Short-Term Ambient Particulate Air Pollution and Hospitalization Expenditures of Cause-Specific Cardiorespiratory Diseases in China: A Multicity Analysis.

BACKGROUND: Ambient air pollution is leading risk factor for health burden in China. Few studies in China have investigated the economic loss related to short-term exposure to ambient PM2.5, which could trigger acute onset of cardiorespiratory diseases within a few days.
METHODS: Daily ambient air pollutants data are obtained for each city from the National Air Quality Monitoring System and daily hospitalization data are obtained from the urban employee-based basic medical insurance scheme database in 74 Chinese cities with an average coverage of 88.5 million urban employees during 2016-2017. A three-stage time-series analytic approach is used in this study to investigate the impact of short-term exposure to ambient fine particulate (PM2.5) air pollution on hospital admissions, expenses and hospital stays of three cause-specific cardiorespiratory diseases, including lower respiratory infections (LRI), coronary heart disease (CHD) and stroke in the included cities.
FINDINGS: Based on the time-series analysis using daily hospitalization data, 28,560 LRI cases, 54,600 CHD cases, and 23,989 stroke cases are attributable to ambient PM2.5 in the 74 cities during the study period, and the related attributable expenses are 220 million CNY (US$ 32.9 million) for LRI, 458 million CNY (US$ 68.5 million) for CHD, and 410 million CNY (US$ 65.8 million) for stroke, respectively. These attributable numbers account for 1.45% to 2.05% of total hospital admissions and 1.10% to 1.51% of total expenses for the three diseases during 2016-2017, respectively. The attributable numbers for the three cause-specific cardiorespiratory diseases would increase to 362,007 hospital admission cases and 3.68 billion CNY expenses ($US550 million) in the entire urban employee population (299 million) in China during 2016-2017, and the related direct economic loss of absence from work would be 798 million CNY (US$ 119.3 million).
INTERPRETATION: Our results support that short-term exposure to ambient PM2.5 pollution could lead to significant health and economic impacts in China. Reducing levels of ambient PM2.5 can avoid substantial health damage and expenditures, and generate appreciable economic benefits from decreasing absence from work. FUNDING: Natural Science Foundation of China (82073509, 71903010, 71903011), and the National Key Research and Development Program of China (2017YFC0211600, 2017YFC0211601).

Evidence before this study

Billions of people live in over polluted regions with ambient PM2.5 levels over the WHO guideline (annual mean: 10 µg/m3), whereas evidence is lacking for the economic loss related to short-term exposure to ambient air pollution, especially in developing countries with high ambient air pollution. Most existing studies have evaluated the health and economic losses related to long-term exposure to ambient air pollution based on the total mortality or morbidity data, and few studies have evaluated the economic loss related to short-term exposure to ambient air pollution based on detailed hospitalization expenditure data.

Added value of this study

To our best knowledge, our study provides the first detailed evaluation about the excess hospitalization burden attributable to short-term exposure to ambient PM2.5 based on daily hospital admission data over multiple cities in China. This study estimates that short-term exposure to ambient PM2.5 leads to significant increases in hospital admissions and hospitalization expenses for three major cause-specific cardiorespiratory diseases responsible for population disease burden including lower respiratory infections, coronary heart disease and stroke in urban employees in 74 Chinese cities during 2016-2017. The attributable number estimates would increase to 362,007 hospital admission cases and 3.68 billion CNY expenses ($US550 million) for the three cause-specific cardiorespiratory diseases in the entire urban employee population (299 million) in China during 2016-2017, and the related direct economic loss of absence from work for the younger adults (15-64 years) subgroup would be 798 million CNY (US$ 119.3 million).

Implications of all the available evidence

Given that billions of people live in the over polluted regions with ambient PM2.5 levels over the WHO guideline, especially in developing countries, the potential economic loss due to short-term exposure to ambient PM2.5 would be enormous over the world. Accurately estimating the impact of ambient air pollution is crucial for the establishment of effective air quality regulations, and our study findings may provide important evidence for ambient air pollution control action from the angle of input-output (policy cost-benefit) balance.

Alt-text: Unlabelled box

Introduction

Ambient air pollution is leading millions of premature deaths all over the world, especially in developing countries, and has become the eighth leading risk factor for death according to the Global Burden of Diseases (GBD) in 20171,2. Ambient particulate matter with an aerodynamic diameter of 2.5 micrometers or less (PM2.5) is the major air pollutant receiving world-wide attention due to its high toxicity. More than 90% people live in the over polluted regions, where PM2.5 exceeds the guideline (annual mean: 10 µg/m3) proposed by the World Health Organization (WHO). In addition, numerous studies indicate that ambient PM2.5 pollution is associated with a broad range of chronic and acute health effects,[3], [4], [5], [6], [7], [8]. Exposure to ambient air pollution not only damages public health but also causes economic loss, resulting in increased healthcare expenditures and decreased work hours ,. The welfare damages for ambient air pollution and household air pollution were estimated to be $3767 billion, accounting for 5.06% of world gross domestic products (GDP) in 201511.

Ambient air pollution, especially ambient PM2.5, has significant contribution on cardiorespiratory mortality,. China has the largest population in the world and is facing severe ambient air pollution issue. The Global Burden of Disease 2017 study shows that approximately 1.1 million people died prematurely, and 21.8 million years of life lost were lost in China in 2015 due to long-term exposure to ambient air pollution. One study using national mortality-specific baseline incidence rates derived from the GBD study estimated that the national long-term PM2.5-attributable mortality in 2016 was 0.96 million, accounting for about 9.98% of the total reported deaths nationwide. In addition, the specific hospital admission morbidities for cardiovascular disease and respiratory disease related to PM2.5 were 0.36 million and 0.61 million, respectively, based on baseline incidence rates derived from previous references. The result also shows that long-term PM2.5 exposure caused an economic loss of 101.4 billion US$, accounting for 0.9% of the national GDP in 2016, based on provincial health cost and per capita GDP information.

Although several studies also reported economic loss related to long-term exposure to ambient air pollution based on data on hospital admissions and average expense per time,, few studies have evaluated the economic loss related to short-term exposure to ambient air pollution based on detailed hospitalization expenditure data. Numerous studies have shown that short-term exposure to ambient PM2.5 within a few days could trigger acute onset of cardiorespiratory diseases including lower respiratory infections (LRI), coronary heart disease (CHD), and stroke[18], [19], [20], which are among the major cardiorespiratory diseases responsible for population disease burden. Several previous studies also show that short-term exposure to PM2.5 could lead to more intensive care admissions and more hospital admissions of cardiorespiratory diseases[21], [22], [23]. Therefore, short-term exposure to ambient PM2.5 has important adverse health effects in addition to that of long-term exposure to ambient PM2.5. The evaluation on the health and economic burdens of both types of exposure constitutes an integrated perspective on the adverse socioeconomic impact related to ambient PM2.5. To date, detailed evaluation on the hospitalization expenditures related to short-term exposure to ambient PM2.5 for these specific diseases has been lacking in China.

In addition, the levels of ambient particulate air pollution and spectrum of disease may vary geographically, especially in China, a large country with the largest population in the world, and it is necessary to investigate the variation in the impact of air pollution on these cause-specific cardiorespiratory diseases (e.g., LRI, CHD, and stroke) across different regions for targeted disease prevention. A previous investigation finds that a 10 µg/m3 increase in inhalable particulate matter (PM10) reduces life expectancy by 0.64 years based on quasiexperimental variation in PM10 generated by China's Huai River policy, which provides free or heavily subsidized coal for indoor heating during the winter to cities north of the Huai River but not to those in the south. The large variations in both ambient particulate air pollution levels and spectrum of disease across China provide a unique opportunity to investigate the relationship between ambient PM2.5 and three cause-specific cardiorespiratory diseases in different regions, e.g., northern region and southern region divided by the Huai River.

This study thus evaluates the health and hospitalization expenditures related to ambient PM2.5 with respect to these cause-specific cardiorespiratory diseases in 74 cities of China in 2016-2017 based on detailed daily morbidity data obtained from the urban employee-based basic medical insurance scheme (UEBMI) of China.

Materials and Methods

This study evaluated the relationship between ambient particulate air pollution and hospital admissions, expenditures and hospital stays of three cause-specific cardiorespiratory diseases, including LRI, CHD and stroke, based on detailed hospital admissions data obtained from the UEBMI in China. Our study extracted daily hospital admission data from 74 administrative divisions of prefecture-level cities (out of 338 prefecture-level cites in total) in China during January 1, 2016 and December 31, 2017.

Health Data

This study population includes all the urban Chinese populations who joined the UEBMI in the included cities. The UEBMI covers all the current urban employees and retirees of past urban employees. Insurance policy is the same after employees retired in China,. Specifically, daily hospital admissions and hospital expenditure data are obtained for 74 cities and hospital stay data are obtained for 69 cities from the UEBMI (covering 299 million employees). Three major cardiorespiratory diseases including LRI, CHD, and stroke are our targeted diseases, which are coded according to the ICD-10 codes (LRI: J12-J22; CHD: I20-I25; stroke: I60-I61, I63-I64),. For each disease, the hospital expenditure and hospital stay for a given day during the study are defined as the average expenses and days per person in that day multiplied by the number of patients who were admitted to the hospital because of the disease in that day. The UEBMI also includes outpatient visits information. However, the disease diagnosis for outpatient visits is less reliable than hospital admissions, and thus we do not include outpatient visits in our study. This study is exempted from institutional review board approval by the Ethics Committee of Peking University Health Science Center, because all data used for the study are collected for administrative purposes without any individual identifiers.

Ambient air pollution data

Daily ambient air pollutants data, including fine particulate matter (PM2.5), nitrogen dioxide (NO2), ozone (O3), sulfur dioxide (SO2) and carbon monoxide (CO), are obtained for each city from the National Air Quality Monitoring System,. Environmental air quality monitoring stations are usually set at fixed-sites which have good regional representativeness and can objectively reflect the ambient air quality level and change rules in a certain space. Daily mean concentrations for all ambient air pollutants in each city are calculated by averaging all valid monitoring measurements if there is more than one monitor in that city. We also collect daily meteorological data on mean temperature (˚C) and relative humidity (%) from the National Meteorological Data Service Center of China (http://data.cma.cn/, accessed 2019-01-01).

Statistical analysis

In this study, we use a three-stage time-series analytic approach to assess the city-specific and overall estimates of the association between ambient PM2.5 and daily hospital admissions, total hospital stays and total expenses for LRI, CHD and stroke. In the first stage, we estimate city-specific associations using a Constrained Distributed Lag Model (CDLM) with a time-series quasi-Poisson regression, as below:where E(Yt) is the expected count of health variables in the analyzed city on day t; bst(air pollutants) is the concentrations of PM2.5 in the analyzed city on day t using a CDLM with 3 degrees of freedom (df) polynomials for 2 lag days; Day of week is the indicator variable for day of week to account for possible variations over weekdays and weekends; Holiday is the indicator variable for public holidays to account for possible variations over holidays and non-holidays; nst(calendar time) is the natural cubic spline for calendar time with 7 df per year to exclude unmeasured time trends longer than 2 months; nst(temperature) and nst(relative humidity) are natural cubic splines with 6 df for 3-day moving average temperature and 3 df for 3-day moving average relative humidity on day t to control for potential non-linear and delayed confounding effects of meteorological conditions, respectively. Specifically, non-linear effects are controlled for by using natural cubic splines, and delayed effects are controlled for by using moving average data. Associations between short-term exposure to ambient PM2.5 and six health variables (hospital admissions and total hospital stays for LRI, CHD and stroke) are assessed separately using the above single-pollutant models. The city-specific effect estimates are expressed as relative risks (RR) and related 95% confidence intervals (CI) by using the equation RR = eβ, where β is the coefficient derived from the above model.

In the second stage, we pool the city-specific effect estimates to obtain the overall effect estimates that represent the effects for all included cities using a random-effect model. The lag 0-2 effects of PM2.5 on health variables are estimated and expressed as RR and related 95% CI associated with a 10 µg/m³ increase in PM2.5. Lag 0-2 is chosen as the main exposure metric because many previous studies found most apparent associations between ambient PM2.5 and health outcomes within this time frame,,, and sensitivity analyses are performed from lag0-1 to lag0-6. Moreover, we calculate health variable-specific overall excess risk in percent change (%) ((RR − 1)*100) related to a 10 µg/m3 increase in ambient PM2.5 at lag 0-2.

In the third stage, we first use overall effect estimates from the second stage to calculate attributable fraction (AF) and attributable number (AN) of health variables during the study period based on the estimated RR at lag 0-2. ANic in each year i for city c are computed through the formula,:where Nic is the annual total number of health variable in year i for city c, β is health variable-specific overall coefficients derived from the second stage and Dic is the annual average concentration of PM2.5 in year i for city c. Annual AN for a given health variable is then summed over years and cities. AF is calculated by dividing the AN by the sum of the corresponding health variable over years and cities. In addition, the upper and lower 95% CI values of the pooled β are used to calculate the 95% CIs of AN and AF using the above equations.

Besides, we estimate AFs and ANs of total expenses for LRI, CHD and stroke using the formula:where ANac is the AN of hospital admissions for city c during the study period, and AEc is the average expense across all admissions for city c during the study period. AF is calculated by dividing the AN by the sum of total expenses over cities.

In order to estimate the national level health burden of ambient PM2.5 pollution, we also project the hospital admissions, total hospital stays, and total expenses attributable to PM2.5 in the entire urban employee population (299 million averaged over 2016-2017) who joined the UEBMI based on results of 74 cities included in the study assuming the same exposure-response relationship between short-term exposure to ambient PM2.5 and the examined health outcomes across China. In addition, we also estimate the direct economic loss of absence from work for the younger adults (15-64 years) subgroup using the city-level average wage from Chinese Statistical Yearbook during the study period,. We multiply total hospital stay in days attributable to PM2.5 by average daily wage in each city, and then multiply the proportion of younger UEBMI participants (15-64 years) in the city to calculate the city-specific economic loss for absence from work. Then we summarize the total economic loss attributable to PM2.5 over 74 cities.

We then conduct subgroup analyses by geographical region, gender and age (15-64 and ≥65). All included cities are divided into northern and southern regions according to the Huai River-Qinling Mountains line. To further understand the potential impact of other air pollutants on the results, two-pollutant models are fitted with average concentrations of NO2, O3, SO2, or CO along with PM2.5 at the same lags in the regression models. The R software (version 3.6.2) is used for all statistical analyses and plotting. R packages dlnm (author: A. Gasparrini) and meta (author: Sara Balduzzi and Gerta Rücker and Guido Schwarzer) are used, and a 2-sided P value<0.05 is considered statistically significant.

Role of the funding source

The sponsors of this study played no role in the study design, data collection, or analysis of data; interpretation of results; or in the writing of this manuscript. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication.

Results

Figure 1 shows the geographic locations of all 74 Chinese cities included in this study. The UEBMI covers an average of 88.5 million population during 2016-2017 across these cities. Table 1 shows the distribution in daily levels of ambient PM2.5 and meteorological variables during the study period. The average daily concentrations of ambient PM2.5 in the 74 cities during the study period is 49.7 µg/m3 (Column 1), and is higher in the northern region (57.6µg/m3) than in the southern region (41.7µg/m3).

Figure 1

Map of the included cities (n=74) in mainland China. The black points represent cities in the northern region (n=37), and the gray points represent cities in the southern region (n=37).

Table 1

Description for daily ambient PM2.5, temperature and relative humidity in 74 Chinese cities during 2016-2017

Measure and Region		Mean	SD	Minimum	Median	Maximum	Interquartile Range
PM2.5 (µg/m³)
	Overall	49.7	37.7	3.0	39.3	389.0	37.1
	Northern regiona	57.6	43.4	3.5	45.1	389.0	42.8
	Southern region	41.7	28.7	3.0	34.3	342.6	31.6
Temperature (°C)
	Overall	15.2	10.2	-22.8	16.7	36.5	15.2
	Northern region	12.4	11.3	-22.8	13.6	35.3	19.0
	Southern region	18.0	8.0	-7.8	18.8	36.5	12.6
Relative humidity (%)
	Overall	68.5	17.9	7.0	71.0	100.0	25.9
	Northern region	60.6	18.6	7.0	61.0	100.0	29.0
	Southern region	76.4	13.2	17.0	77.0	100.0	18.0

Abbreviation: PM2.5: particulate matter with an aerodynamic diameter of 2.5 micrometers or less.

aThere are 37 cities and 37 cities in the southern and northern region, respectively.

Table 2 summarizes descriptive statistics for hospital admission variables during the study period. A total of 6.03 million hospital admission cases, including 1.97 million LRI cases, 2.67 million CHD cases and 1.39 million stroke cases, are identified in the UEBMI database across 74 cities during the study period. The total expenses reach 78,972 million CNY (US$ 11,804 million) for the three cause-specific cardiorespiratory diseases. Among the three diseases, CHD has the largest total expenses of 30,307 million CNY (US$ 4,530 million), higher than those for LRI [19,963 million CNY (US$ 2,984 million)] and stroke [28,702 million CNY (US$ 4,283 million)]. The total hospital stays reach 57.5 million days for the three cause-specific cardiorespiratory diseases in 69 cities during our study period. CHD also has the largest total hospital stays (22.8 million days) among the three diseases, higher than those for LRI (16.8 million days) and stroke (17.9 million days).

Table 2

Demographic characteristics of hospital admission cases in the included Chinese citiesa during 2016-2017

Disease	Total	Northern region	Southern region	Males	Females	Age 15-64 years	Age≥65 years
Hospital admissions (million case)
LRI	1.97	0.98	0.99	0.89	0.75	0.75	0.89
CHD	2.67	2.01	0.66	1.27	0.98	0.82	1.43
Stroke	1.39	0.81	0.59	0.77	0.40	0.42	0.75
Total expenses (million CNY)
LRI	19,963	3,669	16,294	10,354	7,606	6,053	11,906
CHD	30,307	9,201	21,106	16,650	10,549	9,616	17,582
Stroke	28,702	7,490	21,212	16,157	9,787	8,586	17,358
Total hospital stays (million days)
LRI	16.8	7.6	9.2	6.3	8.0	6.1	8.1
CHD	22.8	15.8	7.0	8.8	10.7	6.4	13.1
Stroke	17.9	9.5	8.4	5.2	9.8	5.2	9.8

Abbreviations: CHD: coronary heart disease; CNY: Chinese Yuan; LRI: lower respiratory infections.

aTwo years’ average numbers of the UEBMI participants are 88.5 million for hospital admissions and total expenses (74 cities) and 84.6 million for total hospital stays (69 cities) during 2016-2017; subgroup numbers may not add up to the total number because of rounding, and subgroup numbers excluded admission cases without sex or age information.

Overall excess risks of hospital admissions and total hospital stays for LRI, CHD and stroke at different cumulative lags are shown in Figure 2. The excess risks are generally significant from lag0-2 to lag0-4 for all six health outcomes except the excess risk at lag0-4 for LRI hospital admission. The excess risks are attenuated at lag0-5 to lag0-6.

Figure 2

Overall excess risks in % of hospital admissions and total hospital stays per 10 μg/m3 increase in ambient PM2.5 at different cumulative lag days in included cities.

(A)Hospital admissions for LRI; (B) Hospital admissions for CHD; (C) Hospital admissions for stroke; (D) Total hospital stays for LRI; (E) Total hospital stays for CHD; (F) Total hospital stays for stroke.

Table 3 and Table 4 show the results for the analyses of the associations between ambient PM2.5 (lag0-2) and hospital admissions, total hospital stays and total expenses of LRI, CHD and stroke during 2016-2017. The excess risks in % per 10 µg/m3 increase in ambient PM2.5 (lag0-2) are 0.26%, 0.34% and 0.33% for hospital admissions of LRI, CHD and stroke, respectively, and 0.38%, 0.60% and 0.52% for total hospital stays of LRI, CHD and stroke, respectively (all P<0.05). Overall, the fractions of hospital admissions and total hospital stays attributable to ambient PM2.5 range from 1.45% to 2.05%, and 1.87% to 3.01%, and the fractions of total expenses attributable to ambient PM2.5 range from 1.10% to 1.51% for the three diseases during 2016-2017, respectively.

Table 3

Attributable numbers and fractions of hospital admissions and total hospital stays associated with ambient PM2.5 at lag 0-2 in the included citiesa during 2016-2017

Disease	Health variable	Excess risk in % (95% CI)b	Attributable number (95% CI)	Attributable fraction in %(95% CI)
LRI	Hospital admissions (case)	0.26 (0.03, 0.49)	28 560 (3 796, 52 990)	1.45 (0.19, 2.69)
	Total hospital stays (thousand day)	0.38 (0.10, 0.67)	314 (82, 543)	1.87 (0.49, 3.23)
CHD	Hospital admissions (case)	0.34 (0.10, 0.58)	54 600 (16 808, 91 820)	2.05 (0.63, 3.44)
	Total hospital stays (thousand day)	0.60 (0.31, 0.88)	686 (363, 1003)	3.01 (1.59, 4.40)
Stroke	Hospital admissions (case)	0.33 (0.10, 0.57)	23 989 (7 212, 40 551)	1.72 (0.52, 2.91)
	Total hospital stays (thousand day)	0.52 (0.26, 0.79)	473 (234, 707)	2.64 (1.31, 3.95)

Abbreviations: CHD: coronary heart disease; LRI: lower respiratory infections; aTwo years’ average numbers of the UEBMI participants are 88.5 million for hospital admissions (74 cities) and 84.6 million for total hospital stays (69 cities) during 2016-2017.; b Per 10 µg/m3 increase in ambient PM2.5.

Table 4

Attributable numbers and fractions of total expenses (million CNY) associated with ambient PM2.5 at lag 0-2 in the included citiesa during 2016-2017

Disease	Attributable number (95% CI)	Attributable fraction in % (95% CI)
LRI	220 (29, 408)	1.10 (0.15, 2.05)
CHD	458 (141, 772)	1.51 (0.46, 2.55)
Stroke	410 (123, 694)	1.43 (0.43, 2.42)

Abbreviations: CHD: coronary heart disease; CNY: Chinese Yuan; LRI: lower respiratory infections.

aTwo years’ average number of the UEBMI participants is 88.5 million in 74 cities during 2016-2017.

The largest AF of hospital admissions associated with PM2.5 among the three diseases is the CHD-related AF (2.05%, 95% CI: 0.63%, 3.44%), and the corresponding AN is 54,600 (95% CI: 16,808, 91,820). Moreover, CHD-related AF (3.01%, 95% CI: 1.59%, 4.40%) for total hospital stays is larger than those for the other two diseases as well. For total expenses, CHD-related AF (1.51%, 95% CI: 0.46%, 2.55%) associated with PM2.5 is also more apparent than LRI-related AF (1.10%, 95% CI: 0.15%, 2.05%) and stroke-related AF (1.43%, 95% CI: 0.43%, 2.42%), and the corresponding AN for CHD expenses is 458 million CNY (95% CI: 141, 772), also higher than that for LRI expenses (220 million CNY, 95% CI: 29, 408) and stroke expenses (410 million CNY, 95% CI: 123, 694). In two-pollutant models adjusting for other air pollutants, most AF estimates remain positive but are attenuated, whereas some estimates become even more significant when compared to the results in the single-pollutant models (Tables S1 and S2).

Results of projection of hospital admissions, total expenses and total hospital stays attributable to ambient PM2.5 to the national level show that the AN estimates would increase to 362,007 hospital admission cases, 3.68 billion CNY expenses ($US 550 million) and 5.21 million hospital days for the entire urban employee population in China during 2016-2017 (Table 5).

Table 5

Projection of attributable numbers of hospital admissions, total expenses and total hospital stays associated with ambient PM2.5 at lag 0-2 from the included cities to the national level during 2016-2017

Disease	Health variable	Included citiesa(95% CI)	National levelb(95% CI)
LRI	Hospital admissions (case)	28 560 (3 796, 52 990)	96 491 (12 825, 179 028)
	Total expenses (million CNY)	220 (29, 408)	743 (98, 1 378)
	Total hospital stays (thousand day)	314 (82, 543)	1 110 (291,1 918)
	Economic loss from work day loss (million CNY)	50 (13, 86)	177 (46, 303)
CHD	Hospital admissions (case)	54 600 (16 808, 91 820)	184 468 (56 786, 310 217)
	Total expenses (million CNY)	458 (141, 772)	1 547 (476, 2 608)
	Total hospital stays (thousand day)	686 (363, 1003)	2 426 (1 281, 3 546)
	Economic loss from work day loss (million CNY)	103 (56, 153)	363 (196, 539)
Stroke	Hospital admissions (case)	23 989 (7 212, 40 551)	81 048 (24 366, 137 003)
	Total expenses (million CNY)	410 (123, 694)	1 385 (416, 2 345)
	Total hospital stays (thousand day)	473 (234, 707)	1 670 (829, 2 499)
	Economic loss from work day loss (million CNY)	73 (36, 108)	258 (127, 382)

Abbreviations: CHD: coronary heart disease; CNY: Chinese Yuan; LRI: lower respiratory infections.

aAttributable hospital admissions and total expenses are calculated based on data from 74 cities with an average coverage of 88.5 million population during 2016-2017, whereas attributable total hospital stays and economic loss from work day loss are calculated based on data from 69 cities with an average coverage of 84.6 million population during 2016-2017.

bThe UEBMI covers an average urban employee population of 299 million in China during 2016-2017.

In the subgroup analyses by region for the AF estimates attributable to ambient PM2.5 at lag 0-2, four of the six AF estimates for hospital admissions and total hospital stays (CHD-related hospital admissions, and LRI-, CHD- and stroke-related total hospital stays) are higher in the northern region than in the southern region (Table S3). Results for subgroup analyses by gender show that four of the six AF estimates for hospital admissions and total hospital stays (LRI-, CHD- and stroke-related hospital admissions, and CHD-related total hospital stays) are higher in males than in females (Table S4), and all AF estimates for hospital admissions and total hospital stays are higher in older adults (age≥65 years) subgroup than in younger adults (age 15-64 years) subgroup (Table S5).

In the subgroup analyses for total expenses, the AF for CHD in the northern region (1.90%, 95% CI: 0.25%, 3.52%) is higher than that in the southern region (1.46%, 95% CI: -0.20%, 3.09%), whereas the AF for LRI in the southern region (1.80%, 95% CI: 0.11%, 3.47%) is higher than that in the northern region (0.82%, 95% CI: -0.56%, 2.18%) (Table S6). The AF estimates for total expenses of three diseases are all higher in males than in females, and are all higher in older adults than in younger adults.

The direct economic loss of absence from work for the younger adults (15-64 years) subgroup due to the three cause-specific cardiorespiratory diseases attributable to ambient PM2.5 would be 226 million CNY (US$ 33.8 million) in the included cities, and would increase to 798 million CNY (US$ 119.3 million) for the entire urban employee population in China (Table 5).

Discussion

In general, our study provides the first detailed evaluation about the excess hospitalization expenditure burden attributable to short-term exposure to ambient PM2.5 based on daily hospital admission data from a large population over multiple cities in the context of high air pollution in China. Our results show that ambient PM2.5 would significantly increase the hospital admissions, total expenses and total hospital stays for three major cause-specific cardiorespiratory diseases responsible for population disease burden including LRI, CHD and stroke. The attributable hospital admissions, total expenses and total hospital stays would reach 362,007 hospital admission cases, 3.68 billion CNY expenses ($US 550 million) and 5.21 million hospital days for the entire urban employee population (299 million averaged over 2016-2017) in China assuming the same exposure-response relationship between short-term exposure to ambient PM2.5 and the examined health outcomes as found in the present study. These results suggest that short-term exposure to ambient PM2.5 pollution could lead to significant health and economic impacts in China.

Health expenditure has been increasing in China. A previous study reported that the governmental health expenditures on health care quadrupled from 359 billion CNY to 1.52 trillion CNY from 2008 to 2017. Ambient air pollution is among the leading risk factors for increasing health expenditure. A number of existing studies have evaluated the health and economic burden associated with long-term exposure to ambient air pollution based on measures of mortality and statistics of life (e.g., disability-adjusted life years). whereas few studies have focused on the impact of short-term exposure to ambient air pollution due to a range of barriers. For example, a previous study shows that the total respiratory disease and cardiovascular disease-specific hospital admission morbidities attributable to long-term exposure to ambient PM2.5 are 0.61 million and 0.36 million in China in 2016, respectively, based on baseline incidence rates derived from previous references.

Short-term exposure to ambient PM2.5 may also lead to substantial health burden in addition to long-term exposure to ambient PM2.5 ,, and therefore the potential socioeconomic impact of ambient air pollution may have been underestimated in the literature because of sparse evidence on the economic evaluation for short-term exposure,. A recent study evaluates the health burden and economic loss attributable to short-term exposure to ambient air pollution in China based on data obtained from the national statistical yearbook and other similar sources and exposure-response functions for different ambient air pollutants published in previous epidemiological studies. They estimate that the summed economic loss of hospital admissions for cardiovascular and respiratory diseases attributable to short-term exposure to six major ambient air pollutants is 71.81 billion CNY in 2017 in China. This estimate is much higher than our estimate of 3.68 billion CNY for ambient PM2.5 alone for three cause-specific cardiorespiratory diseases during 2016-2017. Nevertheless, our study does not include other cardiorespiratory diseases beyond the three major cardiorespiratory diseases (i.e., LRI, CHD, stroke) and our study population is also limited to urban employees. However, it should be noted that the exposure-response functions obtained from different studies may be heterogeneous, and levels of different ambient air pollutants are usually highly correlated with each other, and the attributable economic loss estimates for different air pollutants may be substantially overlapped under such conditions. In contrast, our study uses detailed daily hospital admission data from the UEBMI database which could help generate reliable exposure-response function for the estimation of related hospitalization expenditures, and thus may serve as a novel dimension to improve the assessment of economic burden of air pollution on human health. Nevertheless, our study only reflects the direct expenditures on the hospital admissions for the LRI, CHD and stroke, which may still underestimate the economic burden related to short-term exposure to ambient PM2.5.

In addition to the direct expenses due to hospital admissions, ambient PM2.5 will also increase the economic loss due to absence from work for the hospitalized patients. The related direct economic loss of absence from work for the younger adults (15-64 years) subgroup would be 226 million CNY (US$ 33.8 million) in the included cities during the study period, and would increase to 798 million CNY (US$ 119.3 million) for the entire urban employee population in China during the study period (Table 5).

Furthermore, there are also additional costs for other healthcare behaviors such as wages for nursing workers or economic loss due to absence from work for close relatives of the hospitalized patients. In support of these results, one previous study based on the survey data from the Health and Retirement Study and hospitalization data in the US finds that hospital admissions increase out-of-pocket medical spending, unpaid medical bills, and bankruptcy, and reduce earnings which are substantial compared to the out-of-pocket spending increase . Another study on health burden in the European Union also shows that the productivity loss because of early death cost and lost working days account for 41% of total health cost of cancer in EU, which is a little bit higher than the health care cost proportion of 40%. Furthermore, if we include the outpatient visits, the direct expenses would be much higher than the present estimation.

Our study also provides specific evaluation for subgroups divided by region, gender and age, and finds some interesting results. First, the AF estimates of hospital admissions and total expenses for CHD and stroke in the northern region are significant and generally higher than those in the southern region, whereas the AF estimates of hospital admissions and total expenses for LRI in the southern region are higher than those in the northern region. This finding suggests that short-term exposure to ambient PM2.5 may have different impact patterns over the northern and southern regions of China, and CHD and stroke are more likely to be affected by ambient PM2.5 in the northern region whereas LRI is more likely to be affected by ambient PM2.5 in the southern region. The major reason for the regional difference in the AF estimates for CHD and stroke would be the higher average ambient PM2.5 concentration in the northern region than in the southern region, because the AF estimates are calculated based on the average ambient PM2.5 concentration. As for the regional difference in the AF estimates for LRI, it is plausible that high relative humidity in the southern region could provide suitable microenvironment for growth of microorganism, which could be absorbed on the surface of ambient PM2.5 and may further increase the risk of LRI when inhaled into the body. Therefore, a same unit increase in ambient PM2.5 in the southern region may have a higher potential to increase the risk of LRI than in the northern region. In addition, the AN estimates for the total expenses of all three diseases in the southern region are all higher than in the northern region. The higher absolute values of total expenses attributable to ambient PM2.5 in the southern region suggest that hospitalized patients in the southern region may have a stronger affordability than in the northern region, which may be determined by local economic development levels, as per capita GDP and average income are higher in the southern region. This finding suggests that additional reduction of ambient PM2.5 would save more expenses in the southern region than in the northern region. Nevertheless, the difference between different regions may also be associated with the severity of the investigated diseases, which is unavailable in the present study, and future studies are encouraged to investigate this issue in detail.

We also find that the AF estimates tend to be higher in males than in females, and also tend to be higher in older adults (≥65 years) subgroup than in younger adults (15-64 years) subgroup. These results provide clues for the susceptible population in the context of high ambient air pollution. Males are likely to have more outdoor activities compared to females due to differences in occupation characteristics, and thus may have higher exposures to ambient PM2.5. The work intensity and activity level of males are also higher than those of females, resulting in increased respiratory rates in males than in females. Therefore, males are likely to absorb more ambient PM2.5 than females under the same air pollution condition. In addition, animal experiments in mice find that male offspring of dams exposing to diesel exhaust (a major source for ambient PM2.5) along with stress restriction during pregnancy have a greater pro-inflammatory change in microglial-derived cytokine levels compared to female offspring in their adulthood, suggesting a potentially higher health susceptibility of male offspring in response to air pollution exposure vs. female offspring. A human exposure study also supports a more apparent increase in inflammation-related gene expression profiles in males than in females after exposure to wood smoke, another source for ambient air pollution. Individuals aged ≥65 years have declined body function compared to younger individuals, and thus would be more vulnerable to ambient air pollution.

This study has several limitations. First, we only use the city level ambient PM2.5 concentrations from fixed-site air monitoring stations which do not reflect personal exposures to ambient PM2.5, and there exists certain exposure measurement error. Nevertheless, it has been shown that fixed-site air monitoring measurements could reflect ambient air pollution levels of the urban background to a certain extent, and the monitoring data have been commonly used as a proxy for population exposure to ambient air pollution in previous time-series analyses,,. Second, this study only covers the health data in 2016 and 2017 and cannot assess the long-term impact of ambient PM2.5 over multiple years. The digitization is happening quite near in China, and it is difficult to get detailed daily hospital admission data over multiple cities in the past. Future studies with daily hospital admission data over multiple years will provide useful information about the changes in AF estimates associated with exposure to ambient air pollution over stages with different air pollution conditions, and help evaluate the effectiveness of air pollution control action implemented by the government. Third, the UEBMI database used in this study covers the urban employee population but does not include children under 15 years old and the population in the rural areas, which may underestimate the health and economic losses associated with short-term exposure to ambient PM2.5 in China. Fourth, some potential confounding variables, such as traffic noise, light pollution, have also been found to affect human health,, but related data are not available in our study. Future studies are encouraged to investigate the potential confounding effects of these factors in the context of ambient air pollution. Finally, this study only estimates the health and economic losses associated with ambient PM2.5 concentrations, and is unable to consider the impact of different chemical components or emission sources of ambient PM2.5, which may vary over different regions of China. Ambient PM2.5 is a mixture consisting of various chemical components with different toxicological potentials from different emission sources, all of which may have contributed to the observed results.

In summary, our study provides evidence for the increased hospital admissions, total expenses and total hospital stays of three cause-specific cardiorespiratory diseases, including LRI, CHD and stroke, in association with short-term exposure to ambient PM2.5 in a large urban population in China. The projected attributable numbers for short-term exposure to ambient PM2.5 would be 362,007 hospital admission cases, 3.68 billion CNY expenses ($US 550 million) and 5.21 million hospital days for the entire urban employee population (299 million) in China during 2016-2017. Our results also show that ambient PM2.5 pollution may lead to higher attributable proportions of health and economic losses for CHD and stroke in the northern region, and higher attributable proportions of health and economic losses for LRI in the southern region. Most existing studies on the economic evaluation related to ambient air pollution are conducted in developed countries, where ambient PM2.5 concentrations are much lower than that in developing countries. Given that billions of people live in the over polluted regions with ambient PM2.5 levels over the WHO guideline, especially in developing countries, the potential economic loss due to short-term exposure to ambient PM2.5 would be enormous over the world. Our findings may provide important evidence for ambient air pollution control action from the angle of input-output (policy cost-benefit) balance especially in developing countries, because accurately estimating the impact of ambient air pollution is crucial to the establishment of effective air quality regulations.

Contributors

YX, HZ, XLF and SW designed the study; YX, ZL, HZ, and PL performed the data analysis; ZX and TG collected the exposure data; YS, JW, LC and CW provided the health data; YX, ZL, HZ and SW wrote the paper; FD, AAB, ZZ, and XG provided critical comments; SW provided study supervision, and all authors approved the paper for submission. The authors declare no conflict of interest.

Data sharing statement

The environmental data used in this study are publicly available. The health data required for the analysis in this study were obtained from the urban employee-based basic medical insurance scheme (UEBMI) of China operated by Ministry of Human Resources and Social Security of the People's Republic of China. The study is carried out under privacy protection. The health data owner confirmed that the data cannot be made to the public due to privacy reasons.

Editor note: The Lancet Group takes a neutral position with respect to territorial claims in published maps and institutional affiliations.

Declaration of Competing Interest

We declare no competing interests.