A Cohort Study of Traffic-related Air Pollution and Mortality in Toronto, Canada: Online Appendix Michael Jerrett, 1 Murray M. Finkelstein, 2 Jeff R. Brook, 3 M. Altaf Arain, 4 Palvos Kanaroglou, 4 Dave M. Stieb, 5 Nicholas L. Gilbert, 5 Dave Verma, 6 Norm Finkelstein, 4 Kenneth R. Chapman, 7 Malcolm R. Sears 8 1 Division of Environmental Health Sciences, School of Public Health, University of California, Berkeley, Berkeley, California, United States 2 Department of Family and Community Medicine, Mount Sinai Hospital,University of Toronto, Toronto, Ontario, Canada 3 Meteorological Services, Environment Canada 4 School of Geography and Earth Science, McMaster University, Hamilton, Ontario, Canada 5 Air Health Effects Section, Health Canada 6 Occupational Medicine Program, McMaster University, Hamilton, Ontario, Canada 7 Division of Respirology, Department of Medicine, University of Toronto, Toronto, Ontario, Canada 8 Division of Respirology, Department of Medicine, McMaster University, Hamilton, Ontario, Canada Corresponding Author: Michael Jerrett University of California, Berkeley School of Public Health Division of Environmental Health Sciences 50 University Hall (Mailing Address) 7 University Hall (Office and GIS Lab) Berkeley, CA 947-7360 Email: jerrett@berkeley.edu Tel: 5-725-9991 Fax: 5-642-5815 1
Purpose This appendix gives additional details on the measurement and modeling used to estimate nitrogen dioxide exposures for the epidemiological analyses presented in the main paper. As noted in the main body of the paper, we have already published model specifications for the 02 and 04 land use regressions, and complete details can be found elsewhere (Jerrett et al. 07; Finkelstein and Jerrett 07). This online appendix summarizes both models to assist with interpreting the epidemiologic results in the main paper. Descriptive statistics and box plots for data collected in 02 and 04 are listed in Supplemental Materials Table 1 and Figure 1 below. The data shows that there was a little more than two times difference between the years. Supplemental Material Table 1: Descriptive statistics of all NO 2 measurements in parts per billion (02, 04) NO2 (02) NO2 (04) N Range Minimum Maximum Mean Std. Deviation 94 43.67 17.41 61.08 32.2365 9.18813 0.73 7.03 27.76 14.5433 4.48639 In Supplemental Material, Figure 1, we see that with 46 co-located measurements there was strong correlation with R = 0.82 even with the large difference in the magnitude of NO 2 measurements. 2
60 50 41 18 16 40 14 30 12 8 0 N = 46 46 NO2 (04) 6 4 Rsq = 0.6792 30 40 50 60 NO2 (02) NO2 (04) NO2 (02) Supplemental Material Figure 1: Boxplot and scatterplot of co-located monitoring values Supplemental Materials Table 2 and Figure 2 below show the model specifications and the observed on predicted value scatterplots for the Fall 02 model. Supplemental Material Table 2: Summary of the regression results for the logarithmic NO 2 Model (02). Number of obs = 94 Source SS df MS F(7,87) = 27.4 Regression 5.09 7 0.727 Prob > F = 0 Residual 2.29 86 0.027 R-square = 0.69 Total 7.38 93 Adj. R-square = 0.67 Root MSE = 0.163 Variable* Coefficient Std. Error t Prob > t VIF LN(NO 2 ) (Constant) 8.06E+00 1.177 6.85 0.00 RD1_0 1.84E-01 0.0 9.04 0.00 1. RD2_50 5.56E-01 0.300 1.85 0.07 1.11 IND750 1.63E-03 0.001 3.04 0.00 1. DC00 8.28E-05 0.000 4.66 0.00 1.38 X -8.01E-06 0.000-4.31 0.00 1.06 D_WIND1500 1.32E-01 0.040 3.30 0.00 1. TRAF500 1.11E-03 0.001 1.96 0.05 1.32 *RD1_0 measure of expressway within 0m; RD2_50 measure of major roads within 50m; IND750 measure of industrial land use within 750m; DC00 density of dwellings within 00m (Kernel estimate); X UTM NAD83 x-coordinate; D_WIND15 Boolean identifier whether downwind and within 1500m of nearest expressway at PMpeak traffic; TRAF500 Density measure of 24 hour traffic counts within 500m. 3
4.2 4.0 3.8 3.6 3.4 Observed values 3.2 3.0 2.8 2.8 3.0 3.2 3.4 3.6 3.8 4.0 Unstandardized Predicted Value Supplemental Material Figure 2: Logarithmic-observed mean NO 2 (02) on predicted value. We present similar results for the 04 LUR model in Table 3, with the observed on predicted scatterplot shown in Figure 3. Supplemental Material Table 3: Summary of the regression results for NO 2 Model (04). Number of obs = 1 Source SS df MS F(5,96) = 46.1 Regression 1411.45 5 282.291 Prob > F = 0.000 Residual 581.86 95 6.125 R-square = 0.7080943 Total 1993.31 0 Adj. R-square = 0.69 Root MSE = 2.475 Variable* Coefficient Std. Error t Prob > t VIF NO 2 [ppb] (Constant) 9.62E+00 0.690 13.93 0.00 X_MEAN -2.68E-04 0.000-9.36 0.00 1.02 EXCS400 9.83E-01 0.123 8.00 0.00 1.12 RD2_00 2.E-01 0.087 2.52 0.01 1.62 RD2_50 1.E+01 5.6 2.15 0.03 1.28 EA2500 5.02E-04 0.000 3.60 0.00 1.52 * X_MEAN mean deviated UTM NAD83 x-coordinate; EXCS400 measure of area of road expressway encasements within 400m; RD2_00 measure of major roads within 00m; RD2_50 measure of major roads within 50m; EA2500 measure of population density using kernel density with a bandwidth of 2500m of population values at enumeration area centroids. 4
30 Observed Values 0 Rsq = 0.7081 0 30 Unstandardized Predicted Value Supplemental Material Figure 3: Observed NO 2 (04) on predicted values. Even though the models fit different variables, when comparing 02 to 04 they display similar spatial gradients (see Figure 4). We averaged the two surfaces after assignment to the study subjects in the cohort and used the average estimate for the epidemiological analyses. 5
Figure 4: Prediction Maps of 02 and 04 NO2 land use regression models 6
References Finkelstein MM, Jerrett M. 07. A study of the relationships between Parkinson's disease and markers of traffic-derived and environmental manganese air pollution in two Canadian cities. Environ Res 4(3):4-432. Jerrett M, Arain MA, Kanaroglou P, Beckerman B, Crouse D, Gilbert NL, et al. 07. Modelling the intraurban variability of ambient traffic pollution in Toronto, Canada. J Toxicol Environ Health A 70(3-4):0-212. 7