Content uploaded by Yaohua Tian
Author content
All content in this area was uploaded by Yaohua Tian on Jul 18, 2021
Content may be subject to copyright.
Environment, Green Space and Pollution
Fine particulate air pollution and adult hospital
admissions in 200 Chinese cities: a time-series
analysis
Yaohua Tian,
1
Hui Liu,
1,2
Tianlang Liang,
3
Xiao Xiang,
1
Man Li,
1
Juan Juan,
1
Jing Song,
1
Yaying Cao,
1
Xiaowen Wang,
1
Libo Chen,
3
Chen Wei,
3
Pei Gao
1,4†
and Yonghua Hu
1
*
†
1
Department of Epidemiology and Biostatistics, School of Public Health, Peking University, Beijing,
China,
2
Medical Informatics Center, Peking University, Beijing, China,
3
HealthCom Data Technology Co.
Ltd, Beijing, China and
4
Key Laboratory of Molecular Cardiovascular (Peking University), Ministry of
Education, Beijing, China
*Corresponding author. Department of Epidemiology and Biostatistics, School of Public Health, Peking University, No. 38
Xueyuan Road, Beijing 100191, China. E-mail: yhhu@bjmu.edu.cn
†
These authors contributed equally to this work.
Editorial decision 1 May 2019; Accepted 10 May 2019
Abstract
Background: The association between short-term exposure to ambient fine particulate
matter (PM
2.5
) and morbidity risk in developing countries is not fully understood. We
conducted a nationwide time-series study to estimate the short-term effect of PM
2.5
on
hospital admissions in Chinese adults.
Methods: Daily counts of hospital admissions for 2014–16 were obtained from the
National Urban Employee Basic Medical Insurance database. We identified more than
58 million hospitalizations from 0.28 billion insured persons in 200 Chinese cities for sub-
jects aged 18 years or older. Generalized additive models with quasi-Poisson regression
were applied to examine city-specific associations of PM
2.5
concentrations with hospital
admissions. National-average estimates of the association were obtained from a
random-effects meta-analysis. We also investigated potential effect modifiers, such as
age, sex, temperature and relative humidity.
Results: An increase of 10 lg/m
3
in same-day PM
2.5
concentrations was positively associ-
ated with a 0.19% (95% confidence interval: 0.07–0.30) increase in the daily number of hospi-
tal admissions at the national level. PM
2.5
exposure remained positively associated with hos-
pital admissions on days when the daily concentrations met the current Chinese Ambient
Air Quality Standards (75 lg/m
3
). Estimates of admission varied across cities and increased
in cities with lower PM
2.5
concentrations (p¼0.044) or higher temperatures (p¼0.002) and
relative humidity (p¼0.003). The elderly were more sensitive to PM
2.5
exposure (p<0.001).
Conclusions: Short-term exposure to PM
2.5
was positively associated with adult hospital
admissions in China, even at levels below current Chinese Ambient Air Quality Standards.
V
CThe Author(s) 2019; all rights reserved. Published by Oxford University Press on behalf of the International Epidemiological Association 1142
International Journal of Epidemiology, 2019, 1142–1151
doi: 10.1093/ije/dyz106
Advance Access Publication Date: 3 June 2019
Original article
Downloaded from https://academic.oup.com/ije/article/48/4/1142/5510153 by guest on 18 July 2021
Key words: Fine particulate matter, hospital admission, time-series, China
Introduction
Epidemiological and toxicological studies have reported
the adverse health effects of short-term exposure to ambi-
ent air pollution.
1
Among air pollutants, ambient fine par-
ticulate matter (PM
2.5
, particulate matter 2.5 lmin
aerodynamic diameter) has been widely considered the
main air pollutant contributing to hazardous effects
2
and a
primary risk factor for disease. According to the Global
Burden of Disease Study, PM
2.5
led to an estimated
4.2 million premature deaths and 103.1 million disability-
adjusted life-years in 2015, with 59% of these occurring in
East and South Asia.
3
Whereas the increased mortality risk has been well
documented in both developed and developing countries
such as China,
4–9
relatively fewer studies have examined
the association of PM
2.5
with hospital admissions or other
morbidity measures. Hospitalizations can differ markedly
from death events by volume, demographics and diagnostic
composition, and greatly outnumber death events, reflect-
ing a measure of health effects caused by increases in air
pollution in a broader segment of the population.
Furthermore, hospitalization data can better test the tem-
poral pattern between short-term exposure to air pollution
and clinical presentation of disease.
10
Although a few stud-
ies have attempted to assess the association of PM
2.5
with
hospital admission,
11–16
most were conducted in developed
countries and few research data at country-level have been
generated in developing countries, despite their much
higher PM
2.5
levels.
In 2013, China began monitoring PM
2.5
levels and re-
leasing real-time measurements. To date, only a few studies
have evaluated the short-term effects of PM
2.5
on morbid-
ity risk in China, with most conducted in one city or a few
cities.
17,18
The findings derived from single-city studies can
be susceptible to publication bias.
19
In addition, previous
studies have demonstrated geographical heterogeneity in
the short-term effects of PM
2.5
on hospital admis-
sions.
11,20,21
Therefore, published estimates may be insuffi-
ciently representative for policymaking purposes and
setting air-quality standards at the national level.
With the establishment of a basic medical-insurance
scheme,
22
national morbidity data in China have become
available. We therefore assessed the association of PM
2.5
and hospital admissions in China between January 2014
and December 2016.
Methods
Study sites
A total of 200 cities were included in this analysis (shown
in Supplementary Figure 1, available as Supplementary
data at IJE online). These cities were selected based on the
availability of both air-pollution and health data. The total
study period of this study was from 2014 to 2016 and dif-
fered by city based on the availability of PM
2.5
data. Of
the 200 cities, 82 cities have only 2-year data and 118 cities
have 3-year data.
Data source for hospital admissions
China has achieved universal health-insurance coverage in
2011, which now has three main insurance schemes. The
Urban Employee Basic Medical Insurance (UEBMI) covers
urban employees and retired employees. The Urban
Residence Basic Medical Insurance covers urban residents,
including children, students, elderly people without previ-
ous employment and unemployed people. The New Rural
Cooperative Medical Scheme covers rural residents. The
data on city-specific hospital admissions in our study for
Key Messages
•We observed a positive association between short-term PM
2.5
exposure and hospital admissions in 200 Chinese
cities.
•PM
2.5
exposure was associated with hospital admissions even at levels below current Chinese Ambient Air Quality
Standards.
•The association was more evident in cities with lower PM
2.5
levels or higher air temperature and relative humidity.
•First investigation on short-term PM
2.5
exposure and hospital admissions at the national level in China. Our findings
will be useful in informing Chinese policies on ambient-air-quality standards.
International Journal of Epidemiology, 2019, Vol. 48, No. 4 1143
Downloaded from https://academic.oup.com/ije/article/48/4/1142/5510153 by guest on 18 July 2021
2014–16 were obtained from UEBMI, administered by
China’s Ministry of Human Resources and Social Security.
In 2016, the database included 0.28 billion enrollees in 31
provinces, accounting for an estimated 20% of the total
population in China (1.38 billion). Supplementary Table 1,
available as Supplementary data at IJE online, lists the
number of people enrolled in UEBMI in the database, the
number of residents and the coverage of population by
UEBMI in each city in 2016. To receive reimbursement, a
claim for a billable medical service must be submitted on a
standardized electronic form. Each billing claim includes
basic demographic information (age and sex), the date of
health service, treatment and diagnosis. Hospital admis-
sions of individuals aged less than 18 years were too few
and thus were excluded from this study. Considering the
substantial differences in levels and characteristics of air
pollution, weather patterns and geographical conditions
between southern and northern China,
23,24
we grouped the
cities by region (north or south) following the Huai River–
Qinling Mountains line.
Air-pollution and meteorological data
We obtained hourly PM
2.5
concentration data from the
National Air Pollution Monitoring System, which is ad-
ministered by China’s Ministry of Environmental
Protection. There are 1–17 monitoring stations in each
city. To fulfil the quality-control and quality-assurance
programmes mandated by the Chinese government, all
monitoring stations must upload data of real-time hourly
concentrations of criteria air pollutants into the system,
providing reliable and comparable measurements between
stations.
25
We obtained the daily mean concentrations for
PM
2.5
averaged across all operational monitoring stations
in each city.
9
To allow adjustment for other air pollutants,
we also acquired data on sulphur dioxide (SO
2
), nitrogen
dioxide (NO
2
), carbon monoxide (CO) and ozone (O
3
)
from the same sources. Air-pollution data obtained from
this monitoring system have been used extensively to eval-
uate the health effects of air pollution both regionally and
nationally.
9,26,27
Daily mean air temperature and relative
humidity for each city were extracted from the China
Meteorological Data Sharing Service System (http://data.
cma.cn/).
Statistical analysis
National- and regional-average associations of PM
2.5
pol-
lution and hospital admissions were estimated by a two-
stage approach that was used in previous studies.
14,28
Briefly, we obtained city-specific estimates of PM
2.5
in the
first stage using a generalized additive model with quasi-
Poisson regression. We used a natural cubic smooth func-
tion with respect to calendar time with 7 degrees of free-
dom (df) per year to adjust for seasonality and long-term
trends, such as influenza epidemics.
9,11,29
We also con-
trolled for the non-linear and lagged effects of weather
conditions on the risk of admission using natural spline
functions of 3-day moving average temperatures (6 df) and
relative humidity (6 df).
30
Finally, we incorporated indica-
tor variables for day of the week and public holidays to ac-
count for the difference in baseline admissions for each
day. To explore the temporal association of PM
2.5
and
hospital admissions, we fitted the models with different lag
structures from the current day (lag day 0) to 5 lag days
(lag day 5). We also estimated the association with 6-day
(lag days 0–5) moving average PM
2.5
concentrations. To
test whether there was evidence of PM
2.5
effects on hospi-
tal admission among individuals exposed to levels below
the current daily PM
2.5
Chinese Ambient Air Quality
Standards (CAAQS), daily data were categorized into three
groups based on daily PM
2.5
concentrations (25, 25–75
and >75 lg/m
3
). Twenty-five micrograms per cubic metre
is the World Health Organization air-quality guideline for
daily PM
2.5
concentrations and 75 lg/m
3
is the Chinese
Grade II standard.
31,32
In the second stage, a random-
effects meta-analysis was used to obtain regional-average
or national-average estimates for PM
2.5
.
33
The shape of
the association between PM
2.5
levels and hospital admis-
sions was characterized following the approach described
previously.
9,34
Specifically, following the distribution of
PM
2.5
concentrations in each city, we used a cubic spline
with two knots at 60 and 150 lg/m
3
for PM
2.5
. We then
Figure 1. National-average exposure–response curve between PM
2.5
concentrations (lag day 0) and daily hospital admissions in 200 cities in
China, 2014–16. The horizontal scale is the current-day (lag day 0) fine
particulate matter (PM
2.5
) concentrations (lg/m
3
). The vertical scale can
be interpreted as the relative change from the mean effect of PM
2.5
on
hospital admission, after adjusting for temperature, relative humidity,
calendar time, day of the week and public holidays.
1144 International Journal of Epidemiology, 2019, Vol. 48, No. 4
Downloaded from https://academic.oup.com/ije/article/48/4/1142/5510153 by guest on 18 July 2021
estimated five regression coefficients of the spline function
and the 5 5 variance–covariance matrix in each city.
Finally, we applied random-effect models to combine the
city-specific components of spline estimates. The associa-
tions were estimated in subgroups of age (18–64, 65–74
and 75 years), sex (male/female) and the coverage of pop-
ulation by UEBMI (<15 and 15%).
9
The differences in
risk were tested using a meta-regression approach. We also
assessed potential effect modifications by city-level charac-
teristics using meta-regression models, including annual
mean PM
2.5
concentrations, temperature, relative humidity
and the coverage of the population by UEBMI.
9,28
City-
specific relative risk [and their confidence intervals (CIs)]
as the outcome were meta-regressed on each continuous
variable of city characteristics. In order to evaluate the po-
tential public-health impact of the effect estimate, the an-
nual reduction in hospitalizations attributable to a 10-lg/
m
3
reduction in daily PM
2.5
concentrations used in
previous studies was calculated,
9,11,35
defined as
[exp(b)1] N, where bis the national-average estimate
for an increase in PM
2.5
by 10 lg/m
3
and Nis the number
of total hospital admissions in mainland China in 2016.
Sensitivity analyses were conducted as follows: (i) two-
pollutant models with adjustment for SO
2
,NO
2
, CO and
O
3
using the same parameters from the single-pollutant
analysis; (ii) restrict the analysis in cities with only 2- and
3-year data; (iii) use of alternative df values per calendar
period (six to eight per year); (iv) we used penalized spline
functions for calendar time, temperature and relative hu-
midity; (v) excluding hospitalizations for injury, which
were identified using natural language processing in our
database. The effect estimates are presented as percentage
change and its 95% CI in hospital admissions per 10-lg/
m
3
increase in PM
2.5
concentrations. The analyses were
conducted using R version 3.2.2 (R Foundation for
Statistical Computing, Vienna, Austria) and Stata version
12 (StataCorp, College Station, TX, USA).
Results
In total, we identified 58.52 million hospital admissions
during the study period and 23.53 million of them were in
2016. The average coverage of total population by UEBMI
in these cities in 2016 was 22.8% (Supplementary Table 1,
available as Supplementary data at IJE online).
Supplementary Table 2, available as Supplementary
data at IJE online, shows the daily mean number of hospi-
tal admissions, PM
2.5
concentrations and weather condi-
tions. Over the study period, the average daily mean
[standard deviation (SD)] count of hospital admissions was
333 (197) overall, 340 (189) in southern China and 324
(206) in northern China. The daily mean (SD) PM
2.5
concentrations across all cities was 51 (34) lg/m
3
.PM
2.5
levels and weather conditions differed markedly between
southern and northern China, with lower daily mean
PM
2.5
levels and higher daily mean temperature and rela-
tive humidity in the former.
City-specific estimates of the associations between
same-day PM
2.5
levels and hospital admissions are listed in
Supplementary Table 3, available as Supplementary data
at IJE online. There was a notable heterogeneity of the
PM
2.5
-hospitalization associations across cities. Table 1
presents the national- and regional-average estimates for
the effects of PM
2.5
on hospital admissions for different lag
days. We observed immediate PM
2.5
effects (lag day 0) na-
tionally and in the southern region. Each 10-lg/m
3
increase
in PM
2.5
concentrations on lag day 0 was associated with a
0.19% (95% CI: 0.07–0.30) and 0.38% (95% CI: 0.20–
0.55) increase in hospital admissions across all cities and in
southern cities, respectively. In northern cities, PM
2.5
was
positively associated with hospitalizations only on lag day
2 (0.15% change; 95% CI: 0.04–0.27). We further
grouped cities into six geographical regions, namely East,
Middle-south, Southwest, Northwest, North and
Northeast. The regional-average estimates for the six
regions are presented in Supplementary Table 4, available
as Supplementary data at IJE online. There was a signifi-
cant heterogeneity in the PM
2.5
-hospitalization associa-
tions across different regions. The effects were more
evident in the East, Middle-south and Southwest regions.
There was a clear national-average exposure–response
association between PM
2.5
concentrations (lag day 0) and
hospital admissions (Figure 1). The curve was nonlinear,
with a sharp slope at concentrations below 50 lg/m
3
,a
moderate slope at 50–150 lg/m
3
and a relatively stable re-
sponse at concentrations above 150 lg/m
3
.Table 2 shows
the relative risks of daily hospital admission for categories
of daily PM
2.5
levels. There are health effects below the
current Chinese standard nationally and in southern
China. At the national level, using 25 lg/m
3
, we observed
relative risk of 1.011 (95% CI: 1.003–1.019) for 25–75 lg/
m
3
and 1.020 (1.006–1.034) for >75 lg/m
3
.
The total number of hospital admissions in mainland
China in 2016 was reported as 175.28 million,
147.50million of which were in public hospitals.
36
Based
on our estimate, a 10-lg/m
3
decrease in PM
2.5
concentra-
tions would have reduced total hospital admissions by 0.33
(0.12–0.53) million nationwide in 2016 (Supplementary
Table 5, available as Supplementary data at IJE online).
The effect estimate for PM
2.5
was higher among individ-
uals aged 65–74 and 75 years than among individuals
aged 18–64 years (p<0.05). No evidence was found for ef-
fect modification by sex or by the coverage of the popula-
tion (p>0.05, Table 3). We also assessed the association
International Journal of Epidemiology, 2019, Vol. 48, No. 4 1145
Downloaded from https://academic.oup.com/ije/article/48/4/1142/5510153 by guest on 18 July 2021
between the estimated PM
2.5
effects and several city-
specific characteristics such as annual average PM
2.5
con-
centrations, temperature, relative humidity and the cover-
age of the population by UEBMI using meta-regression
analyses (Table 4). PM
2.5
effects on hospital admissions
were stronger in cities with lower annual average PM
2.5
concentrations and in cities with higher annual average
temperature and relative humidity. For each 10-lg/m
3
in-
crease in PM
2.5
concentrations on lag day 0, a city with
10 lg/m
3
lower PM
2.5
concentrations, 1C higher temper-
ature and 1% higher relative humidity with respect to
another city would see an additional 0.071% (0.001–
0.141%), 0.044% (0.017–0.071%) and 0.014% (0.005–
0.023%) increase in hospital admissions, respectively. We
found no evidence for effect modification by a city’s popu-
lation coverage rate by UEBMI (p¼0.865).
In the sensitivity analysis, the risk estimates were
broadly similar in cities with only 2-year data and cities
with 3-year data (Supplementary Table 6, available as
Supplementary data at IJE online). The increase in hospital
admissions per 10-lg/m
3
increase in PM
2.5
concentrations
was 0.19% nationally, 0.21% in cities with only 2-year
data and 0.18% in cities with 3-year data. Altering the df
(6–8) per year for time trend did not substantially affect
the risk estimates. Using penalized spline functions for cal-
endar time and weather conditions had little effect on the
estimate (0.22% change; 95% CI: 0.11–0.33). However,
in the two-pollutant models, the association of PM
2.5
levels
and hospital admissions were weakened towards the null
after adjustment for SO
2
,NO
2
and CO (Table 5). The esti-
mate changed little after excluding hospitalizations for in-
jury (0.20% change; 95% CI: 0.08–0.31).
Table 1. Percentage change with 95% confidence interval (CI) in daily hospital admissions associated with 10-lg/m
3
increase in
PM
2.5
concentrations using different lag days in 200 Chinese cities by region, 2014–16
Nationwide South North
Lag PC
a
95% CI PPC
a
95% CI PPC
a
95% CI P
Lag 0 0.19 0.07 to 0.30 0.001 0.38 0.20 to 0.55 <0.001 0 –0.15 to 0.14 0.942
Lag 1 0.02 –0.08 to 0.11 0.722 –0.05 –0.21 to 0.11 0.520 0.09 –0.01 to 0.20 0.086
Lag 2 0.06 –0.05 to 0.17 0.298 –0.04 –0.23 to 0.16 0.698 0.15 0.04 to 0.27 0.007
Lag 3 0 –0.10 to 0.10 0.926 –0.03 –0.19 to 0.14 0.750 0.02 –0.12 to 0.16 0.751
Lag 4 0.10 –0.01 to 0.21 0.067 0.25 0.08 to 0.43 0.005 –0.04 –0.18 to 0.10 0.583
Lag 5 0.10 –0.01 to 0.20 0.070 0.16 –0.02 to 0.34 0.091 0.04 –0.07 to 0.14 0.478
Lag 0-5 0.25 0 to 0.50 0.044 0.35 –0.06 to 0.76 0.003 0.16 –0.13 to 0.45 0.287
a
Adjusted for temperature, relative humidity, calendar time, day of the week and public holiday.
Table 2. Relative risk of daily hospital admissions for catego-
ries of same-day PM
2.5
concentrations (lag day 0) in 200
Chinese cities by region, 2014–16
Region Relative risk
a
95% CI P
Nationwide
25 1 (Ref.)
25–75 1.011 1.003–1.019 0.008
>75 1.020 1.006–1.034 0.005
South
25 1 (Ref.)
25–75 1.021 1.010–1.031 <0.001
>75 1.034 1.016–1.053 <0.001
North
25 1 (Ref.)
25–75 0.995 0.986–1.005 0.326
>75 1.003 0.984–1.023 0.725
a
Adjusted for temperature, relative humidity, calendar time, day of the
week and public holiday.
Table 3. National-average percentage change with 95% confi-
dence interval in daily hospital admissions associated with
10-lg/m
3
increase in same-day PM
2.5
concentrations (lag day
0) in 200 Chinese cities, 2014–16, classified by sex, age and
city-specific coverage of population by the Urban Employee
Basic Medical Insurance (UEBMI)
Subgroups Percentage
change
a
95% CI P
b
Sex 0.620
Male 0.20 0.07–0.33
Female 0.16 0.03–0.29
Age, years
18–64 0.11 0.01–0.20 1 (Ref.)
65–74 0.23 0.13–0.33 0.001
75 0.36 0.26–0.47 <0.001
Coverage of population by UEBMI (%) 0.208
<15 0.11 0–0.23
15 0.24 0.07–0.42
a
Adjusted for temperature, relative humidity, calendar time, day of the
week and public holiday.
b
P-value obtained from Z-test for the difference between the two risk esti-
mates derived from subgroup analysis.
1146 International Journal of Epidemiology, 2019, Vol. 48, No. 4
Downloaded from https://academic.oup.com/ije/article/48/4/1142/5510153 by guest on 18 July 2021
Discussion
In this national time-series study, we investigated the rela-
tionship between PM
2.5
pollution and hospital admissions
in 200 representative Chinese cities, covering >58 million
hospitalizations. We found that an increase in the same-
day PM
2.5
concentrations, even at levels below the current
CAAQS, was associated with increased hospital admis-
sions. We observed effect modifications by cities’ mean
PM
2.5
levels, temperature and relative humidity. The asso-
ciation was more evident in the elderly. To our knowledge,
this is the first Chinese study to report the short-term
effects of PM
2.5
on hospital admissions on a national scale.
In the present analysis, a 10-lg/m
3
increase in the concur-
rent day PM
2.5
concentrations corresponded to a 0.19% in-
crease in hospital admissions. Generally, the magnitude of
our effect estimates was lower than in previously reported
results from multi-city or meta-analyses. For example, a re-
cent national study in the US estimated a 1.05% increase in
daily all-cause mortality in the entire Medicare population
(adults aged over 65 years) between 2000 and 2012.
37
Our
estimate was also smaller than estimates found in other
multi-city studies in the US and Europe.
4–6,38
In a meta-
analysis of time-series studies of PM
2.5
and mortality, mostly
conducted in the USA and Europe, Atkinson and colleagues
estimated a 1.04% increase in all-cause mortality in associa-
tion with a 10-lg/m
3
increase in PM
2.5
.
16
In our primary
analysis, hospital admissions for natural injury were included
due to the consideration of quantifying the total measure of
health effects. However, in the sensitivity analysis, we ex-
cluded the hospitalizations for injury and found little effect
on the results (Supplementary Table 6, available as
Supplementary data at IJE online), namely an increase from
0.19% (95% CI: 0.07–0.30) to 0.20% (95% CI: 0.08–0.31).
There are several potential explanations for the lower
estimates in this study. First, the inconsistency in the mag-
nitude of estimated effects might be partly attributable to
differences in the age groupings. Several studies focused on
an elderly population,
11,12,15
whereas this analysis covered
all adults. Previous studies have demonstrated a higher vul-
nerability to PM
2.5
exposure in the elderly
9,14,15
and our
findings among individuals aged 75 years are in agree-
ment. Second, the exposure–response curve was slightly
nonlinear with a plateauing trend at high PM
2.5
concentra-
tions. This saturation effect is consistent with the negative
effect modification of cities’ mean PM
2.5
concentrations,
as shown in the meta-regression analysis. Our findings are
supported by those of previous studies that reported a pla-
teauing trend at higher PM
2.5
concentrations.
9,39
The sta-
ble response at higher levels may result from a ‘harvesting
effect’, meaning that individuals vulnerable to PM
2.5
may
have developed symptoms and sought treatment before
PM
2.5
concentrations reached a fairly high level.
40
Third,
the weaker effects observed in our study were due, at least
in part, to the variation in the PM
2.5
composition.
Chemical components of PM
2.5
have been shown to exert
varied effects on hospital admissions.
12,21
PM
2.5
in China’s
air has a greater proportion of crustal constituents,
41
which may result in lower toxicity.
42
Finally, the differen-
ces in socio-economic status, weather patterns, geographi-
cal conditions and population susceptibility may also
partly explain the heterogeneity in the magnitude of risk
estimates.
To further compare our findings with prior reports
from China, we conducted a systematic review of short-
term effects of PM
2.5
on all-cause hospital visits or mortal-
ity in China (Supplementary eAppendix, available as
Table 4. Effects of city-level characteristics on the association between PM
2.5
and daily hospital admissions in 200 Chinese cities
using meta-regression, 2014–16
Variables Percentage change 95% CI P
PM
2.5
(10 lg/m
3
) –0.071 –0.141 to –0.001 0.044
Temperature (C) 0.044 0.017 to 0.071 0.002
Relative humidity (%) 0.014 0.005 to 0.023 0.003
Coverage of population by UEBMI (%) 0.001 –0.006 to 0.007 0.865
UEBMI, Urban Employee Basic Medical Insurance.
Table 5. Percentage change with 95% confidence intervals in daily hospital admissions associated with 10 lg/m
3
increase in
same-day PM
2.5
concentrations (lag day 0) in two-pollutant models in 200 Chinese cities by region, 2014–16
Variable Adjust SO
2
Adjust NO
2
Adjust CO Adjust O
3
Nationwide –0.01 (–0.15 to 0.13) –0.26 (–0.43 to –0.08) 0.05 (–0.15 to 0.25) 0.17 (0.06 to 0.28)
South 0.14 (–0.04 to 0.32) –0.23 (–0.46 to 0.01) 0.28 (0 to 0.55) 0.36 (0.19 to 0.54)
North –0.17 (–0.38 to 0.04) –0.28 (–0.54 to –0.02) –0.20 (–0.47 to 0.07) 0 (–0.14 to 0.13)
International Journal of Epidemiology, 2019, Vol. 48, No. 4 1147
Downloaded from https://academic.oup.com/ije/article/48/4/1142/5510153 by guest on 18 July 2021
Supplementary data at IJE online). Details of the system-
atic literature search and procedures are presented. We
identified 29 studies that assessed the acute effects of
PM
2.5
on total mortality. For an increase of 10 lg/m
3
in
PM
2.5
, the combined excess risk of mortality was 0.49%
(95% CI: 0.39–0.59). Reviews published in 2013
43
and
2015
44
also concluded that a 10-lg/m
3
increase in PM
2.5
would be associated with a 0.38% (95% CI: 0.31–0.45)
and 0.36% (95% CI: 0.26–0.46) increase in total mortal-
ity, respectively. Our findings were consistent with these
previous Chinese meta-analyses. Only four studies in our
systematic review, however, examined the association be-
tween PM
2.5
pollution and overall morbidity risk.
17,18,45,46
The pooled risk of morbidity per increase of 10 lg/m
3
in
PM
2.5
was 0.30% (95% CI: 0.10–0.51). These studies
were either single-city study
17,45,46
or covered only largest
hospitals.
18
The combined estimates from single-city stud-
ies tend to be higher than our national estimate. Aspects of
city selection, model specification and the publication bias
in single-city studies may all have led to upward
estimates.
19
Moreover, the majority of previous studies
were conducted in major large Chinese cities. It has been
hypothesized that the toxicity of particulate matter may be
greater for particles originating in larger cities.
35
The com-
position and sources of PM
2.5
may also vary across cities
of different sizes. Interestingly, we have previously esti-
mated a 0.19% increase in all-cause admissions in 26
Chinese cities.
32
Moreover, our estimate is comparable to
that of a recent national study of the association between
PM
2.5
and mortality risk in China. Chen and colleagues es-
timated a comparable increase (0.22%) in mortality from
total non-accidental causes per 10-lg/m
3
increase in 2-day
moving average PM
2.5
concentrations (lag days 0–1) in
272 Chinese cities.
9
We observed spatial heterogeneity in the short-term
effects of PM
2.5
on hospital admissions between cities. The
association appeared to be stronger in southern cities. This
spatial heterogeneity is consistent with a regional effect
modification observed in prior Chinese studies
8,9
; a nation-
wide analysis in 272 Chinese cities reported weak or non-
significant effects of PM
2.5
on mortality in north-eastern
and north-western regions.
9
There are several possible
explanations for the spatial heterogeneity in health effects
with respect to PM
2.5
. First, it may be attributable to the
variations in the composition and source of PM
2.5
across
cities. PM
2.5
in northern cities has a higher proportion of
crustal constituents,
41,47
which have been suggested to ex-
ert relatively less-hazardous effects than other PM
2.5
com-
ponents.
42
Second, the overall mean daily PM
2.5
concentrations during the study period were 26.1% higher
in northern cities. We observed a negative effect modifica-
tion by mean PM
2.5
concentrations, consistent with
findings in previous multi-city studies, where higher daily
mortality risk estimates were calculated for cities with
lower ambient particulate matter levels.
8,9,27,28
This may
be related to the ‘harvesting effect’ mentioned above.
Third, the spatial variation in risk estimates could also be
explained by the substantial differences in weather condi-
tions across regions. The overall mean daily temperature
and relative humidity were approximately 72.7 and 35.6%
higher in southern China, respectively, than in the north.
We reported positive associations between estimated
PM
2.5
effects and cities’ mean temperature and relative hu-
midity, which is consistent with findings from a recent na-
tional study.
9
This positive effect modification may be
associated with exposure patterns and with better exposure
characterization, as people spend more time outdoors in
warmer weather.
48
In 2012, China launched a national air-quality standard
for PM
2.5
. To date, few Chinese studies have characterized
the health effects of PM
2.5
at concentrations below the reg-
ulatory limit. In our analysis, we found a positive associa-
tion between hospital admissions and PM
2.5
for days when
daily PM
2.5
concentrations met the CAAQS. The shape as-
sociation also indicated that PM
2.5
at low levels could still
increase the risk of hospital admission. Our findings were
supported by those of studies that reported PM
2.5
-related
health effects at levels below regulatory limits in the
USA.
49,50
Our findings suggest that a more stringent PM
2.5
standard than the current PM
2.5
CAAQS may be needed in
China from the perspective of public health. In addition,
the exposure–response curve was slightly nonlinear with a
plateauing trend at high PM
2.5
concentrations, indicating
that a unit reduction of PM
2.5
at relatively lower levels
might generate more health benefits.
We estimated a 0.19% increase in total hospital admis-
sions per 10-lg/m
3
increase in PM
2.5
; although small, such
an increase may have major public-health implications.
China is a large country with a population of more than
1.3 billion. In our data, the annual mean PM
2.5
concentra-
tion across all cities was 51 mg/m
3
. A 10-lg/m
3
reduction
in PM
2.5
levels would reduce total hospital admissions by
0.33 (0.12–0.53) million (2016 data), suggesting that air-
quality improvements in China could yield remarkable
public-health benefits.
In the two-pollutant models, the PM
2.5
effect estimates
decreased substantially and became statistically insignifi-
cant (and even negative) after controlling for SO
2
,NO
2
and CO. We conducted a separate analysis for SO
2
,NO
2
,
CO and O
3
. These pollutants were all associated with hos-
pital admissions (Supplementary Table 7, available as
Supplementary data at IJE online). Three recent studies in
272 Chinese cities reported that the effects of SO
2
,NO
2
and CO on mortality remained significant after controlling
1148 International Journal of Epidemiology, 2019, Vol. 48, No. 4
Downloaded from https://academic.oup.com/ije/article/48/4/1142/5510153 by guest on 18 July 2021
for PM
2.5
.
51–53
NO
2
generally serves as a surrogate mea-
sure for vehicular pollution because of its close association
with vehicle exhaust emissions.
54
SO
2
is largely from com-
bustion of sulphur-containing fuels such as coal and oil. In
China, a substantial proportion of PM
2.5
originates from
vehicle exhaust emissions and fossil-fuel combustion.
47,55
The notable collinearity between pollutants made it diffi-
cult to precisely assess the independent effects of PM
2.5
on
hospital admissions.
One of the strengths of this study was the uniform ap-
proach used to examine city-specific associations between
PM
2.5
levels and hospital admissions in 200 representative
cities, avoiding the potential publication bias in single-city
studies and providing an overall effect estimate for China.
Several limitations also should be noted. First, we used the
citywide average PM
2.5
concentrations as a proxy for per-
sonal exposure, which may have caused exposure measure-
ment error, potentially biasing the estimates downward.
56
Second, the lack of data on PM
2.5
constituents and sources
limited our ability to further investigate the heterogeneity
of PM
2.5
health effects between cities.
57
Third, although
the two-pollutant models were fitted to examine the ro-
bustness of the association between PM
2.5
levels and hospi-
tal admissions, the collinearity between pollutants
precluded an assessment of the independent effects of
PM
2.5
on hospital admissions. Fourth, as applied in previ-
ous nationwide studies,
9,11,35,37
we used the same df values
for models across cities. Examining the city-specific associ-
ations between PM
2.5
levels and hospital admissions can
provide an overall effect estimate and increase the compa-
rability of the effect estimates across cities; however, China
is a large country and weather conditions and topography
vary by location. Using the same df for models across cities
may result in residual confounding, since the health effects
of meteorological factors have been shown to vary
spatially.
58
Fifth, we linked PM
2.5
levels to hospital admis-
sions by date of admission rather than by the date of symp-
tom onset. This may have introduced non-differential error
in exposure measurement and biased the effect estimates
towards the null.
56
Finally, though total hospital admis-
sions have been validated as an effective measure of
morbidity in assessing air-pollution-associated health
effects,
50,59–61
they include unrelated causes such as
planned surgeries. However, in China, there is no general
practitioner-based referral system.
62
Hospital visits are
generally unscheduled and are on a first-come, first-served
basis.
63
Therefore, the impact of planned hospital admis-
sion is expected to be minor and hospital records could
provide reliable morbidity information for a geographi-
cally defined population.
63
The use of total hospital admis-
sion data has become an important tool in evaluating the
effects of air pollution on public health in China.
18,61,63
In addition, we know of no reason as to why the frequency
of planned hospital admissions would be associated with
PM
2.5
levels—a condition necessary to bias our results.
In summary, we found a significant association between
PM
2.5
concentrations and total hospital admissions, even
at levels below the current CAAQS, indicating a more gen-
eral role for air pollution in human health than cardio-
respiratory diseases alone. As the first nationwide Chinese
study reporting the effects of PM
2.5
on total hospital
admissions, our findings should be useful for assessing hu-
man health effects of PM
2.5
pollution and for policymaking
in China, although the association between PM
2.5
and
cause-specific hospitalizations requires further study. Our
findings strengthen the rationale for more stringent limits
on PM
2.5
levels in China.
Funding
This work was supported by the National Natural Science
Fund Projects of China (81872695, 91846112, 91546120,
81473043) and the Key Project of Natural Science Funds of China
(81230066).
Supplementary data
Supplementary data are available at IJE online.
Conflict of interest: None declared.
References
1. Brunekreef B, Holgate ST. Air pollution and health. Lancet
2002;360:1233–42.
2. Pope CA, 3rd, Dockery DW. Health effects of fine particulate air
pollution: lines that connect. J Air Waste Manag Assoc 2006;56:
709–42.
3. Cohen AJ, Brauer M, Burnett R et al. Estimates and 25-year
trends of the global burden of disease attributable to ambient air
pollution: an analysis of data from the Global Burden of
Diseases Study 2015. Lancet 2017;389:1907–18.
4. Zanobetti A, Schwartz J. The effect of fine and coarse particulate
air pollution on mortality: a national analysis. Environ Health
Perspect 2009;117:898–903.
5. Samoli E, Stafoggia M, Rodopoulou S et al. Associations be-
tween fine and coarse particles and mortality in Mediterranean
cities: results from the MED-PARTICLES project. Environ
Health Perspect 2013;121:932–38.
6. Dai L, Zanobetti A, Koutrakis P, Schwartz JD. Associations of
fine particulate matter species with mortality in the United
States: a multicity time-series analysis. Environ Health Perspect
2014;122:837–42.
7. Alessandrini ER, Stafoggia M, Faustini A et al. Association
between short-term exposure to PM2.5 and PM10 and mor-
tality in susceptible subgroups: a multisite case-crossover
analysis of individual effect modifiers. Am J Epidemiol 2016;
25:25.
International Journal of Epidemiology, 2019, Vol. 48, No. 4 1149
Downloaded from https://academic.oup.com/ije/article/48/4/1142/5510153 by guest on 18 July 2021
8. Li G, Xue M, Zeng Q et al. Association between fine ambient
particulate matter and daily total mortality: an analysis from
160 communities of China. Sci Total Environ 2017;600:108–13.
9. Chen R, Yin P, Meng X et al. Fine particulate air pollution and
daily mortality. a nationwide analysis in 272 Chinese cities. Am
J Respir Crit Care Med 2017;196:73–81.
10. Villeneuve PJ, Chen L, Stieb D, Rowe BH. Associations between
outdoor air pollution and emergency department visits for stroke
in Edmonton, Canada. Eur J Epidemiol 2006;21:689–700.
11. Dominici F, Peng RD, Bell ML et al. Fine particulate air pollu-
tion and hospital admission for cardiovascular and respiratory
diseases. JAMA 2006;295:1127–34.
12. Peng RD, Bell ML, Geyh AS et al. Emergency admissions for car-
diovascular and respiratory diseases and the chemical composi-
tion of fine particle air pollution. Environ Health Perspect 2009;
117:957–63.
13. Stafoggia M, Samoli E, Alessandrini E et al. Short-term associa-
tions between fine and coarse particulate matter and hospitaliza-
tions in Southern Europe: results from the MED-PARTICLES
project. Environ Health Perspect 2013;121:1026–33.
14. Lanzinger S, Schneider A, Breitner S et al. Ultrafine and fine par-
ticles and hospital admissions in central Europe: results from the
UFIREG study. Am J Respir Crit Care Med 2016;194:1233–41.
15. Barnett AG, Williams GM, Schwartz J et al. The effects of air
pollution on hospitalizations for cardiovascular disease in elderly
people in Australian and New Zealand cities. Environ Health
Perspect 2006;114:1018–23.
16. Atkinson RW, Kang S, Anderson HR, Mills IC, Walton HA.
Epidemiological time series studies of PM2.5 and daily mortality
and hospital admissions: a systematic review and meta-analysis.
Thorax 2014;69:660–65.
17. Wang X, Chen R, Meng X et al. Associations between fine parti-
cle, coarse particle, black carbon and hospital visits in a Chinese
city. Sci Total Environ 2013;460:1–6.
18. Chen G, Zhang Y, Zhang W et al. Attributable risks of emer-
gency hospital visits due to air pollutants in China: a multi-city
study. Environ Pollut 2017;228:43–49.
19. Anderson HR, Atkinson RW, Peacock JL, Sweeting MJ,
Marston L. Ambient particulate matter and health effects: publi-
cation bias in studies of short-term associations. Epidemiology
2005;16:155–63.
20. Bell ML, Ebisu K, Peng RD et al. Seasonal and regional short-
term effects of fine particles on hospital admissions in 202 US
counties, 1999–2005. Am J Epidemiol 2008;168:1301–10.
21. Bell ML, Ebisu K, Peng RD, Samet JM, Dominici F. Hospital
admissions and chemical composition of fine particle air pollu-
tion. Am J Respir Crit Care Med 2009;179:1115–20.
22. Hu S, Tang S, Liu Y et al. Reform of how health care is paid for
in China: challenges and opportunities. Lancet 2008;372:
1846–53.
23. Almond D, Chen Y, Greenstone M, Hongbin L. Winter heating
or clean air? Unintended impacts of China’s Huai river policy.
American Economic Review 2009;99:184–90.
24. Ebenstein A, Fan M, Greenstone M, He G, Zhou M. New evi-
dence on the impact of sustained exposure to air pollution on life
expectancy from China’s Huai River Policy. Proc Natl Acad Sci
USA 2017;114:10384–89.
25. Technical specifications for operation and quality control of am-
bient air quality continuous monitoring system for particulate
matter (PM10/PM2.5), 2015. http://www.mep.gov.cn/gkml/
hbb/bgth/201506/W020150619564063514615.pdf (7 November
2017, date last accessed).
26. Yin P, Chen R, Wang L et al. Ambient ozone pollution and daily
mortality: a nationwide study in 272 Chinese cities. Environ
Health Perspect 2017;125:117006.
27. Chen R, Kan H, Chen B et al. Association of particulate air pol-
lution with daily mortality: the China Air Pollution and Health
Effects Study. Am J Epidemiol 2012;175:1173–81.
28. Yin P, He G, Fan M et al. Particulate air pollution and mortality
in 38 of China’s largest cities: time series analysis. BMJ 2017;
356:j667.
29. Bell ML, McDermott A, Zeger SL, Samet JM, Dominici F.
Ozone and short-term mortality in 95 US urban communities,
1987–2000. JAMA 2004;292:2372–78.
30. Guo Y, Barnett AG, Pan X, Yu W, Tong S. The impact of tem-
perature on mortality in Tianjin, China: a case-crossover design
with a distributed lag nonlinear model. Environ Health Perspect
2011;119:1719–25.
31. Wellenius GA, Burger MR, Coull BA et al. Ambient air pollution
and the risk of acute ischemic stroke. Arch Intern Med 2012;
172:229–34.
32. Liu H, Tian Y, Xiang X et al. Ambient particulate matter con-
centrations and hospital admissions in 26 of China’s largest cit-
ies: a case-crossover study. Epidemiology 2018;29:649–57.
33. Higgins JP, Thompson SG. Quantifying heterogeneity in a meta-
analysis. Stat Med 2002;21:1539–58.
34. Samoli E, Analitis A, Touloumi G et al. Estimating the exposure-
response relationships between particulate matter and mortality
within the APHEA multicity project. Environ Health Perspect
2005;113:88–95.
35. Peng RD, Chang HH, Bell ML et al. Coarse particulate matter
air pollution and hospital admissions for cardiovascular and re-
spiratory diseases among Medicare patients. JAMA 2008;299:
2172–79.
36. National Health and Family Planning Commission, 2017. China
Health and Family Planning Yearbook 2016. http://www.moh.
gov.cn/guihuaxxs/s10748/201708/d82fa7141696407abb4ef764
f3edf095.shtml? winzoom¼1(20 May 2018, date last accessed).
37. Di Q, Dai L, Wang Yet al. Association of short-term exposure to
air pollution with mortality in older adults. JAMA 2017;318:
2446–56.
38. Alessandrini ER, Stafoggia M, Faustini A et al. Association
between short-term exposure to PM2.5 and PM10 and mortality
in susceptible subgroups: a multisite case-crossover analysis
of individual effect modifiers. Am J Epidemiol 2016;184:
744–54.
39. Katsouyanni K, Touloumi G, Samoli E et al. Confounding and
effect modification in the short-term effects of ambient particles
on total mortality: results from 29 European cities within the
APHEA2 project. Epidemiology 2001;12:521–31.
40. Costa AF, Hoek G, Brunekreef B, Ponce de Leon AC. Air pollu-
tion and deaths among elderly residents of Sao Paulo, Brazil: an
analysis of mortality displacement. Environ Health Perspect
2017;125:349–54.
1150 International Journal of Epidemiology, 2019, Vol. 48, No. 4
Downloaded from https://academic.oup.com/ije/article/48/4/1142/5510153 by guest on 18 July 2021
41. Yang F, Tan J, Zhao Q et al. Characteristics of PM 2.5 speciation
in representative megacities and across China. Atmos Chem
Phys 2011;11:5207–19.
42. Thurston GD, Burnett RT, Turner MC et al. Ischemic heart dis-
ease mortality and long-term exposure to source-related compo-
nents of U.S. fine particle air pollution. Environ Health Perspect
2016;124:785–94.
43. Shang Y, Sun Z, Cao J et al. Systematic review of Chinese studies
of short-term exposure to air pollution and daily mortality.
Environ Int 2013;54:100–11.
44. Lu F, Xu D, Cheng Y et al. Systematic review and meta-analysis
of the adverse health effects of ambient PM2.5 and PM10
pollution in the Chinese population. Environ Res 2015;136:
196–204.
45. Wang X, Li G, Jin X et al. Study of relationship between atmo-
spheric fine particulate matter concentration and one grade a ter-
tiary hospital emergency room visits during 2012 and 2013 in
Beijing. Zhonghua Yu Fang Yi Xue Za Zhi 2016;50:73–78.
46. Liu R, Zeng J, Jiang X et al. The relationship between airborne
fine particle matter and emergency ambulance dispatches in a
southwestern city in Chengdu, China. Environ Pollut 2017;229:
661–67.
47. Ren L, Yang W, Bai Z. Characteristics of major air pollutants in
China. Adv Exp Med Biol 2017;1017:7–26.
48. Chen R, Peng RD, Meng X et al. Seasonal variation in the acute
effect of particulate air pollution on mortality in the China Air
Pollution and Health Effects Study (CAPES). Sci Total Environ
2013;451:259–65.
49. Shi L, Zanobetti A, Kloog I et al. Low-concentration PM2.5 and
mortality: estimating acute and chronic effects in a population-
based study. Environ Health Perspect 2016;124:46–52.
50. Makar M, Antonelli J, Di Q et al. Estimating the causal effect of
low levels of fine particulate matter on hospitalization.
Epidemiology 2017;28:627–34.
51. Wang L, Liu C, Meng X et al. Associations between short-term
exposure to ambient sulfur dioxide and increased cause-specific
mortality in 272 Chinese cities. Environ Int 2018;117:33–39.
52. Chen R, Yin P, Meng X et al. Associations between ambient ni-
trogen dioxide and daily cause-specific mortality: evidence from
272 Chinese cities. Epidemiology 2018;29:482–89.
53. Liu C, Yin P, Chen R et al. Ambient carbon monoxide and car-
diovascular mortality: a nationwide time-series analysis in 272
cities in China. Lancet Planet Health 2018;2:e12–e18.
54. Seaton A, Dennekamp M. Hypothesis: ill health associated with
low concentrations of nitrogen dioxide—an effect of ultrafine
particles? Thorax 2003;58:1012–15.
55. Gao J, Wang K, Wang Y et al. Temporal-spatial characteristics
and source apportionment of PM2.5 as well as its associated
chemical species in the Beijing-Tianjin-Hebei region of China.
Environ Pollut 2018;233:714–24.
56. Goldman GT, Mulholland JA, Russell AG et al. Impact of expo-
sure measurement error in air pollution epidemiology: effect of
error type in time-series studies. Environ Health 2011;10:10–61.
57. Krall JR, Strickland MJ. Recent approaches to estimate associa-
tions between source-specific air pollution and health. Curr
Envir Health Rpt 2017;4:68–78.
58. Zhao Q, Zhang Y, Zhang W et al. Ambient temperature and
emergency department visits: time-series analysis in 12 Chinese
cities. Environ Pollut 2017;224:310–16.
59. Gwynn RC, Burnett RT, Thurston GD. A time-series analysis of
acidic particulate matter and daily mortality and morbidity in
the Buffalo, New York, region. Environ Health Perspect 2000;
108:125–33.
60. Gurung A, Son JY, Bell ML. Particulate matter and risk of hospi-
tal admission in the Kathmandu Valley, Nepal: a case-crossover
study. Am J Epidemiol 2017;186:573–80.
61. Chen R, Chu C, Tan J et al. Ambient air pollution and hospital
admission in Shanghai, China. J Hazard Mater 2010;181:
234–40.
62. Li Q, Xie P. Outpatient workload in China. Lancet 2013;381:
1983–84.
63. Xu X, Li B, Huang H. Air pollution and unscheduled hospital
outpatient and emergency room visits. Environ Health Perspect
1995;103:286–89.
International Journal of Epidemiology, 2019, Vol. 48, No. 4 1151
Downloaded from https://academic.oup.com/ije/article/48/4/1142/5510153 by guest on 18 July 2021