Vital Surveillances: A Time-Series Analysis on the Association Between Fine Particulate Matter and Daily Mortality — Shijiazhuang City, Hebei Province, China, 2015–2020

The study obtained daily mortality data from January 1, 2015, to December 31, 2020 from Shijiazhuang CDC’s Cause of Death Reporting System. Causes of death were classified according to the International Classification of Diseases, 10th revision (ICD-10) (6), including total non-accidental causes (“ALL”, codes A00-R99), cardiovascular disease (“CVD”, codes I00-I99), and respiratory diseases (“RESP”, codes J00-J99). We categorized non-accidental deaths into strata by sex, age group (5–64 years and 65 years or older), and education level (low: less than or equal to 9 years of education; high: more than 9 years of education). Deaths of children five years and under were too few to analyze and were excluded. Meteorological factors (daily average temperature and relative humidity) and air pollution data [daily 24-hour average concentration of PM_2.5, particulate matter with particle size below 10 microns (PM₁₀), SO₂, and NO₂, and maximum eight-hour mean concentration of O₃] were obtained from Shijiazhuang Meteorological Bureau and Shijiazhuang Ecology and Environment Bureau, respectively.

We examined associations between PM_2.5and cause-specific mortality using generalized additive models (GAM) (7) with a quasi-Poisson link function to account for over-dispersion of daily cause-specific deaths. We controlled for seasonal patterns, long-term trends, temperature, and relative humidity using natural cubic regression smoothing. Our analyses allowed 7 degrees of freedom (df) per year for time long-term trends, 6dffor daily mean temperature, and 3dffor daily mean relative humidity, to minimize the Akaike’s Information Criterion (AIC) value of GAM. We stratified analyses by year (2015–2017 and 2018–2020).

The description of the model, methods and results for analyses of different periods (2015–2017 and 2018–2020) as shown inSupplementary Table S1, Spearman’s correlation coefficients as shown inSupplementary Table S2, sensitivity analyses as shown inSupplementary Table S3, two-pollutant models as shown inSupplementary Table S4, and stratification analyses were presented in the Supplementary materials. Analyses were conducted using the packages “mgcv” in R statistical software (version 3.5.1; The R Foundation for Statistical Computing, Vienna, Austria). We used two-tailed tests;Pvalues less than 0.05 were considered statistically significant.

Table 1showed mortality, PM_2.5,and meteorological data and daily average counts of non-accidental (ALL), cardiovascular (CVD), and respiratory (RESP) deaths. During the study period, there were averages of 35 ALL, 19 CVD, and 5 RESP deaths per day. Among the 76,859 ALL deaths, there were 41,473 (54.0%) CVD deaths and 9,955 (13.0%) RESP deaths. Daily concentration of PM_2.5ranged from 6.3 μg/m³to 625.3 μg/m³, and the annual-mean concentration was 77.6 μg/m³. There were 767 days (35% of the study period days) in which PM_2.5concentration was over 75 μg/m³, the national second ambient air quality standard in China. As shown inSupplementary Table S1, PM_2.5pollution was lower during 2018 to 2020 compared with 2015 to 2017, as the PM_2.5average concentration decreased from 91.1 μg/m³to 64.1 μg/m³, while the maximum concentration decreased from 625.3 μg/m³to 355.0 μg/m³and the number of days exceeding the national standard decreased from 480 to 287 days.

As shown inFigure 1, the delayed effects of PM_2.5on ALL mortality were statistically significant for lag1, lag2, lag01, lag02, lag03, and lag04; the largest delayed effects estimates were for lag01, in which a 10 μg/m³increase in PM_2.5was associated with an increment in ALL deaths of 0.47% (95% CI: 0.24%, 0.70%). For CVD mortality, the delayed effects of PM_2.5were statistically significantly for lag1, lag2, lag01, lag02, lag03, lag04, and lag05; the largest delayed effects estimates were also for lag01, with a 10 μg/m³increase in PM_2.5corresponding to a 0.49% (95% CI: 0.19%, 0.79%) increment in death. For RESP mortality, in single-day lag models, significant associations were limited to the first day after PM_2.5exposure, with a 10 μg/m³increase in PM_2.5corresponding to a 0.78% (95% CI: 0.33%, 1.23%) increment in death. When PM_2.5exposures were lagged over multiple days, the associations were strongest for exposures during lag01 (Estimates: 0.72%, 95% CI: 0.22%, 1.23%).

Figure 1.

Percent changes (95% CI) in daily cause-specific mortality per 10 μg/m³increase in PM_2.5concentrations using different lag days; (A) ALL mortality, (B) CVD mortality, (C) RESP mortality.

Abbreviations: ALL=total non-accidental death; CVD=cardiovascular disease; RESP=respiratory disease; PM _2.5=particulate matter with an aerodynamic diameter less than or equal to 2.5 μm.

ForFigure 2Aand 2B, the exposure-response curves for ALL and CVD showed increasing trends. When PM_2.5concentrations were lower than 120 μg/m³or higher than 300 μg/m³, slopes of curves showed marked increases. When PM_2.5concentrations were between 120 μg/m³and 300 μg/m³, slope was flat. InFigure 2C, the exposure-response curve for respiratory mortality was nearly linear and positive. When PM_2.5concentration was over 300 μg/m³, confidence intervals were wider than when PM_2.5concentration was less than 300 μg/m³.

Figure 2.

Exposure-response curves for associations of daily PM_2.5concentrations (lag 01) with (A) ALL mortality, (B) CVD mortality, and (C) RESP mortality.

Notes: The y-axis can be interpreted as the relative change from the mean effect of PM _2.5on mortality. The solid lines represent mean estimates, and the shaded areas represent 95% confidence intervals. Abbreviations: ALL=total non-accidental death; CVD=cardiovascular disease; RESP=respiratory disease; PM _2.5=particulate matter with an aerodynamic diameter less than or equal to 2.5 μm.

As shown inTable 2, compared with 2015–2017, during 2018–2020, the effect of PM_2.5on ALL mortality was larger, and the estimated effect value changed from 0.50% to 0.63%, but the difference was not statistically significant. The effect on CVD mortality was slightly less and not statistically significantly different. The effect of PM_2.5on RESP mortality was significantly less and was statistically significantly different. The association between PM_2.5and total mortality varied by demographic characteristics. Throughout the 2015–2020 study period, an increase in PM_2.5of 10 μg/m³corresponded to a 0.53% increment in deaths of males and a 0.39% increment in deaths of females. The 5–64-year-old group and ≥65-year-old-group had similar mortality associations. The association between PM_2.5and total mortality was a 0.51% increment for people with lower educational achievement and a 0.37% increment for those with higher educational achievement. There were no statistically significant differences among sex, age, and education in stratified analyses.

During the study period, PM_2.5pollution in Shijiazhuang improved compared with previous years (8), but pollution was still serious. Our general additive model results showed that ALL mortality, CVD mortality, and RESP mortality were related to PM_2.5concentration. Sensitivity analysis and 2 pollutant models indicated that PM_2.5had an independent health effect on ALL mortality, CVD mortality, and RESP mortality.

Whether improvement of PM_2.5can reduce mortality is still controversial. During the study periods of 2015–2017 and 2018–2020, after the concentration of PM_2.5was reduced, the most direct manifestation of the reduction was a significant decrease in RESP mortality. The respiratory system is a target organ of PM_2.5direct action, and PM_2.5concentration is directly related to the occurrence and development of respiratory diseases. With improvement of people’s health awareness, residents wear masks and use air purifiers on polluted days. No significant change was observed for ALL mortality and CVD mortality. Due to its small particle size, PM_2.5can enter the blood circulation through the gas-blood barrier, thus affecting the circulation system, in which the composition of PM_2.5plays a major role. PM_2.5components should be further analyzed in the future to find harmful components and take targeted control measures.

ALL mortality in our study consisted mainly of circulatory system diseases, tumors, respiratory system diseases, endocrine and metabolic diseases, and digestive system diseases. In addition to circulatory system diseases and respiratory system diseases, the connection between other system diseases and PM_2.5needs to be explored in future studies.

The effect values we observed differed slightly from those observed in previous studies. There is significant spatial heterogeneity between PM_2.5concentration and daily mortality in different countries and regions (2,9). Our study found that every 10 μg/m³increase in PM_2.5concentration was associated with a 0.47% increase in ALL mortality — a value that was higher than results of a recent study of 272 cities in China (0.22%) (3). Many factors may be responsible for the difference, including different PM_2.5components, long-term air pollution levels, and population susceptibility. The larger effect estimates observed in our study may also be due to higher PM_2.5concentrations.

The shape of the exposure-response curves was linear without discernible thresholds, which was consistent with findings from previous studies (10). As shown in our study and a previous multisite study of 272 representative cities in China, there was a marked increase in E-R values at lower PM_2.5levels and a leveling off at relatively high concentrations (3). The stability observed at higher concentrations may be the result of a “harvesting effect,” since susceptible populations may die before air pollutant concentrations reach very high levels (11). When PM_2.5concentrations were above 300 μg/m³, slopes of the curves markedly increased. Therefore, reducing outdoor activity on heavily polluted days may reduce risk of death. Associations of PM_2.5on RESP mortality for PM_2.5concentrations over 300 μg/m³were characterized by wider confidence intervals, implying that the mortality risk has greater uncertainty.

When stratified by sex, age, and educational attainment, we found higher association in males, 5–64-year-olds, and individuals with lower educational achievement, but differences between sex, age group, and education were not statistically significant in stratified analyses. These findings may be due to occupational factors, as young men with low education levels may engage in more outdoor work, resulting in exposure to higher concentrations of air pollution. Lin et al. (12) found that older people may be more susceptible to PM_2.5in 6 cities of the Pearl River Delta region. Lee et al. (13) found that the most vulnerable population in three southeastern states was people with low educational achievement. A study in Shenzhen showed a high effect of PM_2.5on males and the elderly (14), whereas an analysis from 160 communities of China showed females, older individuals, and widows appeared to be more vulnerable to PM_2.5(9). There were differences in lifestyle, physiological factors, immunity, housing, and medical conditions by sex, age, and education, all of which can lead to different research results. Identification of potentially susceptible populations is crucial to public health and to the development of targeted intervention strategies.

This study was subject to some limitations. First, we used pollutant concentration data from urban environmental monitoring stations instead of population exposure concentrations. People spend much of their time indoors, and there is a significant difference between indoor and outdoor pollution (15). Therefore, there will be differences between our results and the real effects of PM_2.5. Second, the primary causes of death were categorized into circulatory and respiratory diseases. Categorization should be further refined to screen out sensitive diseases. Finally, our research only analyzed Shijiazhuang, which has relatively high pollution levels, thus, caution should be exercised when generalizing these findings to other locations.

Our findings showed an effect of air pollution on mortality in Shijiazhuang and emphasized the necessity of further controlling PM_2.5and continuing the significant achievements in PM_2.5control that have been made in Shijiazhuang. Futher studies on associations of components of PM_2.5with cause-specific mortality are still needed to guide environmental health policies to improve population health.

No conflicts of interest.