Ambient particulate matter pollution and adult hospital admissions for pneumonia in urban China: A national time series analysis for 2014 through 2017.

BACKGROUND: The effects of ambient particulate matter (PM) pollution on pneumonia in adults are inconclusive, and few scientific data on a national scale have been generated in low- or middle-income countries, despite their much higher PM concentrations. We aimed to examine the association between PM levels and hospital admissions for pneumonia in Chinese adults. METHODS AND
FINDINGS: A nationwide time series study was conducted in China between 2014 and 2017. Information on daily hospital admissions for pneumonia for 2014-2017 was collected from the database of Urban Employee Basic Medical Insurance (UEBMI), which covers 282.93 million adults. Associations of PM concentrations and hospital admissions for pneumonia were estimated for each city using a quasi-Poisson regression model controlling for time trend, temperature, relative humidity, day of the week, and public holiday and then pooled by random-effects meta-analysis. Meta-regression models were used to investigate potential effect modifiers, including cities' annual-average air pollutants concentrations, temperature, relative humidity, gross domestic product (GDP) per capita, and coverage rates by the UEBMI. More than 4.2 million pneumonia admissions were identified in 184 Chinese cities during the study period. Short-term elevations in PM concentrations were associated with increased pneumonia admissions. At the national level, a 10-μg/m3 increase in 3-day moving average (lag 0-2) concentrations of PM2.5 (PM ≤2.5 μm in aerodynamic diameter) and PM10 (PM ≤10 μm in aerodynamic diameter) was associated with 0.31% (95% confidence interval [CI] 0.15%-0.46%, P < 0.001) and 0.19% (0.11%-0.30%, P < 0.001) increases in hospital admissions for pneumonia, respectively. The effects of PM10 were stronger in cities with higher temperatures (percentage increase, 0.031%; 95% CI 0.003%-0.058%; P = 0.026) and relative humidity (percentage increase, 0.011%; 95% CI 0%-0.022%; P = 0.045), as well as in the elderly (percentage increase, 0.10% [95% CI 0.02%-0.19%] for people aged 18-64 years versus 0.32% [95% CI 0.22%-0.39%] for people aged ≥75 years; P < 0.001). The main limitation of the present study was the unavailability of data on individual exposure to PM pollution.
CONCLUSIONS: Our findings suggest that there are significant short-term associations between ambient PM levels and increased hospital admissions for pneumonia in Chinese adults. These findings support the rationale that further limiting PM concentrations in China may be an effective strategy to reduce pneumonia-related hospital admissions.

Introduction

Pneumonia is a major cause of mortality and morbidity worldwide [1, 2]. In China, an estimated 2.5 million pneumonia cases occur annually, and 5% of these individuals die of pneumonia-related illness [3]. Furthermore, pneumonia is closely related to complications such as pleurisy, lung abscess, and septicemia, as well as to cardiovascular disease [4, 5]. The risks of pneumonia and pneumonia-related mortality increase with age [6]; thus, the incidence of pneumonia is projected to increase as a result of global population aging [7].

Air pollution has emerged as a significant public health problem worldwide [8]. Despite plausible hypotheses linking air pollution with pneumonia [9-12], limited evidence of the association is available, especially in adults. Some studies have started to report some evidence of an association between ambient particulate matter (PM) pollution and respiratory infections, including pneumonia. A time series study suggested an association between fine particulate matter (PM2.5, PM ≤2.5 μm in aerodynamic diameter) and increased admissions for respiratory tract infections in Medicare enrollees (aged >65 years) in 204 United States urban counties [13]. Similarly, another study reported significant effects of PM10 (PM ≤10 μm in aerodynamic diameter) on pneumonia admissions among 36 US citizens aged 65 years or older [14]. Strosnider and colleagues assessed age-specific acute effects of PM2.5 on respiratory emergency department visits in 17 US states and reported that short-term exposure to PM2.5 was associated with increased emergency department visits for pneumonia in people aged 19–65 years, but not in people aged <19 or ≥65 years [15]. A recent case-crossover study from the US demonstrated significant associations between short-term elevations in PM2.5 and greater healthcare utilization for acute lower respiratory infection in both children and adults [16]. Host and colleagues found increased admissions for lower respiratory tract infections following PM2.5 exposure in six French cities [17], and two time series studies in Hong Kong provided further evidence on the acute effects of PM on respiratory tract infections/pneumonia [18, 19]. However, these studies were conducted primarily in high-income countries/cities where characteristics of air pollution and socioeconomic status differ from those in low- or middle-income countries. The effects of short-term exposure to PM pollution on pneumonia in low- or middle-income counties require further investigation.

China, the largest low- or middle-income country, has one of the highest PM2.5 concentrations worldwide [20]. However, only a few studies have explored the associations between PM pollution and pneumonia, and were subjected to single city/hospital study and small sample size [21, 22]. In this study, we examined the short-term associations between concentrations of ambient PM pollution and daily hospital admissions for pneumonia in adults in China between 2014 and 2017.

Methods

Study sites

A total of 184 Chinese cities were finally included in this study. shows the locations of the 184 cities, representing a geographic distribution across China. Cities were selected according to the availability of daily hospital admission and air pollution data. Cities with ≤1-year hospital admission records were excluded given the fact that a long enough study period in a time series analysis is needed to ensure credible precision and power. Individuals’ detailed disease diagnosis (International Classification of Diseases [ICD] code and text of disease diagnosis) was required to identify pneumonia admissions. Cities with no information on disease diagnosis recorded in the database were also excluded (e.g., Beijing and Shanghai). In addition, we applied additional rules to control the quality of the original data. We regressed the city-specific average numbers of pneumonia admission per day on the number of enrollers of the city. Cities with an absolute value of standardized residuals >3 were considered as the outliers and were removed from the analysis. China has gradually introduced PM2.5 in the national air quality monitoring network and publicized real-time monitoring data since 2013. In this study, the total study period encompassed 2014–2017, and the years of analysis differed by city based on the availability of air pollution data. Of the 184 cities, 78 cities had 3-year PM data records and 106 cities had 4-year PM data records. The years of the study period for each city were presented in .

Locations of the 184 Chinese cities included in the study.

The base map was obtained from Natural Earth (https://naturalearthdata.com).

Study population

In China, there are three main health insurance schemes: the Urban Employee Basic Medical Insurance (UEBMI), the Urban Residence Basic Medical Insurance, and the New Rural Cooperative Medical Scheme, covering more than 92% of the population [23]. Private medical insurance covers little in China and is generally supplementary to the basic schemes. Daily pneumonia admission data were obtained from the UEBMI database. The UEBMI covers urban employees and retired employees (aged ≥18 years). All employers in urban areas, including government agencies and institutions, state-owned enterprises, private businesses, social organizations, and other private entities and their employees (retirees included), are obligated to enroll in UEBMI [24]. At the end of 2016, 282.93 million individuals were recorded in the database. The original data source was medical claims data for urban employees, which included sex, age, date of admission, and cause of admission. Information regarding the number of individuals enrolled in the database, city residents, and coverage rates of these cities for the UEBMI database were published previously [25]. We have legal access to this national insurance information through the big data platform of national medical insurance data for risk prevention and control, established by the Ministry of Human Resources and Social Security of the People’s Republic of China and Beijing HealthCom Data Technology. Hospitalizations with a principal diagnosis of pneumonia (ICD-10 codes J12–J18) were identified from the database. Because the health information was primarily the city-specific daily count of pneumonia admissions, i.e., summarized data (overall, and by age and sex subgroups) without any individual identifiers, this study was exempted from Institutional Review Board approval by the Ethics Committee of Peking University Health Science Center, Beijing, China. The need for informed consent was also waived by the Institutional Review Board. This study is reported as per the STROBE guideline ().

Environmental data

Data on hourly PM2.5 and PM10 measurements in each city were obtained from the National Air Pollution Monitoring System. Each city has 1–17 monitoring stations. China has issued detailed standards on placement of monitoring stations and the air pollution monitoring process [26]; the monitoring stations cannot be located in the direct vicinity of apparent emission sources; thus, measurements reflect the general urban background concentrations. These measurements have therefore been used extensively to represent population exposure [27-29]. Data on sulfur dioxide (SO2), nitrogen dioxide (NO2), carbon monoxide (CO), and ozone (O3) concentrations were also collected from the same source. We calculated 24-hour averages by averaging hourly air pollution measurements within a day. In each city, the measurements from all monitoring stations were averaged to represent population exposure [27, 28, 30]. Rates of missing data were 0.66% for PM2.5 and 0.15% for PM10. Days with missing monitoring measurements were excluded from the analysis. Data on hourly atmospheric temperature and relative humidity were provided by the China Meteorological Data Sharing Service System. Each city has 1–3 meteorological monitoring stations, all located in urban areas. Rates of missing temperature and relative humidity data were only 0.25%. Daily (24-hour) mean temperature and relative humidity were calculated by averaging all valid monitoring measurements in each city. More detailed information about health, air pollution, and meteorological data sources is given in our previous publications [24, 25].

Statistical analysis

We used a common two-stage method to examine the associations between PM and pneumonia admissions [13, 24, 27, 30–32]. The method and the model used in this study were designed before the analyses were conducted, and the prespecified statistical analysis plan is present in . In stage 1, we assessed the city-specific associations between PM and pneumonia admissions using a quasi-Poisson regression model that allows for over-dispersed admission counts. Several covariates were included in the model, following previous published studies [13, 27, 32, 33]: (1) a natural cubic spline for time with 7 degrees of freedom (df) per year to adjust for seasonality and long-term trends, (2) natural cubic splines for 3-day moving average temperature with 6 df and relative humidity with 3 df to control for nonlinear and delayed effects of meteorology, and (3) indicators of day of the week and public holiday. The selection of df values was based on the parameter used in several recent large national studies [27, 28, 32, 33]. The model was as shown below: where E(Yt) is the expected number of pneumonia admissions on day t; β indicates the log-relative risk of pneumonia associated with a unit increase of PM concentrations; ns() indicates a natural cubic spline function; and temperature and relative humidity indicate 3-day moving averages. This model was in line with those used in previous studies [13, 24, 27, 30, 31] and was tested in sensitivity analyses, described in detail below. We also conducted unadjusted analyses with only PM and a natural cubic spline for time, with 7 df per year in the regression model. In stage 2, we conducted random-effects meta-analyses to combine the city-specific effect estimates. To explore regional heterogeneity of the associations, we divided the 184 cities into six geographical regions, i.e., East, Middle-south, Southwest, Northwest, North, and Northeast regions [27]. We used single lags of 0, 1, and 2 days to assess the temporal association between PM and pneumonia. Considering that single-day lag models may underestimate PM effects, we also estimated associations with 3-day (lag 0–2) moving average PM concentrations.

We plotted the national-average concentration-response curves of the associations of PM with pneumonia admissions using the approach of the Air Pollution and Health—A European Approach (APHEA) project [27, 32, 34]. Briefly, we replaced the linear term of PM in the main model with a B-spline function with two knots at 60 and 150 μg/m3 for PM2.5 and 100 and 200 μg/m3 for PM10, based on the distribution of PM concentrations in each city. We then estimated five regression coefficients of the spline function and the 5 × 5 variance-covariance matrix in each city. Finally, we fitted multivariable random-effect models to combine the city-specific components of spline estimates. Variance-covariance matrix is a matrix whose element in the i, j position is the covariance between the i-th and j-th elements of a random vector. The regression coefficients derived from the main model with a B-spline function for PM.

We examined effect modification of the relationship between short-term ambient PM concentrations and pneumonia hospital admissions in analyses by sex, age, and region (northern and southern China). The regional division followed the Huai River-Qinling Mountains line [27, 30]. We assessed the differences in the estimates using a Z-test [35]. We also fitted meta-regression models to investigate several city characteristics as potential effect modifiers, including cities’ annual-average air pollutants concentrations, air temperature and relative humidity, gross domestic product (GDP) per capita, and coverage rates by the UEBMI [27, 30]. Cities’ annual-average air pollutants concentrations, temperature, and relative humidity were calculated from daily measurements during the study period. City-specific relative risks and their confidence intervals (CIs) as the outcome were meta-regressed on each continuous variable of city characteristics. We fitted the random-effects meta-regression model using city-level estimations from the first stage as yi = βxi + μi + ei, where yi is the city-specific estimation, xi is the city-level characteristic variable, μi is the city-specific random effect characterized by the between-city variance, and ei is a normal error term [36]. Random-effects meta-regression can be considered as an extension to the random-effects meta-analysis (i.e., DerSimonian and Laird model) that includes study-level (i.e., city-level in our case) covariates. The model is implemented by the metareg function in STATA. The algorithm for random-effects meta-regression firstly estimated the between-study (between-city) variance and then estimated the coefficients, β, by weighted least squares by using the weights 1/(σi2 + τ2), where σi2 is the standard error of the estimated effect in study (city) i, and τ2 is the between-study (between-city) variance, advocated by Thompson and Sharp [37].

Sensitivity analysis

We conducted several sensitivity analyses: (1) We tested the potential confounding effects of gaseous pollutants in two-pollutant models; (2) we separately examined the associations of PM with pneumonia in cities with only 3- or 4-year data to compare with the primary analysis; (3) we examined the associations between PM and pneumonia in cities with different levels of population coverage by the UEBMI (<20% and ≥20%); (4) we examined the associations excluding cities with ≤2 monitoring stations; (5) we conducted a sensitivity analysis with additional adjustment of the hospitalization of influenza; and (6) to examine whether we appropriately specified the regression model used in this analysis, we used alternative df for time (6–10 per year) or used penalized spline functions for time and meteorological variables in the model.

Statistical analyses were conducted in R version 3.2.2 (R Development Core Team 2008) and STATA software, version 12 (StataCorp, College Station, TX), and a two-sided P < 0.05 was considered statistically significant. The effect estimates are presented as percentage increases and 95% CIs in pneumonia admissions per 10-μg/m3 increase in PM concentrations.

Results

There were 4.2 million hospital admissions for pneumonia in the 184 Chinese cities between 2014 and 2017 in our study. presents summary statistics on citywide pneumonia admissions, PM concentrations, and weather conditions at an annual-average level during the study period. The mean number of daily hospital admissions for pneumonia was 26 (range: 1 to 376). The national-average PM2.5 concentration was 50 μg/m3 (range: 15 to 102 μg/m3), and the mean PM10 concentration was 89 μg/m3 (range: 28 to 193 μg/m3). The average citywide annual-mean temperature was 14°C (range: 0 to 24°C). City-specific characteristics, including the number of individuals enrolled in the database, city residents, coverage rates of the population by UEBMI database, and annual-average PM2.5 and PM10 concentrations, temperature, relative humidity, and daily pneumonia hospitalizations are presented in .

shows the Spearman correlation coefficient values for the environmental variables. At the national level, daily PM2.5 and PM10 concentrations were positively correlated with SO2 (PM2.5: r = 0.56, P < 0.001; PM10: r = 0.59, P < 0.001), NO2 (PM2.5: r = 0.64, P < 0.001; PM10: r = 0.64, P < 0.001), and CO (PM2.5: r = 0.61, P < 0.001; PM10: r = 0.57, P < 0.001) concentrations. O3 concentrations were not correlated with PM2.5 (r = −0.02, P = 0.452) or PM10 (r = 0.05, P = 0.215). There were inverse and weak correlations between PM2.5 and PM10 concentrations and temperature (PM2.5: r = −0.26, P < 0.001; PM10: r = −0.26, P < 0.001) and relative humidity (PM2.5: r = −0.08, P < 0.001; PM10: r = −0.28, P < 0.001).

City-specific estimates of the associations between 3-day moving average (lag 0–2) concentrations of PM2.5 and PM10 and hospital admissions for pneumonia are presented in . shows the national-average percentage increases in daily hospital admissions for pneumonia associated with a 10-μg/m3 increase in PM2.5 and PM10 concentrations at lags 0, 1, 2, and 0–2 days. Overall, we observed similar lag patterns for PM2.5 and PM10. For single-day lags, the strongest associations occurred at lag 0. The associations weakened considerably at lag days 1 and 2. Lag 0–2 generated the highest estimates. On average, a 10-μg/m3 increase in PM2.5 and PM10 concentrations at lag 0–2 corresponded to a 0.31% (0.15% to 0.46%, P < 0.001) and 0.19% (0.11% to 0.30%, P < 0.001) increase in pneumonia admissions, respectively.

presents the results of stratified analyses. The estimates were similar in males (PM2.5: percentage increase, 0.31%; 95% CI 0.12%–0.49%; PM10: percentage increase, 0.20%; 95% CI 0.13%–0.28%) and females (PM2.5: percentage increase, 0.33%; 95% CI 0.13%–0.53%; PM10: percentage increase, 0.15%; 95% CI 0.05%–0.25%) (the P values for the differences in the estimates were 0.887 for PM2.5 and 0.422 for PM10). The associations between PM and pneumonia were stronger in people aged ≥75 years, but the difference between the strata was only significant for PM10 (the P values for the differences in the estimates were 0.201 for PM2.5 and P < 0.001 for PM10). The effect estimates were higher in the southern region (PM2.5: percentage increase, 0.40%; 95% CI 0.16%–0.65%; PM10: percentage increase, 0.35%; 95% CI 0.16%–0.54%) than in the northern region (PM2.5: percentage increase, 0.23%; 95% CI 0.04%–0.43%; PM10: percentage increase, 0.12%; 95% CI 0%–0.24%), but the between-region difference was not significant for PM2.5 (the P values for the regional differences in the estimates were 0.276 for PM2.5 and 0.045 for PM10). We grouped the cities further into six geographical areas; the average estimates for the six areas are presented in . There was a significant heterogeneity in the PM–hospitalization associations across different regions. The associations were more evident in the Middle-south, East, and North regions.

We noted a slightly nonlinear concentration-response curve between PM2.5 and pneumonia admissions, where there appears to be a plateau at higher concentrations. This evidence of linearity is consistent with studies that examined short-term PM exposure and other outcomes [13–15, 19, 27]. The curve of the association between PM10 and pneumonia increased sharply at concentrations below 100 μg/m3 and then climbed relatively moderately as concentrations increased (). The estimated risk of PM10 concentrations for pneumonia was 0.38% (0.18% to 0.57%, P < 0.001) at PM10 < 100 μg/m3 and 0.10% (−0.10% to 0.30%, P = 0.754) at PM10 ≥ 100 μg/m3.

National-average concentration-response curves of 3-day moving average (lag 0–2) concentrations of PM2.5 and PM10 and hospital admissions for pneumonia in 184 cities in China, 2014–2017.

Note, y-axes of the two graphs are scaled differently. The solid line represents relative changes and the dashed lines represent the 95% CIs. The vertical lines represent the air quality guidelines or standards for 24-hour average concentrations of PM2.5 and PM10. PM2.5: 25 μg/m3 is the World Health Organization (WHO) air quality guideline for daily PM2.5 concentrations, 35 μg/m3 is the Chinese Grade I standard for daily PM2.5 concentrations, and 75 μg/m3 is the Chinese Grade II standard; PM10: 50 μg/m3 is both the WHO air quality guideline and Chinese Grade I standard for daily PM10 concentrations, and 150 μg/m3 is the Chinese Grade II standard for daily PM10 concentrations. PM2.5, particulate matter ≤2.5 μm in aerodynamic diameter; PM10, particulate matter ≤10 μm in aerodynamic diameter.

summarizes the results of meta-regression analyses of effect modification on the associations between PM and pneumonia by city-level characteristics. We observed stronger associations between PM10 and pneumonia in cities with higher annual-average temperatures or relative humidity. For each 10-μg/m3 increase in PM10 concentrations, a city with 1°C higher annual-mean temperature with respect to another city would see an additional 0.031% (0.003% to 0.058%, P = 0.026) increase in pneumonia admissions; a city with 1% higher annual-mean relative humidity with respect to another city would see an additional 0.011% (0% to 0.022%, P = 0.045) increase in pneumonia admissions. We observed the same direction of effect modification by temperature (P = 0.129) and relative humidity (P = 0.378) on the association between PM2.5 and pneumonia, but the effects were not significant. No evidence was found for effect modification by annual-average PM2.5 (percentage increase, −0.019%; 95% CI −0.113% to 0.075%, P = 0.688) or PM10 (percentage increase, 0; 95% CI −0.064% to 0.064%, P = 0.997). Annual-average NO2 (the P values were 0.737 for PM2.5 and P = 0.861 for PM10) or CO (the P values were 0.466 for PM2.5 and P = 0.367 for PM10) concentrations, GDP per capita (the P values were 0.864 for PM2.5 and P = 0.655 for PM10), or the coverage rate of the population (the P values were 0.835 for PM2.5 and P = 0.665 for PM10) did not significantly modify the associations.

The associations of PM2.5 and PM10 with pneumonia remained after adjusting for SO2 (PM2.5: percentage increase, 0.22%; 95% CI 0.07%–0.37%; P = 0.006; PM10: percentage increase, 0.12%; 95% CI 0.02%–0.22%; P = 0.015) and O3 (PM2.5: percentage increase, 0.31%; 95% CI 0.16%–0.46%; P < 0.001; PM10: percentage increase, 0.20%; 95% CI 0.09%–0.30%; P < 0.001); the effects remained positive but not significant after controlling for CO (PM2.5: percentage increase, 0.12%; 95% CI −0.03%–0.27%; P = 0.127; PM10: percentage increase, 0.06%; 95% CI −0.04%–0.16%; P = 0.256) and NO2 (PM2.5: percentage increase, 0.10%; 95% CI −0.05%–0.25%; P = 0.271; PM10: percentage increase, 0.04%; 95% CI −0.06%–0.14%; P = 0.431) ().

lists the results of the sensitivity analyses. Consistent associations were observed in datasets with different lengths of data-years. Specifically, for PM2.5, the increases in pneumonia admissions were 0.28% (0% to 0.56%, P = 0.044) in 78 cities with 3-year data and 0.37% (0.20% to 0.56%, P < 0.001) in 106 cities with 4-year data; for PM10, the corresponding values were 0.13% (−0.07% to 0.34%, P = 0.213) in cities with 3-year data and 0.24% (0.11% to 0.37%, P < 0.001) in cities with 4-year data. The estimates were similar in cities with different levels of population coverage (<20% and ≥20%). The estimates were slightly attenuated after excluding cities with ≤2 monitoring stations (PM2.5: percentage increase, 0.30%; 95% CI 0.15%–0.46%, P < 0.001; PM10: percentage increase, 0.21%; 95% CI 0.10%–0.32%, P < 0.001) or after adjustment of hospitalization of influenza (PM2.5: percentage increase, 0.29%; 95% CI 0.13%–0.45%, P < 0.001; PM10: percentage increase, 0.18%; 95% CI 0.07%–0.29%, P < 0.001). The national-average estimates generated from models with alternative df for time (6–10 per year, all P < 0.05) or with penalized spline functions for time and meteorological variables (PM2.5: percentage increase, 0.31%; 95% CI 0.17%–0.45%, P < 0.001; PM10: percentage increase, 0.22%; 95% CI 0.12%–0.32%, P < 0.001) were comparable to those generated from the base model.

Discussion

Using comprehensive data, we examined the associations between ambient PM pollution and pneumonia hospitalizations in 184 Chinese cities. We also investigated potential effect modifications by a variety of demographic, geographical, and meteorological characteristics. We found that short-term elevations in PM2.5 and PM10 were associated with higher pneumonia-related hospital admissions in Chinese adults. To our knowledge, this is the first study in China, or even in the low- or middle-income countries, to investigate whether short-term changes in ambient PM concentrations are related to increases pneumonia hospital admissions on a national scale.

We observed increased risk of hospital admissions for pneumonia in association with both PM2.5 and PM10 at lag 0 and lag 1, in line with previous findings from studies examining short-term PM exposures [14, 21]. The highest estimates for single-day lag models were observed for lag 0, indicating immediate PM effects on pneumonia. In addition, the estimate at lag 0–2 was slightly higher than at the estimate for lag 0, suggesting that there might be some cumulative effects of PM on pneumonia admissions as well. We also explored the concentration-response curve of the associations between PM and pneumonia hospitalizations. China is one of the most polluted countries worldwide [38]. The average city-specific annual-mean PM2.5 and PM10 concentrations during the study period were 50 μg/m3 and 89 μg/m3, respectively, allowing us to assess the concentration-response pattern at high concentrations. The curves plateau at high concentrations, indicating that the increase in risk of pneumonia admission is greater at lower compared with higher concentrations. A similar shape of the association was also observed in other studies [27]. The leveling off at high concentrations might be explained in that people vulnerable to PM exposure may have developed symptoms and sought treatment before PM concentrations reached high concentrations. Other reasons could also be that there may be different health risks associated with different PM sizes and that people avoid spending time outdoors or start wearing face masks when air pollution is severe.

The PM2.5 estimates were higher than those of PM10 at all lag days in this study, in line with previous findings on pneumonia [22, 39]. PM10 is representing exposure to more particles in the coarse range (PM10–2.5) than PM2.5. The sources, composition, and lung deposition patterns of PM10–2.5 vary from those of PM2.5 [40, 41]. The chemical composition of PM differs by size, with more crustal materials in PM10–2.5 and more combustion-related constituents in PM2.5 [31, 41]. It was reported that PM2.5 can penetrate deep into the lungs, reaching the bronchioles and depositing inside the alveoli. In addition, PM2.5 has a larger surface area than PM10 and thus can absorb more toxic substances per unit mass [42].

There are several possible mechanisms linking PM pollution to pneumonia. It has been postulated that air pollution may act as an irritant and evoke defensive responses in the airways, including increased mucus secretion and bronchial hyperreactivity [9]. PM is a potent oxidant that can produce free radicals and cause oxidative stress in lung cells [10]. An animal study demonstrated that short-term exposure to concentrated ambient particles could increase the levels of reactive oxygen species in the lung [11]. PM-induced oxidative stress might impair the cellular defense and immune system, increasing the susceptibility to infection [19]. In addition, PM might enhance the susceptibility to infection through the suppression of the immune response [43].These toxicological evidences were consistent with our findings on the short-term association between ambient PM concentrations and pneumonia hospitalizations. However, it should be noted that our study does not reveal the mechanisms related to PM that may be active in these processes, and thus further study is still required to verify the causality.

In the two-pollutant models, the associations between PM2.5 and PM10 and pneumonia remained positive but were not significant after adjusting for CO and NO2 concentrations. Similarly, a case-crossover study by Tsai and Yang demonstrated a positive association between PM2.5 and hospital admissions for pneumonia in a single-pollutant model, but the estimate decreased and became nonsignificant when controlling for CO and NO2 [44]. We previously reported positive effects of PM2.5 and PM10 on hospital admissions for respiratory disease, but the estimates decreased dramatically and even became negative after adjusting for NO2 [45]. NO2 and CO are closely associated with vehicle exhaust emissions [46, 47]. In China, a substantial fraction of PM2.5 originates in traffic emissions [48]. The high correlation between pollutants in the regression model may have weakened the observed effects, although the CIs of the estimates were not remarkably inflated in the two-pollutant model. Another possible explanation is that the effects of PM may be confounded by those of other air pollutants. Additional studies are warranted to investigate the independent effects of PM on pneumonia.

There were some effect modifiers in the PM-pneumonia associations. First, the elderly had a higher risk of PM-associated pneumonia. Biological functions decline in the elderly, and this population has a higher prevalence of chronic health conditions [9, 49] that may contribute to the increased vulnerability to air pollution. Second, the effects were stronger in cities with higher temperatures (or higher relative humidity, as they are highly correlated). There are several possible explanations. High temperature itself is associated with increased risk of pneumonia [50, 51]. It was reported that high temperature impacts the emission, transportation, dilution, chemical transformation, and deposition of pollutants [52]. Thus, there may exist the interaction between high temperatures and PM. Previous studies also show moderate evidence of the modifying effect of temperature on PM10. A meta-analysis reported that the relative risks for respiratory mortality per 10 μg/m3 increase in PM10 in the low, middle, and high temperature level were 1.005 (95% CI 1.000–1.010), 1.008 (1.006–1.010), and 1.019 (1.010–1.028), respectively [53]. This is consistent with our findings. In addition, the larger estimate might be associated with exposure patterns. People generally spend more time outdoors in warmer cities, resulting in smaller measurement errors [52, 54].

This study provided detailed estimates of the risk of PM-associated hospital admission for pneumonia through the use of data from a national health insurance program that serves approximately one fifth of the population in China, and offered a unique advantage in identifying potential modifiers of the associations. Our study also has some limitations. First, the use of citywide monitoring measurements as a proxy for population exposure could cause exposure measurement errors, which tend to bias the estimates downward [55]. Second, children (aged <18 years) were not included in this study because UEBMI is for urban working and retired employees, although young children may be more sensitive to air pollution, similarly to the elderly. In addition, only urban employed and retired individuals were included in this study based on the type of insurance enrolled. Due to the differences in sociodemographic characteristics and PM concentrations between rural and urban areas, the generalizability of our findings should be interpreted with caution. Third, both one-stage and two-stage methods could be applied in multicity analysis. In this study, the two-stage method was applied, considering the issue of the computational complexity [56], in line with previous studies [13, 27, 32]. The variances estimated in the first step in the two-stage analysis may not be exactly those in the one-stage method. However, it was reported that one-stage and two-stage methods often generate very similar results [57]. Fourth, in this study, we examined the associations between PM and hospital admissions for pneumonia, not including emergency room visits. As hospital admissions generally include the more severe cases, future investigations are needed to assess the effects of PM on other morbidity outcomes (e.g., emergency room visits). Fifth, as in other environmental health studies using a large administrative health database [13, 27, 30, 58], data on several patient-level variables, such as medical history and smoking status, were not available, limiting the ability to identify potentially susceptible populations. More detailed individual information would be needed to assess the modifying effects of individual characteristics on the associations between PM pollution and pneumonia.

In summary, we found that short-term elevations in PM were associated with increased pneumonia-related hospital admissions in Chinese adults. Our findings support the rationale to further limit PM concentrations in China.

Summary statistics on city-level characteristics in 184 Chinese cities.

(DOCX)

Click here for additional data file.

City-specific percentage increase with 95% CI in daily hospital admissions for pneumonia associated with a 10 μg/m3 increase in PM2.5 and PM10 concentrations (lag 0–2) in 184 Chinese cities, 2014–2017.

(DOCX)

Click here for additional data file.

Regional-average percentage increase with 95% CI in daily hospital admissions for pneumonia associated with a 10 μg/m3 increase in PM2.5 and PM10 concentrations (lag 0–2) in 184 Chinese cities, 2014–2017.

(DOCX)

Click here for additional data file.

National-average percentage increase with 95% CI in daily hospital admissions for pneumonia associated with a 10 μg/m3 increase in PM2.5 and PM10 concentrations (lag 0–2) in two-pollutant models in 184 Chinese cities, 2014–2017.

(DOCX)

Click here for additional data file.

Results of sensitivity analyses.

(DOCX)

Click here for additional data file.

(DOC)

Click here for additional data file.

Statistical analysis plan.

(DOCX)

Click here for additional data file.

9 Oct 2019

Dear Dr. Hu,

Thank you very much for submitting your manuscript "Ambient Particulate Matter Pollution and Adult Hospital Admissions for Pneumonia in Urban China: A national time-series analysis" (PMEDICINE-D-19-02197) for consideration at PLOS Medicine.

Your paper was evaluated by a senior editor and discussed among all the editors here. It was also discussed with an academic editor with relevant expertise, and sent to four independent reviewers, including a statistical reviewer. The reviews are appended at the bottom of this email and any accompanying reviewer attachments can be seen via the link below:

[LINK]

In light of these reviews, I am afraid that we will not be able to accept the manuscript for publication in the journal in its current form, but we would like to consider a revised version that addresses the reviewers' and editors' comments. Obviously we cannot make any decision about publication until we have seen the revised manuscript and your response, and we plan to seek re-review by one or more of the reviewers.

In revising the manuscript for further consideration, your revisions should address the specific points made by each reviewer and the editors. Please also check the guidelines for revised papers at http://journals.plos.org/plosmedicine/s/revising-your-manuscript for any that apply to your paper. In your rebuttal letter you should indicate your response to the reviewers' and editors' comments, the changes you have made in the manuscript, and include either an excerpt of the revised text or the location (eg: page and line number) where each change can be found. Please submit a clean version of the paper as the main article file; a version with changes marked should be uploaded as a marked up manuscript.

In addition, we request that you upload any figures associated with your paper as individual TIF or EPS files with 300dpi resolution at resubmission; please read our figure guidelines for more information on our requirements: http://journals.plos.org/plosmedicine/s/figures. While revising your submission, please upload your figure files to the PACE digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email us at PLOSMedicine@plos.org.

We expect to receive your revised manuscript by Oct 30 2019 11:59PM. Please email us (plosmedicine@plos.org) if you have any questions or concerns.

***Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.***

We ask every co-author listed on the manuscript to fill in a contributing author statement, making sure to declare all competing interests. If any of the co-authors have not filled in the statement, we will remind them to do so when the paper is revised. If all statements are not completed in a timely fashion this could hold up the re-review process. If new competing interests are declared later in the revision process, this may also hold up the submission. Should there be a problem getting one of your co-authors to fill in a statement we will be in contact. YOU MUST NOT ADD OR REMOVE AUTHORS UNLESS YOU HAVE ALERTED THE EDITOR HANDLING THE MANUSCRIPT TO THE CHANGE AND THEY SPECIFICALLY HAVE AGREED TO IT. You can see our competing interests policy here: http://journals.plos.org/plosmedicine/s/competing-interests.

Please use the following link to submit the revised manuscript:

https://www.editorialmanager.com/pmedicine/

Your article can be found in the "Submissions Needing Revision" folder.

To enhance the reproducibility of your results, we recommend that you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. For instructions see http://journals.plos.org/plosmedicine/s/submission-guidelines#loc-methods.

Please ensure that the paper adheres to the PLOS Data Availability Policy (see http://journals.plos.org/plosmedicine/s/data-availability), which requires that all data underlying the study's findings be provided in a repository or as Supporting Information. For data residing with a third party, authors are required to provide instructions with contact information for obtaining the data. PLOS journals do not allow statements supported by "data not shown" or "unpublished results." For such statements, authors must provide supporting data or cite public sources that include it.

We look forward to receiving your revised manuscript.

Sincerely,

Caitlin Moyer, Ph.D.

Associate Editor

PLOS Medicine

plosmedicine.org

-----------------------------------------------------------

Requests from the editors:

1.Did your study have a prospective protocol or analysis plan? Please state this (either way) early in the Methods section.

a) If a prospective analysis plan (from your funding proposal, IRB or other ethics committee submission, study protocol, or other planning document written before analyzing the data) was used in designing the study, please include the relevant prospectively written document with your revised manuscript as a Supporting Information file to be published alongside your study, and cite it in the Methods section. A legend for this file should be included at the end of your manuscript.

b) If no such document exists, please make sure that the Methods section transparently describes when analyses were planned, and when/why any data-driven changes to analyses took place.

c) In either case, changes in the analysis—including those made in response to peer review comments—should be identified as such in the Methods section of the paper, with rationale.

2. Abstract Methods and Findings: Line 29: Please change 0.28 billion adults to a more intuitive number (perhaps report in millions instead of billions).

3. Abstract Methods and Findings: Please include the study design.

4. Abstract Methods and Findings: Please quantify the results of the association between PM and hospital admissions with p values.

5. Abstract Methods and Findings: Please quantify the results of the effects of PM with higher temperatures/humidity and in elderly adults with 95% CIs and p values.

6. Abstract Conclusions: Please avoid assertions of primacy (“As the first study in China to investigate…”).

7. Abstract Conclusions: Your study is observational and therefore causality cannot be inferred. Please remove language that implies causality (...may assist in understanding how PM causes lung-inflammation diseases.).

8. At this stage, we ask that you include a short, non-technical Author Summary of your research to make findings accessible to a wide audience that includes both scientists and non-scientists. The Author Summary should immediately follow the Abstract in your revised manuscript. This text is subject to editorial change and should be distinct from the scientific abstract. Please see our author guidelines for more information: https://journals.plos.org/plosmedicine/s/revising-your-manuscript#loc-author-summary

9. Methods: Please provide details on the 184 cities included in the study, including criteria for the selection of cities and the years of the study period for each city.

10. Results: Line 200: Please provide a range for the number of daily hospital admissions for pneumonia.

11. Results: Line 268: Please change “statistically insignificant” to “effects were not significant after controlling for…”

12. Results: Lines 270-277: Please quantify these results with 95% CIs and P values, and clarify that the results are presented in S4 Table.

13. Discussion: Lines 280-285: Your study is observational and thus causality cannot be inferred. Please revise throughout and remove causal language (such as, “...we assessed the effects of ambient PM pollution on pneumonia admissions…” and “...investigate the acute effects of PM on pneumonia…”).

14. Discussion: Lines 289-290: Please revise causal language (“...indicating that short-term PM exposure (for even less than a day) could trigger pneumonia.”)

15. Discussion: Lines 293-294, Lines 299-300, Line 305,: Please revise causal language.

16. Discussion: Line 307: Please revise the text to avoid over-interpretation of the conclusion here.

17. Discussion: Line 362-363: Please revise the text to avoid causal language (...”assist in understanding how PM causes lung-inflammation…”)

18. Figure 1: Please revise Figure 1 such that it is possible to see the locations of the cities more clearly. Please clarify the meaning of the variation in dot size.

19. Figure 2: Please provide units for the axes.

20. Table 3 and Table 4, and S2, S3, and S4 tables: Please indicate if these analyses are adjusted for any factors, and if so, please specify which factors are being adjusted for, and provide unadjusted values.

21. Thank you for including the completed STROBE checklist as Supporting Information. When completing the checklist, please use section and paragraph numbers, rather than page numbers. Please add the following statement, or similar, to the Methods: "This study is reported as per the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guideline (S1 Checklist)."

Comments from the reviewers:

Reviewer #1: I confine my remarks to statsitical aspects of this paper. These were very well done indeed and I have no problem recommending publication.

Peter Flom

Reviewer #2: The authors have conducted an interesting analysis on a health outcome that has not been well studied, which could aid in both the overall interpretation of PM effects on respiratory infections generally, but also specifically the potential health implications of PM exposures within China. Generally, the analysis is well conducted, but additional clarification is needed in many instances, and the authors should consider revising the approach used to examine effect modification. The following comments identify those areas and issues for the authors to consider in revising the manuscript:

- Throughout the manuscript there is too much reliance on statistical significance in presenting results from the study as well as in the discussion of studies in the introduction and discussion. The authors need to remove this and focus on patterns of associations.

Abstract:

Line 28: I would specify that these are hospital admissions because I could see some getting confused as to whether these are hospital admissions or emergency department visits.

Line 32: should note here what effect modifiers will be examined, otherwise you don't know which ones until in the results section and seeing Table 5.

Line 41-42: This does not seem like something that should be in the abstract.

Introduction:

Line 69: This does not seem like something that should be in the abstract.

Line 72: This doesn't make sense to put here because this paper is focusing on short-term exposure and the estimates here are based on health impact functions for long-term PM2.5 exposures. It's not really making the correct point.

Line 74: should be association, not correlation.

Line 75: From this point on, there needs to be more context put around these studies versus just listing them out and what they found. Something like, some studies have started to report some evidence of an association between air pollution exposure (or PM2.5 specifically) and respiratory infections, including pneumonia. After that it will help in discussing these studies and their results. Also, it would be better to not just list studies, but find some commonality amongst them.

Line 87: Throughout it says PM levels, but it should actually be PM concentrations. Need to fix this throughout the paper.

Line 88 - 92: This sentence is unclear and needs clarification.

Methods:

Line 98 - 101: This doesn't seem like the correct place for this statement. Also, should make it clearer that if cities did not collect the correct health data they were excluded from the analysis in these sentences.

Line 102 - 104: Should say the years of analysis differed by city. Also, it's not clear what is meant by 3-year data records, health data?

Figure 1: this map is confusing. Unable to clearly distinguish where the cities are located because so many dots. May be clearer to remove all names of provinces and only have dots for the cities. If you still want to detail the most populous cities it can be done in a supplemental table or have the size of the dots change depending on population of the city. Do the size of the dots mean something?

Study Population:

need to specify the age range examined because in the discussion it says adults, but not defined in the methods section.

Line 121: Should be legal instead of legally.

Statistical analysis:

Line 162: what sensitivity analyses? if discussed later should note that.

Line 168 - 173: This discussion on the concentration-response relationship is unclear and could use a revision. It should be concentration versus exposure response. The last few lines of the paragraph needs additional explanation, it's unclear how 5 coefficients were examined and what the variance-covariance matrix represents. Additional explanation is needed to clarify how statements about the shape of the C-R curve where made.

Line 175: Would be better to say something like: We examined effect modification of the relationship between short-term PM exposure and pneumonia hospital admissions in analyses...

Line 178: The part about annual average PM levels is unclear and needs additional explanation. Actually, the whole discussion on how effect modification was examined should be reconsidered. The approach used in the paper is confusing. It would be easier to interpret results if the authors examined effect modification based on the distribution of the factor examined across cities. For example, like the NMMAPS or APHENA studies where they examined how the risk of mortality changed when moving from the 25th to 75th percentile of the factor examined.

Sensitivity Analyses:

Line 185 - 186: Clarify, does this mean the authors compared associations between cities with 3 years of data versus 4 years of data?

Line 187: Prior to the sentence starting "Fourth" : I would add a sentence prior to these last two analyses to note, that model specification was assessed using these two analyses.

Line 188: The sentence starting "Fifth", would be beneficial to start the sentence by saying: to examine whether we appropriately specified the regression model used in this analysis, we...

Results:

Line 216: should clarify whether single day lags or the average of 0-2 day.

Line 219: at the lag days examined instead of at different lag days.

Line 223: Separate paragraph should start here to talk about effect modification results.

Line 224: better to present the risk estimates and confidence intervals than to focus on P values.

Line 228 - 230: I don't recall this being mentioned earlier. How were these delineations defined? Need additional explanation.

Table 3: this table could be simplified by having the CIs directly next to the percent change. Don't need the P value column.

Table 4: it's really % increase, not % change, replace PC with % Increase

Line 240: based off the figures I wouldn't call the curves linear across the full range of concentrations examined. There is actually some evidence of supralinearity where the curve plateaus at some higher concentration. Granted those concentrations are above 100, but it tends to be consistent with some of the evidence from the Integrated Exposure Response Function. See Burnett et al. (2014) - EHP. This statement "in line with reports from other studies" is not entirely correct because it makes it sound like the other studies also examined pneumonia when there are mortality studies being cited. Could say instead that "this evidence of linearity is consistent with studies examined short-term PM exposure and other outcomes."

Line 245: should be concentration-response

Line 250: From this point forward, It's not clear how this analysis was done. I'm surprised the authors did not examine the distribution across all cities of these variables and then examine how the risk estimate may change when going from the 25th percentile to the 75th percentile of the factor being examined. Additionally, there is no discussion of the influence of annual PM concentrations on the results even though it is in the table.

Line 268 - 269: Again, too much reliance on statistical significance, instead focus on whether the associations remained positive. Please revise to reflect this.

Line 269 - 271: I'm not following this discussion on C-R here. It's not clear. Also, I would not use the word threshold, you've really identified an inflection point where the risk changes due to non-linearities in the C-R curve.

Line 272 - 274: Similar comment as to above, group the model specification analyses together and note whether the results were consistent with the base model.

Line 277: clarify minor effects. Slightly attenuated, remained relatively unchanged?

Discussion:

Line 283: need to specify age range examined in the methods section.

Line 284: "acute effects", this is not correct, it's not the acute effects, but to investigate whether short-term PM exposures contribute to increases pneumonia hospital admissions.

Line 287: PM2.5 and PM10?

Line 288: short-time course? from studies examining short-term PM exposures.

Line 289: "for even less than a day", not sure how this is being deduced. We don't actually know this since the exposures being assigned are 24-h avg exposures and not hourly exposures.

Line 291: the evidence examining sub-daily exposures is very limited and this study cannot inform that since the authors are focusing on 24-h avg exposures. An association at lag 0 only indicates an immediate effect, it cannot inform that the effect is due to exposures of a shorter duration because the exposure metric used was 240h average, not individual hour values. Please revise.

Line 291 - 292: bringing in sub-daily cardiovascular results does not make sense, and again this study cannot inform sub-daily exposures because it is focusing on 24- avg exposures.

Line 292 - 294: a 0-2 lag is still considered an immediate effect. Delayed would be if you say associations at lag 2-5 days. Remove delayed and keep cumulative effect.

Line 296: Please see earlier comment on this characterization of the C-R curve.

Line 297 - 299: this is not the correct interpretation. if a health benefit per ton analysis was occurring sure, this statement could be made, but the analysis is focusing on risk estimates so the proper intepretation is that risk is higher and concentrations lower than some value and they plateau above a value.

Line 299 - 300: This sentence is unclear and I don't think it's needed.

Line 302 - 304: Jumping to harvesting seems like a stretch. It could just be that PM2.5 has a greater health risk than PM10 and that PM10 is representing exposure to more particles in the coarse range (PM10-2.5) than fine particles. I don't think the harvesting statement makes sense based on this study. If the authors examined a longer duration of lags to support this statement then it would be a different story.

Line 305: But a threshold analysis was not conducted, the analysis only focused on assessing linearity. I don't think this statement is needed. If anything the authors could state the point at which confidence intervals expand and there is less certainty in the shape of the C-R at both lower and higher PM concentrations.

Line 307: I'd remove statements about health benefits because this paper is not focusing on that and it is a little confusing.

Line 309 - 311: Seems odd to cite a respiratory mortality study instead of a pneumonia HA one here. I would revisit other penumonia studies to confirm the findings of this study instead of mortality.

Line 312-313: I think this is a better statement to make about differences between PM10 and PM2.5 versus the statement above about surface area. The makeup of the particles themselves is much different.

Line 314: I would delete this sentence, especially since there is no citation.

Line 321-322: This paragraph is on the right tract, but it needs more at the end, the PM-metals part really doesn't fit in with the rest of the paragraph.

Line 324-325: I couldn't access the supplement. Did they stay positive? Need to better convey the results versus relying on statistical significance.

Line 325: From this point on, Too much reliance on statistical significance in this whole paragraph.

Line 330: Need to specify PM2.5.

Line 352-355: this needs additional explanation. Unclear what the authors are trying to convey.

Reviewer #3: This study investigates short term effects of air pollution on risk of pneumonia hospital admissions. This is a well conducted study and well written. Some issues need to be clarified, as outlined below.

Line 121: Can the authors remind the reader of the coverage rate of UEBMI?

Can authors elaborate further on using hospital admissions and not including emergency room visits. Hospital admissions will include the more severe cases only.

Statistical analysis section: Can authors provide a bit more details on how the degrees of freedom were chosen for the variables that were used in natural cubic splines? I suppose they looked at model fit, but this should be further explained.

Line 228: Authors refer to the online supplement when further stratifying regions, but can they provide a description of those findings in the text?

Can authors add the variables that were used for adjustment under each table? This is mentioned in the Statistical analysis section, but should also be reported under each table. For example, is Table 5 adjusted for temperature and humidity, in addition to looking at meta-regression models by city-level annual average temperature and humidity?

Did authors consider looking at multi-pollutant models (i.e. adjusting for more than two pollutants in their models)?

Line 276-277: Where are those results reported? Also, why using a cut-off of 20% for health insurance coverage?

Line 298: Can authors provide more discussion on why risks would be higher per unit increase at low levels? This is a very important topic going forward in air pollution epidemiology.

Line 316: Possible mechanistic pathways are outlined, but pneumonia is an infection and it seems there are missing discussion points on how air pollution could be increasing the risk of getting this type of infection. Further discussion is needed in this paragraph.

Can authors describe why they did not adjust for influenza epidemics in their models? This is a common approach when looking at short term air pollution-respiratory disease associations.

Concentration-response curves: Can authors add the 24-hour standards for PM10 and PM2.5 as vertical dashed lines in the graphs? This would give a sense of what standards are currently and how effects are being observed in regards to those standards.

Reviewer #4: The manuscript assesses the relationship between short-term exposure to particulate matter (PM) and hospital admissions for pneumonia in China for a 4-years period (2014-2017). The key finding is that PM is associated with increased pneumonia admissions, especially in the elderly population. Moreover, effects are stronger in cities with higher temperatures and relative humidity.

The study bases on a huge dataset with data from 184 cities across China, representing differing climates across the country. Further, the question and the results are of interest to general practitioners and pulmonary specialists as well as public health officials; the topic is quite timely as air pollution is still affecting populations worldwide, even at places with low air pollution levels. The manuscript is well written and the statistical approach is appropriate. However, I have some concerns:

Major comments:

* My biggest concern is about the selection of pneumonia as the only outcome. For example, it would be interesting to look also at respiratory infections or influenza. Further, individuals with certain chronic medical conditions, such as COPD - which includes emphysema and chronic bronchitis - and asthma are especially at risk for pneumonia. Therefore, I strongly suggest including further outcomes.

* Statistical analysis, lines 152-161: As the association between PM and pneumonia might be confounded by influenza, I suggest adjusting for influenza in a sensitivity analysis.

* Statistical analysis, lines 163-166: Why did the authors choose lags of 0, 1 and 2 days? Previous studies have seen more delayed effects of PM, especially with respiratory diseases.

* Discussion, lines 324-335: The authors could also assess effect modification by other pollutants (NO2 or CO) using categories for pollutant levels.

* Discussion, lines 337-343: The authors need to provide a more detailed discussion and potential mechanisms for their finding of stronger effects in cities with higher temperature and relative humidity.

Minor comments:

1. Abstract, line 31: The authors should drop "overdispersed" here as the use of a quasi-Poisson model already implies consideration of overdispersion. Further, the authors should add here that they adjusted for confounding.

2. Study sites, lines 96-104: The authors should give more details on the choice/inclusion of cities. Why, for example, is Beijing not included, although it is the capital of China? What about Shanghai? What was the original number of cities? Further, authors should provide more details on what they mean with "Cities with no information on ICD code".

3. Study population: As the study uses data from the UEBMI system, results are not transferable to the rural population. This should be acknowledged in the limitations.

4. Lines 124-125: Please provide an ICD code here.

5. Environmental data: The authors state that they obtained hourly air pollution concentrations. How did they calculate 24-hour averages? Was the exclusion of "days with missing monitoring measurements" done for cities with only one monitoring site?

6. Lines 178-179: Please be more specific here and replace "weather conditions" with "air temperature and relative humidity".

7. Results: The authors need to give more details on the cities included. For example, what is the population of the cities? What is the UEBMI coverage per city? What is the distribution of daily hospital admissions, air pollutants and meteorological variables.

8. Figure 2: Please provide units for the axes.

9. Table 5: Please mention the increment for the effects estimates.

10. Line 272: Why did the authors choose 100 μg/m³ as cut-off? The decline in the exposure-response function starts earlier.

Any attachments provided with reviews can be seen via the following link:

[LINK]

2 Nov 2019

Submitted filename: Response to Editors and Reviewers.docx

Click here for additional data file.

19 Nov 2019

Dear Dr. Hu,

Thank you very much for re-submitting your manuscript "Ambient Particulate Matter Pollution and Adult Hospital Admissions for Pneumonia in Urban China: A national time-series analysis" (PMEDICINE-D-19-02197R1) for review by PLOS Medicine.

I have discussed the paper with my colleagues and the academic editor and it was also seen again by 2 reviewers. I am pleased to say that provided the remaining editorial and production issues are dealt with we are planning to accept the paper for publication in the journal.

The remaining issues that need to be addressed are listed at the end of this email. Any accompanying reviewer attachments can be seen via the link below. Please take these into account before resubmitting your manuscript:

[LINK]

Our publications team (plosmedicine@plos.org) will be in touch shortly about the production requirements for your paper, and the link and deadline for resubmission. DO NOT RESUBMIT BEFORE YOU'VE RECEIVED THE PRODUCTION REQUIREMENTS.

***Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.***

In revising the manuscript for further consideration here, please ensure you address the specific points made by each reviewer and the editors. In your rebuttal letter you should indicate your response to the reviewers' and editors' comments and the changes you have made in the manuscript. Please submit a clean version of the paper as the main article file. A version with changes marked must also be uploaded as a marked up manuscript file.

Please also check the guidelines for revised papers at http://journals.plos.org/plosmedicine/s/revising-your-manuscript for any that apply to your paper. If you haven't already, we ask that you provide a short, non-technical Author Summary of your research to make findings accessible to a wide audience that includes both scientists and non-scientists. The Author Summary should immediately follow the Abstract in your revised manuscript. This text is subject to editorial change and should be distinct from the scientific abstract.

We expect to receive your revised manuscript within 1 week. Please email us (plosmedicine@plos.org) if you have any questions or concerns.

We ask every co-author listed on the manuscript to fill in a contributing author statement. If any of the co-authors have not filled in the statement, we will remind them to do so when the paper is revised. If all statements are not completed in a timely fashion this could hold up the re-review process. Should there be a problem getting one of your co-authors to fill in a statement we will be in contact. YOU MUST NOT ADD OR REMOVE AUTHORS UNLESS YOU HAVE ALERTED THE EDITOR HANDLING THE MANUSCRIPT TO THE CHANGE AND THEY SPECIFICALLY HAVE AGREED TO IT.

Please ensure that the paper adheres to the PLOS Data Availability Policy (see http://journals.plos.org/plosmedicine/s/data-availability), which requires that all data underlying the study's findings be provided in a repository or as Supporting Information. For data residing with a third party, authors are required to provide instructions with contact information for obtaining the data. PLOS journals do not allow statements supported by "data not shown" or "unpublished results." For such statements, authors must provide supporting data or cite public sources that include it.

If you have any questions in the meantime, please contact me or the journal staff on plosmedicine@plos.org.

We look forward to receiving the revised manuscript by Nov 26 2019 11:59PM.

Sincerely,

Caitlin Moyer, Ph.D.

Associate Editor

PLOS Medicine

plosmedicine.org

------------------------------------------------------------

Requests from Editors:

1.Prospective Analysis Plan: Thank you for including your pre-specified statistical analysis plan (S1 Appendix). If your statistical analysis plan was developed prospectively, please include text in the document to indicate the date the plan was developed.

2. Thank you for your response to Reviewer 2, point 10. The sentence "Cities with only 1-year hospital admission records were excluded owing to the feasibility of model fit" is not clear. Please clarify what is meant by "feasibility of model fit".

3. Thank you for your response to Reviewer 2, point 16. However, the text "Variance-covariance matrix is a matrix whose element in the i, j position is the covariance between the i-th and j-th elements of a random vector. The regression coefficients derived from the main model with a B-spline function for PM" is missing from the manuscript. Please update the manuscript text to match your response.

4. Thank you for your response to Reviewer 2, point 18. However, please update the text of the methods as described in your response.

5. Thank you for your response to reviewer 2, point 21. However, please update the text as you describe in your response.

6. Thank you for your response to reviewer 2, point 27. However, please present both 95% CIs and p values for your results, according to PLOS Medicine policy. Please also do this throughout the manuscript.

7. Thank you for your response to Reviewer 2, point 44. However, the statement "the curves plateau at high concentrations, indicating that hte risk of pneumonia admission is lower at high concentrations." is confusing. Does this sentence mean that the increase in risk of pneumonia admission is greater at lower compared to high concentrations? If so, please clarify.

8.Thank you for your response to reviewer 2, point 47. However, deleting the term "harvesting" does not adequately address the reviewers point that there may be different health risks associated with different PM sizes, or other explanations.

9. Title: Please revise the title to: “Ambient Particulate Matter Pollution and Adult Hospital Admissions for Pneumonia in Urban China: A national time-series analysis for 2014 through 2017.” or similar, as we would prefer to have the dates of your study included in the title to provide time relevance.

10. Abstract: Background: Line 29-30: Please revise to, “We aimed to examine the association between PM levels and hospital admissions for pneumonia in Chinese adults.” or similar, to reflect that your study did not measure individuals’ exposures to PM.

11. Abstract: Conclusions: Line 51-52: Please revise to, “Our findings suggest that there are significant short-term associations between ambient PM levels and increased hospital admissions for pneumonia in Chinese adults.” or similar to reflect that you were not evaluating individual exposures.

12. Abstract: Conclusions: Please revise to, “These findings support the rationale that further limiting PM concentrations in China may be an effective strategy to reduce pneumonia-related hospital admissions.” or similar.

13. Author Summary: “Why was this study done?”: Here and throughout the text of the manuscript, please refer to low or middle income countries rather than "developing countries". Please refer to high income countries rather than "developed" countries.

14. Author Summary: “What did the authors do and find?”: Please combine the first two bullet points into a single point such as: “We conducted a nationwide time-series analysis using data on more than 4.2 million hospital admissions for pneumonia in 184 cities in China between 2014 and 2017 to estimate city-specific, and national and regional average associations between ambient PM pollution and pneumonia hospitalizations."

15. Author Summary: “What did the authors do and find?”: Please replace “exposures to” with “increases in” for the third bullet point.

16. Author Summary: “What do these findings mean?”: Please revise the first bullet point to “...short term associations of PM levels with hospital admissions…”

17. Author Summary: “What do these findings mean?”: Please revise the second bullet point to: “Our findings support the rationale for further limiting PM concentrations in low-middle income countries. -or similar.

18. Introduction: Line 115-117: Here, and throughout the text of the manuscript, please replace “developed countries” with “high income countries”, and “developing countries” with “low or middle income countries”.

19. Introduction: Line 123-124: Please revise to, “In this study, we examined the short term associations between concentrations of ambient PM pollution and daily hospital admissions for pneumonia in adults in China between 2014 and 2017.” to make the goal of the study clear.

20. Methods: Line 133-134: “Cities with no information on disease diagnosis recorded in the database were also excluded” Following the comments of Reviewer 4, this may be a good place to clarify some cities excluded, including Beijing and Shanghai.

21. Methods: Lines 222-223: Please revise to: “We examined effect modification of the relationship between short-term ambient PM concentrations and pneumonia hospital admissions in analyses by sex, age, and region…”

22. Methods: Lines 246-249: For statistical testing, please specify the significance level used (e.g., P<0.05, two-sided) and the statistical test used to derive a p value.

23. Results: Line 265 and Table 2 title: Please change “exposure variables” to “environmental” or “air quality” variables.

24. Results: Line 265-269: Please provide p values and accompanying r values for the reported correlations between environmental/air quality variables.

25. Results: Line 276-282: Please provide 95% CI and p values for the percent increases of hospital admissions for pneumonia with increases in concentrations of air quality measures.

26. Results: Lines 288-291: For the comparisons between males and females, please provide the specific p values, and also provide values for each comparison (PM 2.5 and PM 10).

27. Results: Lines 293-297: Please provide p values associated with differences between the regions for each PM 2.5 and PM 10.

28. Results: Lines 291-294: Please provide p values associated with comparisons between age groups.

29. Results: 346-352: Please provide p values associated with the 95% CIs reported for these analyses.

30. Results: 359-360, and throughout the manuscript: Please provide the specific p value in place of [NS or P>.05].

31. Discussion: Line 374: Please revise this to “...higher pneumonia-related hospital admissions…”

32. Discussion: Line 376: Please revise this to: “...investigate whether short-term changes in ambient PM concentrations are related to increases pneumonia hospital admissions…”

33. Discussion: Line 379: Please revise this to: “We observed increased risk of hospital admissions for pneumonia in association with both PM2.5 and PM10…”

34. Discussion: Line 411: Please revise this to: “...findings on the short-term association between ambient PM concentrations and pneumonia hospitalizations…”

35. Discussion: Line 442: Please revise this to: “This study provided detailed estimates of the risk of PM-associated hospital admissions for pneumonia through the use of data…”

36. Discussion: Line 466-468: Please revise this to: In summary, we found that short-term elevations in PM were associated with increased pneumonia- related hospital admissions in Chinese adults. Our findings support the rationale to further limit PM concentrations in China.

37. Table 1, 2, 3, 4 and 5: Please define abbreviations for “PM2.5” and “PM10” in the legends.

38. Table 2: Please provide p values associated with all statistical comparisons, include actual values for p<0.01, unless p<0.001.

39. Table 3: Please provide both the 95% CIs and p values for these results.

40. Table 3: Please also provide the results for the unadjusted analyses.

41. Table 4: Please also provide the results for the unadjusted analyses.

42. Table 4: Please provide p values to accompany the 95% CI for all comparisons shown.

43. Table 5: Please replace “P” with “P value”

44. Figure 2: To facilitate comparisons, please display the two graphs on the same y-axis scale, or please note in the legend that the y-axes of the two graphs are scaled differently.

45. S2 Table, S3 Table, S4 Table, S5 Table: Please also provide results for unadjusted analyses. Please provide p values to accompany the 95% CIs for these analyses.

Comments from Reviewers:

Reviewer #3: Authors have made tremendous efforts to address all comments. I have no issues in recommending publication.

Reviewer #4: The manuscript assesses the relationship between short-term exposure to particulate matter (PM) and hospital admissions for pneumonia in China for a 4-years period (2014-2017). The key finding is that PM is associated with increased pneumonia admissions, especially in the elderly population. Moreover, effects are stronger in cities with higher temperatures and relative humidity.

The study bases on a huge dataset with data from 184 cities across China, representing differing climates across the country. Further, the question and the results are of interest to general practitioners and pulmonary specialists as well as public health officials; the topic is quite timely as air pollution is still affecting populations worldwide, even at places with low air pollution levels. The manuscript is well written and the statistical approach is appropriate.

I appreciate the effort the authors have spent on addressing the concerns raised in the initial review. The modifications clearly have improved the quality and clarity of the manuscript.

Minor comments:

1. Study sites, lines 96-104: Although the authors now give more details on the inclusion criteria, it is still unclear why, for example, Beijing is not included, although it is the capital of China? Same is true for Shanghai.

2. Lines 390-392: The authors speculate that "The leveling-off at high concentrations might be speculated by that people vulnerable to PM exposure may have developed symptoms and sought treatment before PM concentrations reached high concentrations". Other reasons could be that people avoid spending time outdoors or start wearing facemasks.

Any attachments provided with reviews can be seen via the following link:

[LINK]

28 Nov 2019

Submitted filename: Response to Editors and Reviewers.docx

Click here for additional data file.

4 Dec 2019

Dear Prof. Hu,

On behalf of my colleagues and the academic editor, Dr. Aziz Sheikh, I am delighted to inform you that your manuscript entitled "Ambient Particulate Matter Pollution and Adult Hospital Admissions for Pneumonia in Urban China: A national time-series analysis for 2014 through 2017" (PMEDICINE-D-19-02197R2) has been accepted for publication in PLOS Medicine.

PRODUCTION PROCESS

Before publication you will see the copyedited word document (in around 1-2 weeks from now) and a PDF galley proof shortly after that. The copyeditor will be in touch shortly before sending you the copyedited Word document. We will make some revisions at the copyediting stage to conform to our general style, and for clarification. When you receive this version you should check and revise it very carefully, including figures, tables, references, and supporting information, because corrections at the next stage (proofs) will be strictly limited to (1) errors in author names or affiliations, (2) errors of scientific fact that would cause misunderstandings to readers, and (3) printer's (introduced) errors.

If you are likely to be away when either this document or the proof is sent, please ensure we have contact information of a second person, as we will need you to respond quickly at each point.

PRESS

A selection of our articles each week are press released by the journal. You will be contacted nearer the time if we are press releasing your article in order to approve the content and check the contact information for journalists is correct. If your institution or institutions have a press office, please notify them about your upcoming paper at this point, to enable them to help maximize its impact.

PROFILE INFORMATION

Now that your manuscript has been accepted, please log into EM and update your profile. Go to https://www.editorialmanager.com/pmedicine, log in, and click on the "Update My Information" link at the top of the page. Please update your user information to ensure an efficient production and billing process.

Thank you again for submitting the manuscript to PLOS Medicine. We look forward to publishing it.

Best wishes,

Caitlin Moyer, Ph.D.

Associate Editor

PLOS Medicine

plosmedicine.org

Table 1

Summary statistics of citywide annual-mean hospital admissions for pneumonia, air pollutants, and weather conditions in 184 Chinese cities, 2014–2017.

Variables	Mean ± SD	Minimum	Percentile			Maximum
			25th	50th	75th
Daily pneumonia admissions	26 ± 58	1	5	9	19	376
PM2.5 (μg/m3)	50 ± 19	15	38	49	59	102
PM10 (μg/m3)	89 ± 40	28	66	84	103	193
Temperature (°C)	14 ± 5	0	11	15	18	24
Relative humidity (%)	68 ± 12	34	59	70	78	92

Abbreviations: PM2.5, particulate matter ≤2.5 μm in aerodynamic diameter; PM10, particulate matter ≤10 μm in aerodynamic diameter

Table 2

Spearman correlation coefficients among the environmental variables in 184 Chinese cities, 2014–2017.

Variables	PM2.5	PM10	SO2	NO2	CO	O3	Temp	RH
PM2.5	1.00	0.90*	0.56*	0.64*	0.61*	−0.02 (P = 0.452)	−0.26*	−0.08*
PM10	―	1.00	0.59*	0.64*	0.57*	0.05 (P = 0.215)	−0.26*	−0.28*
SO2	―	―	1.00	0.55*	0.52*	−0.09*	−0.38*	−0.33*
NO2	―	―	―	1.00	0.54*	−0.13*	−0.31*	−0.12*
CO	―	―	―	―	1.00	−0.21*	−0.28*	−0.04*
O3	―	―	―	―	―	1.00	0.53*	−0.23*
Temp	―	―	―	―	―	―	1.00	0.26*
RH	―	―	―	―	―	―	―	1.00

*P < 0.001.

Abbreviations: PM2.5, particulate matter ≤2.5 μm in aerodynamic diameter; PM10, particulate matter ≤10 μm in aerodynamic diameter; RH, relative humidity; Temp, temperature

Table 3

National-average percentage increase with 95% CI in daily hospital admissions for pneumonia associated with a 10 μg/m3 increase in PM2.5 and PM10 concentrations using different lag days in 184 Chinese cities, 2014–2017.

Lag day	Percentage increase (95% CI)	P value
Unadjusted analyses
PM2.5
Lag 0	0.14 (0.04–0.24)	0.007
Lag 1	0.11 (0.01–0.21)	0.027
Lag 2	0.15 (0.06–0.25)	0.002
Lag 0–2	0.23 (0.08–0.37)	0.002
PM10
Lag 0	0.11 (0.04–0.18)	0.001
Lag 1	0.06 (−0.01–0.13)	0.093
Lag 2	0.10 (0.02–0.17)	0.008
Lag 0–2	0.15 (0.04–0.25)	0.005
Adjusted analyses*
PM2.5
Lag 0	0.23 (0.14–0.33)	<0.001
Lag 1	0.15 (0.05–0.25)	0.004
Lag 2	0.11 (0.01–0.21)	0.039
Lag 0–2	0.31 (0.15–0.46)	<0.001
PM10
Lag 0	0.18 (0.12–0.25)	<0.001
Lag 1	0.08 (0.02–0.15)	0.015
Lag 2	0.04 (−0.02–0.11)	0.218
Lag 0–2	0.19 (0.11–0.30)	<0.001

*Adjusted for temperature, relative humidity, calendar time, day of the week, and public holiday.

Abbreviations: PM2.5, particulate matter ≤2.5 μm in aerodynamic diameter; PM10, particulate matter ≤10 μm in aerodynamic diameter

Table 4

National-average percentage increase (PI) with 95% CI in daily hospital admissions for pneumonia associated with a 10 μg/m3 increase in PM2.5 and PM10 concentrations (lag 0–2), stratified by sex, age, and geographical region.

Variables	PM2.5		PM10
	PI (95% CI)	P value#	PI (95% CI)	P value#
Unadjusted analyses
Sex		0.401		0.279
Male	0.26 (0.10–0.41), P = 0.001		0.18 (0.07–0.28), P = 0.001
Female	0.16 (−0.01–0.33), P = 0.074		0.09 (−0.03–0.21), P = 0.153
Age, years
18–64	0.04 (−0.12–0.21), P = 0.605	1 (Ref.)	0.01 (−0.11–0.13), P = 0.870	1 (Ref.)
65–74	0.28 (0.10–0.46), P = 0.003	0.051	0.18 (0.05–0.30), P = 0.006	0.060
≥75	0.46 (0.29–0.63), P < 0.001	<0.001	0.33 (0.21–0.45), P < 0.001	<0.001
Region		0.007		<0.001
South	0.46 (0.21–0.72), P < 0.001		0.45 (0.27–0.63), P < 0.001
North	0.05 (−0.11–0.22), P = 0.529		−0.06 (−0.17–0.05), P = 0.313
Adjusted analyses*
Sex		0.887		0.422
Male	0.31 (0.12–0.49), P < 0.001		0.20 (0.13–0.28), P < 0.001
Female	0.33 (0.13–0.53), P < 0.001		0.15 (0.05–0.25), P = 0.001
Age, years
18–64	0.24 (0.04–0.44), P = 0.016	1 (Ref.)	0.10 (0.02–0.19), P = 0.017	1 (Ref.)
65–74	0.35 (0.12–0.58), P < 0.001	0.479	0.20 (0.10–0.30), P < 0.001	0.126
≥75	0.42 (0.23–0.62), P < 0.001	0.201	0.32 (0.22–0.39), P < 0.001	<0.001
Region		0.276		0.045
South	0.40 (0.16–0.65), P < 0.001		0.35 (0.16–0.54), P < 0.001
North	0.23 (0.04–0.43), P = 0.022		0.12 (0–0.24), P = 0.046

*Adjusted for temperature, relative humidity, calendar time, day of the week, and public holiday.

#P value obtained from Z-test for the difference between the two risk estimates derived from subgroup analysis.

Abbreviations: PM2.5, particulate matter ≤2.5 μm in aerodynamic diameter; PM10, particulate matter ≤10 μm in aerodynamic diameter; Ref., reference

Table 5

Meta-regression results of the modification effects of city-level characteristics on the associations between PM2.5 and PM10 and hospital admissions for pneumonia in 184 Chinese cities, 2014–2017.

City-level characteristics	Percentage increase	95% CI	P value
PM2.5
PM2.5 (10 μg/m3)	−0.019	−0.113–0.075	0.688
NO2 (10 μg/m3)	−0.030	−0.208–0.148	0.737
CO (1 mg/m3)	0.177	−0.300–0.657	0.466
Temperature (°C)	0.031	−0.009–0.071	0.129
Relative humidity (%)	0.007	−0.009–0.023	0.378
GDP per capita (¥10,000)	0.007	−0.073–0.087	0.864
Coverage rate (%)	−0.010	−0.106–0.087	0.835
PM10
PM10 (10 μg/m3)	0	−0.064–0.064	0.997
NO2 (10 μg/m3)	−0.011	−0.135–0.113	0.861
CO (1 mg/m3)	0.153	−0.180–0.488	0.367
Temperature (°C)	0.031	0.003–0.058	0.026
Relative humidity (%)	0.011	0–0.022	0.045
GDP per capita (¥10,000*)	−0.012	−0.066–0.042	0.655
Coverage rate (%)	−0.002	−0.008–0.005	0.665

*¥10,000 = £1,169; $1,456; €1,377.

Abbreviations: GDP, gross domestic product; PM2.5, particulate matter ≤2.5 μm in aerodynamic diameter; PM10, particulate matter ≤10 μm in aerodynamic diameter