about the significance of the weather terms, the results are consistent with previous reports. Mortality was significantly elevated on hot days, as reported by Buechley et al. (1972), and in the Steubenville study (Schwartz and Dockery, 1991a). Another recent study in Philadelphia (Schwartz and Dockery, 1991b) also found hot days associated with increased mortality. The magnitude of the elevation (about 5%) was also similar to the results in Steubenville and Philadelphia. Other than that, increased temperature was protective overall, but less so in the winter.

The association between predicted TSP and daily mortality was highly significant (P < 0.0001) in regressions controlling for weather, time trends, year to year fluctuations, and serial correlation in the data. Moreover, the magnitude of the relationship was quite close to that observed in Steubenville and Philadelphia. The relative risk for a 100 txg/m 3 increase in TSP in Detroit was 1.06 (95% CI 1.09--1.03) versus 1.04 (95% CI 1.05-1.02) in Steubenville and 1.07 (95% CI 1.04-1.10) in Philadelphia. This consistency of effect size is striking and adds considerable weight to a conclusion of a causal relationship. Moreover, it is broadly consistent with the results from the episode study of Wichmann et al. (1989), where a 200- 250 t~g/m 3 increase in TSP concentration was associated with about an 8% increase in mortality. In London in 1952, the increase in mortality to 2.1 times that in the previous week accompanied an increase of about 1200 i~g/m 3 in particulates. Assuming the same relative risk model used here, in Steubenville and Philadel- phia, this translates to a coefficient of approximately 0.00062 per microgram com- pared to 0.000546 here. Hence essentially the same coefficient explains both the London episode, which few doubt was causal, and the results at much lower concentrations of particles, reported here. This strongly argues for a causal relationship. Figure 1 shows clear evidence of a dose-response relationship, which was also seen in Steubenville and Philadelphia. That also gives strength to the argument for causality. What is also striking in this figure is the low concentrations to which the relationship continues---concentrations well below half of the ambient air quality standard.

In contrast to the particulate relationship, the correlation with SO2 appeared to result only from its correlation with particulates, which is consistent with the results reported in London and Steubenville. It is also consistent with data from Schimmel and Murawski (1976), showing that the percentage of mortality in New York City that was attributable to air pollution in within-year analyses remained constant over a period of years when SO2 concentrations fell by more than two- thirds but particulate levels remained constant. These finding make it unlikely that SO2 is associated with mortality.

Further support for the causality of the observed relationship with particulates is found in studies linking daily particulate concentrations to increased hospital utilization. The 1952 London episode was accompanied by a significant increase in requests for bed admissions to the London Emergency Bed Service, for in-


stance (Her Majesty's Public Health Service, 1954). Wichmann and co-workers also reported increased ambulance calls and hospital admissions during the 1985 European episode. At lower concentrations, daily TSP was associated with emergency room visits in Steubenville (Samet et al., 1981). More recently, studies by Pope have reported associations between hospital admissions for respiratory conditions in the Utah Valley (1989) and in Salt Lake County (1991) in Utah and inhalable particulate (PM10) concentrations. These relationships also continued to concentrations well below the ambient standard. Respirable particulates were linked to hospital emergency room use in Israel (Gross et al., 1984). Bates and Sizto (1987) have also reported an association between sulfate particulates and hospital admissions for respiratory conditions in Southern Ontario.

Acute respiratory symptoms have also been associated with particulate exposure. Lawther et al. (1970) found increased symptoms associated with smog in COPD patients in a London diary study, and Ostro and Rothschild (1989) found increased days when respiratory symptoms severe enough to limit activity were associated with particulates in a broadly representative sample of adults. The latter results again appeared to occur at concentrations well below the ambient air quality standard for particulates. Particulates were also associated with increased episodes of coughing and lower respiratory symptoms in children in two diary studies where concentrations never exceeded the ambient standard (Schwartz et al., 1989; Braun-Fahrlander et al., 1991). Here too, the association was independent of SO 2, but not vice versa. In addition, Pope et al. (1991) reported that PM10 was associated with both peak flow decrements and increases in asthma symptoms in a diary study of asthmatic children in the Utah valley, where PM10 is virtually the only pollutant present. Schwartz et al. (1991) have also reported an association between particulate exposure and the incidence of croup symptoms in children in five cities in Germany. The pattern of association between particle exposure and increased respiratory symptoms, hospitalization, and mortality sug- gests a mechanism whereby respiratory irritation increases infectivity or pulmonary irritation. These may be direct particle effects. However, the role of particulate air pollution may be to alter the site and degree of deposition of other compounds increasing their biological effect. More detailed studies on particulate deposition and studies of the combination of particles and gases (or biologic agents) are clearly needed. Amdur and Chen's (1989) recent work showing that acids adsorbed on fine particles have a greater effect than acids in water aerosols are an example of such needed work. This manifests as increased illness and symptoms, as increased hospitalization for severe symptoms in subjects with chronic respiratory disease, and as increased mortality in the elderly, persons with chronic respiratory disease, and persons in fragile health. Given the pattern of findings across a range of increasing severity of outcome in increasingly sen- sitive populations, the multiple studies with similar finding and effect sizes, and the manifestation of a dose-response relationship, the relationships appear to be causal.

ACKNOWLEDGMENTS

Anne Nozniztky provided valuable computational support. Larry Kalkstein provided the Detroit weather data for this analysis.

212 JOEL SCHWARTZ

REFERENCES

Amdur, M. O., and Chen, L. C. (1989). Furnace generated acid aerosols: Speciation and pulmonary effects. Environ. Health. Perspect. 79, 147-150.

Bates, D. V., and Sizto, R. (1987). Hospital admissions and air pollutants in Southern Ontario: The Summer acid haze effect. Environ. Res. 43, 317-331.

Braun-Fahrlander, C., Ackermann-Liebrich, U., Schwartz, J., Grehm, H. P., Rutishauser, M., and Wanner, H. U. (1991). Air pollution and respiratory symptoms in pre-school children. Am. Rev. Respir. Dis. in press.

Brimblecombe, P. (1987). "The Big Smoke: A History of Air Pollution in London Since Medieval Times," p. 185. Methuen, New York.

Buechley, R. W., et al. (1972). Heat island--Death island? Environ. Res. 5, 85-92. Fairley, D. (1990). The relationship of daily mortality to suspended particulates in Santa Clara County,

1980-1986. Environ. Health Perspect. 89, 159-168. Firket, M. (1936). Fog along the Meuse Valley. Trans. Faraday Soc. 32, 1192-1197. Gross, J., Goldsmith, J. R., Zangwill, L., and Lerman, S. (1984). Monitoring of hospital emergency

room visits as a method for detecting health effects of environmental exposures. Sci. Total En- viron. 32, 289-302.

Her Majesty's Public Health Service (1954). "Mortality and Morbidity during the London fog of December 1952," Public Health and Medical Subjects Report No. 95. Her Majesty's Stationary Office, London.

Holland, W. W., Bennet, A. E., Cameron, I. R., Florey, Cd. V., Leeder, S. R., Shilling, R. S. F., Swann, A. V., and Waller, R. E. (1979). Health effects of particulate air pollution: Re-appraising the evidence. Am. J. Epidemiol. 110(5), 525-659.

Lawther, P. J., Waller, R. E., and Henderson, M. (1970). Air pollution exacerbations of bronchitis. Thorax 25, 525-529.

Liang, K. Y., and Zeger, S. L. (1986). Longitudinal data analysis using generalized linear models. Biometrika 73, 13-22.

Mazumdar, S., Schimmel, H., and Higgins, I. T. T. (1982). Relation of daily mortality to air pollution: An analysis of 14 London winters, 1958/59-1971/72. Arch. Environ. Health. 37, 213-220.

Ostro, B. D., and Rothschild, S. (1989). Air pollution and acute respiratory morbidity: An observa- tional study of multiple pollutants. Environ. Res. 50, 238-247.

Ozkaynak, H., Schatz, A. D., and Thurston, G. D. (1986). Relationships between aerosol extinction coefficients derived from airport visual range observations and alternative measures of airborne particle mass. J. Air. Pollut. Cont. Assoc. 35, 1176-1185.

Ozkaynak, H., Spengler, J. D., Garsd, A., and Thurston, D. G. (1986). Assessment of population health risks resulting from exposures to airborne particles. In "Aerosols" (S. D. Lee, T. Schnei- der, L. D. Grant, and P. J. Verkerk, Eds.), pp. 1067-1080. Lewis Publishers, Chelsea, MI.

Pope, C. A. (1989). Respiratory disease associated with community air pollution and a steel mill, Utah Valley. Am. J. Public Health 79, 623-628.

Pope, C. A. (1991). Respiratory hospital admissions associated with PM10 pollution in Utah, Salt Lake, and Cache valleys. Arch. Environ. Health 46, 90-97.

Pope, C. A., Dochery D. W., Spengler, J. D., Raizenne, M. E. (1991). Respiratory health and PM~o Pollution: a daily time series analysis. Am. Rev. Respir. Dis. 144, 668-674.

Samet, J. M., Speizer, F. E., Bishop, Y., Spengler, J. D., and Ferris, B. G., Jr. (1981). The relationship between air pollution and emergency room visits in an industrial community. J. Air Pollut. Control Assoc. 31, 236-240.

Schimmel, H., and Murawski, T. J. (1976). The relation of air pollution to mortality. J. Occup. Med. 18, 316.

Schwartz, J., Dockery, D. W., Ware, J. H., Spengler, J. D., Wypij, D., Koutrakis, P., Speizer, F. E., and Ferris, B. G., Jr. (1989). "Acute Effects of Acid Aerosols on Respiratory Symptom Reporting in Children." Air Pollution Control Association Paper 89-92.1.

Schwartz, J., and Dockery, D. W. (1990). Particulate air pollution and daily mortality in Steubenville, Ohio. Am. Rev. Respir. Dis., 141, A71.


Schwartz, J., and Marcus, A. (1990). Mortality and air pollution in London: A time series analysis. Am. J. Epidemiol. 131, 185-194.

Schwartz, J., and Dockery, D. W. (1991a). Particulate air pollution and daily mortality in Steubenville, Ohio. Am. J. Epidemiol., in press.

Schwartz, J., and Dockery, D. W. (1991b). Increased mortality in Philadelphia associated with daily air pollution concentrations. Am. Rev. Respir. Dis., in press.

Schwartz, J., Spix, C., Wichmann, H. E., and Malin, E. (1991). Air pollution and Acute Respiratory illness in fire german communities. Environ. Res., in press.

US EPA. (1982). "Air Quality Criteria for Particulate Matter and Sulfur Oxides." Office of Research and Development, Environmental Criteria and Assessment Office, Research Triangle Park, NC.

Wichmann, H. E., Mueller, W., Allhoff, P., Beckmann, M., Bocter, N., Csicsaky, M. J., Jung, M., Molik, B., and Schoenberg, G. (1989). Health effects during a smog episode in West Germany in 1985. Environ. Health Perspect. 79, 8%99.

Wyzga, R. (1978). The effect of air pollution upon mortality: A consideration of distributed lag models. J. Am. Stat. Assoc. 73, 463-472.

particulate air pollution and daily mortality in detroit

Documents