Table 1 presents the injury deaths per million inhabitants per year by external cause (transport vs. other) and in total for each city. In many instances, Nanjing’s rate is less than half that of 8 of the 9 other cities. The three year upward trend in total injury death rates in several cities is mainly due to unintentional injuries other than those experienced in transport.
Table 1. Annual Transport and Other Fatal Unintentional Injuries Per Million Inhabitants Among Cities in Jiangsu Province, China, 2015-2017
Transport Other Injury Total
2015 2016 2017 2015 2016 2017 2015 2016 2017
Nanjing 81 79 80 120 147 138 202 226 218
Wuxi 115 115 112 273 315 314 387 430 426
Xuzhou 228 224 221 183 197 231 411 421 452
Changzhou 150 168 170 246 295 301 395 463 471
Nantong 187 214 232 274 365 380 461 579 612
Lianyungang 212 212 227 295 320 338 507 532 565
Huaian 210 196 189 218 241 244 428 437 433
Yancheng 206 218 261 199 289 282 404 508 543
Suqian 162 179 183 137 151 180 299 330 364
Totals may not add exactly because of rounding.
Table 2. Means and Standard Deviations of Transformed Predictor Variables
|
Temperature (℃)
|
Precipitation (mm)
|
Ozone (µg/m3)
|
|
Mean (SD)
|
Mean (SD)
|
Mean (SD)
|
Nanjing
|
21.2 (9.1)
|
1.3 (2.7)
|
4.6 (0.4)
|
Wuxi
|
21.5 (9.1)
|
0.4 (2.6)
|
4.6 (0.5)
|
Xuzhou
|
20.5 (9.6)
|
0.9 (1.9)
|
4.5 (0.4)
|
Changzhou
|
21.4 (9.2)
|
1.5 (2.6)
|
4.5 (0.5)
|
Nantong
|
20.9 (8.9)
|
1.4 (2.6)
|
4.5 (0.5)
|
Lianyungang
|
19.6 (9.4)
|
0.8 (1.9)
|
4.6 (0.4)
|
Huaian
|
20.3 (9.2)
|
1.1 (2.2)
|
4.6 (0.4)
|
Yancheng
|
20.0 (9.1)
|
1.2 (2.4)
|
4.6 (0.4)
|
Suqian
|
20.4 (9.3)
|
0.9 (2.1)
|
4.6 (0.3)
|
Table 2 gives the means and standard deviations of the weather and pollution variables by city. Nanjing is not consistently higher or lower on the predictor variables. These findings led to the decision to include Nanjing as a separate variable in the analysis.
Figure 1 shows the transport related deaths per billion person days of exposure to specific maximum temperatures. Each data point is the sum of the deaths on the days that the maximum temperature reached the indicated degrees divided by the sum of the approximate number of people residing in a given city each day the temperature reached the indicated degrees adjusted to billions of person days. A maximum daily temperature at freezing or below in Jiangsu Province is rare. Person days when the maximum temperature was zero (C) or less were two-tenths of one percent of the total so deaths on those days and the person years are included in the zero category in the graphs. There does not appear to an association of transport death risk with maximum temperatures other than a dip at the highest temperatures. In contrast, Figure 2 indicates that unintentional injuries other than in transport declined slightly up to about 25 degrees maximum daily temperature, accelerated upward at higher temperatures and did so extraordinarily at temperatures above 34 degrees.
These findings led to the decision to analyze the data separately for cool (<25 degrees), moderate (25-34 degrees) and hot (35 plus degrees) temperatures.
The Poisson regression coefficients are presented in Table 3 along with 95 percent confidence intervals and criteria for goodness of fit. Both transport and other fatal injuries are consistently lower for residents of Nanjing. Corrected for that and the estimated effects of other factors, transport deaths were related to increasing deaths when temperatures were low but the correlation reversed at higher temperatures. Non-transport injury deaths increased substantially in relation to higher temperatures when temperatures were in the moderate range and even more so at temperatures 35 degrees (C) and higher. Deaths were lower on rainy days when temperatures were cool and moderate with the exception of non-transport injuries when temperatures were moderate. Higher ozone concentrations were associated with more deaths except when temperatures were low. Transport fatalities were consistently lower on weekends and on holidays at high and low temperatures. Other injury deaths were higher on holidays when temperatures were low to moderate.
Table 3. Poisson Regression Coefficients beta and 95 Percent Confidence Intervals (CI): Unintentional Injury Fatalities Per Day
Transport Deaths
Temperature Cool (< 25 °C) Moderate (25-34 °C) Hot (> 34 °C)
Beta 95% CI beta 95% CI beta 95% CI
Temperature .0077 (.0050, .0104) -.0074 (-.0145,-.0003) -.0796 (-.1202, -.0334)
√Precipitation -.0189 (-.0260, -.0118) -.0129 (-0.0209, -.0049) -.0031 (-.0365, .0303)
Log(ozone) -.1263 (-.1681, -.0845) 0.0563 (.0000,.1126) .1995 (.0061, .3929)
Weekend -.0531 (-.0862, -.0120) -.0534 (-.0959,-.0109) -.1793 (-.3019, -.0567)
Holiday -.0863 (-.1441, -.0285) .0765 (.0094, .1435) -.3979 (-.7640, -.0318)
Nanjing -.8241 (-.8827,-.7655) -.9118 (.9898, -.8338) -.7803 (-.9695, -.5911)
Intercept -13.9409 -14.4360 -12.6328
Deviance/df 1.35 1.37 1.39
Other Unintentional Injury Deaths
Temperature Cool (< 25 °C) Moderate (25-34 °C) Hot (> 34 °C)
beta 95% CI beta 95% CI beta 95% CI
Temperature -.0045 (-.0068, -.0021) .0314 (,0251, .0377) .2277 (.2037, .2517)
√Precipitation -.0170 (-.0233, -.0107) .0095 (.0026, .0164) .0101 (-.0313, .0111)
Log(ozone) -.1586 (-.1945, -.1227) .0248 (-.0248, .0744) .1788 (.0536, .3040)
Weekend -.0051 (-.0337, .0235) -.0146 (-.0524, .0232) -.0325 (-.1034, .0384)
Holiday .0269 (-.0211, -.0749) .0208 (-.0415, .0831) -.2220 (-.4780, .0380)
Nanjing -.6288 (-.6758, -.5817) -.6376 (-.6996, -.5756) -.6667 (-.7728, -.5606)
Intercept -13.3868 -15.2458 -22.6963
Deviance/df 1.35 1.40 3.18
The criteria for goodness of fit are near 1 indicating good fit of the models.
Of concern in use of regression coefficients is the degree to which the predictor variables are correlated. The squared OLS correlation coefficients among the predictor variables are displayed in Table 4. None of the correlations are high enough to be of concern. O3 is uncorrelated with precipitation and is only moderately correlated with temperature at cool temperatures and less so at moderate and high temperatures in this study. The correlation of these variables with weekends, Nanjing and population was less than 0.01. Nanjing’s larger population resulted in an R2 of 0.22 with population.
Table 4. Squared OLS Correlation Coefficients Among Transformed Daily Weather and Pollution Predictor Variables At Cool, Moderate and Hot Temperatures, Jiangsu Province, China. 2015-2017
|
Cool
|
|
Moderate
|
|
Hot
|
|
|
Precipitation
|
Ozone
|
Precipitation
|
Ozone
|
Precipitation
|
Ozone
|
Temperature
|
0.04
|
0.21
|
<0.01
|
0.01
|
<0.01
|
0.04
|
Precipitation
|
|
<0.0
|
|
0.11
|
|
0.02
|