Short-term effects of different PM2.5 ranges on daily all-cause mortality in Jinan, China

To examine the effects of different PM2.5 concentration ranges on daily all-cause mortality, 8768 all-cause deaths were recorded in the database of the Shandong Provincial Hospital Affiliated to Shandong First Medical University. Data of air pollutants (PM2.5 and O3) concentration were provided by the Jinan Environment Monitoring Center. The relative risk of all-cause mortality was assessed using a quasi-Poisson regression model after adjusting for confounding factors. The concentrations of PM2.5 were divided into four ranges 0–35 μg/m3; 35–75 μg/m3; 75–115 μg/m3; 115–150 μg/m3. There was no significant relationship between PM2.5 exposure and all-cause deaths in individuals aged < 60 years. However, for individuals aged ≥ 60 years, there was a significant positive association between exposure concentrations and all-cause deaths within the ranges 0–35 μg/m3, 35–75 μg/m3, and 115–150 μg/m3 with a mortality increase of 1.07 (1.01, 1.13), 1.03 (1.00, 1.05), and 1.05 (1.01, 1.08), respectively. When the population aged ≥ 60 years was stratified into gender groups, exposure to PM2.5 in the range 0–35 μg/m3 increased the mortality risk in men but not women. All-cause mortality in women, but not men, increased significantly with exposure to PM2.5 in the ranges of 35–75, 75–115, and 115–150 μg/m3.

To establish the four PM2.5 concentration ranges, we set all concentrations outside the range as "NA".
A quasi-Poisson regression model with natural splines was used to assess the impact of different PM 2.5 ranges on daily all-cause mortality because the daily death counts in Jinan approximately followed a Poisson distribution. This regression model is used to adjust inference for overdispersion 25 . The natural cubic spline for mean temperatures with 5 degrees of freedom and relative air humidity with 3 degrees of freedom (df) was controlled to analyze all-cause mortality based on Akaike's Information Criterion (AIC) for lag effects of up to 3 days 26 . Confounding factors such as day of the week and holidays were included as dummy variables.
The natural cubic spline smoothing function degree of freedom for mean temperature and relative air humidity is determined as follows: Yt represents the death counts on day t. E(Yt) represents the expected death counts on day t, ns stands for the natural cubic spline smoothing function, Temp represents the mean temperature, RH represents the relative air humidity, DOW and Holiday stands for the day of the week effect and legal holidays respectively, β 1 and β 2 are the coefficient of DOW and Holiday respectively. The degree of freedom of the mean temperature factor is N (N = 2,3,…,6). Obtain the magnitude of the corresponding AIC of the equation when N is different, and the minimum value of AIC is the optimal degree of freedom. Different PM 2.5 ranges were added into the above basic model to establish a single-pollutant model. Multipollutant models with PM 2.5 and O3, with multi-day moving average lag structures [from a lag of 0 to 1 day (mean) to a lag of 0 to 3 days (mean)], were used for sensitivity analysis to determine the stability of the model.
The relative risk (RR) and corresponding 95% confidence interval (CI) for an increase of 10 μg/m 3 in pollutant concentration were estimated to assess the impact of different PM 2.5 ranges on daily counts of all-cause mortality. P-values smaller than 0.05 were considered statistically significant.
Stratified analyses of exposure to different PM 2.5 ranges based on gender (male or female) and age (< 60 years and ≥ 60 years) were performed to find associations with daily all-cause mortality. www.nature.com/scientificreports/ tively. The distributions of the daily concentrations of air pollutants, weather parameters, and deaths are shown in Table 1. According to MEP, air quality was good (green category) in 4.11% of the days, moderate (yellow category) in 38.36% of the days, poor for sensitive groups (orange category) in 32.24% of the days, poor (red category) in 12.42% of the days, and very poor (purple category) in 12.88% of the days for all populations. PM 25 concentration and air quality index values in the study period are shown in Table 2.
Stratified analysis based on gender and age indicated that there was a significant relationship between allcause mortality and a PM 2.5 range of 0-35 μg/m 3 in men in lags 1 and 01. All-cause deaths in women significantly   The results of sensitivity analysis indicated that the relative risk at different PM 2.5 ranges for daily all-cause mortality generally decreased slightly after including O3 in the multi-day moving average lag structures (Fig. 3). This may be related to the strong collinearity between different Pollutants. Some studies also believe that dualpollutant models will increase the standard deviation of model fitting, so the significance of statistical analysis is weak 27 .

Discussion
To our knowledge, this epidemiologic study is the first to examine the association of PM 2.5 concentration ranges with all-cause mortality in Asia. The results indicated that, except for the PM 2.5 range of 75-115 μg/m 3 , the concentrations of PM 2.5 in the ranges 0-35 μg/m 3 , 35-75 μg/m 3 , and 115-150 μg/m 3 were significantly associated with mortality from all causes for individuals aged ≥ 60 years. At the optimum lag structure, the statistically significant relative risks RR (95% CI) for daily all-cause mortality from exposure in the ranges 0-35 μg/m 3 , 35-75 μg/m 3 , and 115-150 μg/m 3 of PM 2.5 were 1.10 (1.02, 1.18), 1.04 (1.01, 1.07), and 1.06 (1.02, 1.11) respectively. This indicated that the adverse impacts on public health do not decrease as pollutant levels decrease. The statistically significant relative risks RR (95% CI) for daily all-cause mortality from exposure in the range 0-35 μg/m 3 of PM 2.5 were 1.10 (1.02, 1.18). This means that in the concentration range of 0 to 35 μg/m 3 , the effects of PM 2.5 were stronger as the concentration increased. In the same way, in the concentration ranges of 35 to 75 μg/m 3 and 115 to 150 μg/m 3 , we can get the same conclusion. . In addition, as shown in Table 3, in the PM 2.5 concentration range of 0 to 35 μg/m 3 , RR (95% CI) for daily all-cause mortality was 1.10 (1.02, 1.18), and Fig. 2 indicated that there was a significant relationship between all-cause mortality and a PM 2.5 range 0-35 μg/m 3 in men and individuals aged ≥ 60 years in lags 1 and 01, both of these indicated that consistent with other studies [28][29][30] , even for concentrations lower than 35 μg/m 3 , PM 2.5 is a significant risk factor for all-cause mortality.
For individuals aged ≥ 60 years, the association between all-cause deaths and PM 2.5 exposures was statistically significant at ranges of 0-35 μg/m 3 , 35-75 μg/m 3 , and 115-150 μg/m 3 . The lack of significance in the 75-115 μg/ m 3 range may be because of the relatively fewer deaths in this range. Furthermore, the daily temperatures corresponding to concentrations of 75-115 μg/m 3 were higher than those at 0-35, 35-75, and 115-150 μg/m 3 . The impact of different PM 2.5 ranges on mortality may be due to differences temperatures 31,32 .
The results of a previous study on the gender-specific effects of particulate matter were inconsistent 33 . The results of the gender-stratified analysis demonstrated that female subjects were more sensitive to the PM 2.5 in the ranges of 35-75 μg/m 3 , 75-115 μg/m 3 , and 115-150 μg/m 3 , whereas male subjects were more sensitive to PM 2.5 in the range of 0-35 μg/m 3 , indicating that men are more susceptible to lower PM 2.5 concentrations than women.
Smoking is a critical environmental risk factor, and one study suggested that the estimated impact of air pollution might be stronger in nonsmokers than smokers 34 . A potential reason for this difference may be that women have  www.nature.com/scientificreports/ www.nature.com/scientificreports/ slightly stronger airway reactivity and smaller airways than men 35 . Moreover, the adverse impacts of additional exposure to PM 2.5 may be overcome by the oxidative and inflammatory effects of smoking 36 .
Older individuals had increased susceptibility to PM 2.5 ranges of 35-75 μg/m 3 , 75-115 μg/m 3 , and 115-150 μg/ m 3 compared with younger individuals, possibly because the former group has a weaker immune system and higher sensitivity to these particles 37,38 . However, there was no significant association between PM 2.5 exposure and all-cause mortality in individuals aged < 60 years, indicating that the general population should avoid high levels of PM 2.5 (≥ 75 μg/m 3 ).
This study has some limitations. First, the study selected the mean air pollutant concentration from each monitoring site in Jinan as the exposure concentration; nonetheless, individual exposure may depend on other factors, including the type of outdoor activity, physical fitness, and living habits, potentially causing exposure measurement errors or underestimating the impact of air pollution. In addition, this study belongs to the field of ecological research, and the conclusions cannot prove causality but merely indicate the relationship between air pollutants and all-cause mortality.