Fine Particulate Air Pollution and First Hospital Admissions for Ischemic Stroke in Beijing, China

The primary objective of this study was to assess the association between short-term changes in ambient fine particulate matter (PM2.5) and first hospital admissions for ischemic stroke. We identified 63,956 first hospital admissions for ischemic stroke from the Beijing Medical Claim Data for Employees from January 1, 2010, through June 30, 2012. A generalized additive Poisson model was applied to explore the association between PM2.5 and admissions for ischemic stroke. We also explore the effect modification of risk by age and gender. The exposure–response relationship between PM2.5 and admissions for ischemic stroke was approximately linear, with a relatively stable response at lower concentrations (<100 μg/m3) and a steeper response at higher concentrations. A 10 μg/m3 increase in the same-day PM2.5 concentration was associated with 0.31% (95% CI, 0.17–0.45%, P < 1.57 × 10−5) increase in the daily admissions for ischemic stroke. The association was also statistically significant at lag 1, 2, 3, 0–2 and 0–4 days. Subgroup analyses showed that the association was not different between patients ≥65 years and <65 years old or between males and females. In conclusion, short-term exposure to PM2.5 was positively associated with first hospital admissions for ischemic stroke in Beijing, China.

(particles with an aerodynamic diameter ≤10 μm) were significantly associated with increased risk of ischemic stroke 13 . However, only a mere handful of studies have been undertaken which specifically evaluated the health effects of PM 2.5 on ischemic stroke, due primarily to the lack of PM 2.5 monitoring data. Researchers have suggested that PM 2.5 was more harmful to human health than PM 10 , because PM 2.5 can penetrate more deeply into the lung and carry larger concentrations of adsorbed or condensed toxic air pollutants per unit mass with its greater surface area 14,15 . In addition, previous studies were primarily conducted in Western developed countries; however, because of the considerable differences in the level of pollution, weather patterns, and population susceptibility, the association between PM 2.5 and ischemic stroke is still unclear in developing countries.
The primary interest outcome in prior studies that evaluated the short-term effects of PM 2.5 on ischemic stroke hospitalizations was the daily total number of admissions regardless of first or repeated admissions, and all admissions were treated as independent events. A meta-analysis reported that 1-year and 5-year stroke recurrence rates were 11.1% and 26.4%, respectively 16 . Thus, admissions for recurrent stroke may account for a considerable proportion of daily hospital admissions for stroke. While first-ever and recurrent stroke share similar risk factors and clinical manifestations, their clinical characteristics and pathogenetic features are not exactly the same 17,18 . Furthermore, several studies indicated that a history of the disease had considerable influence on the association between air pollution and the specific health outcome 7,19 . On the other hand, repeated admissions could cause a temporal dependence among the hospitalization counts, leading to an underestimation of the variance of air pollution risk estimates 20 . Based on these findings, it seems plausible that air pollution may have a differential effect on first-ever vs. recurrent stroke events. Therefore, studies that specifically examine the association between PM 2.5 and first-ever stroke are needed to better understand the real impact of ambient fine particulate matter on public health.
The primary objective of this study was to examine the short-term effect of PM 2.5 on first admissions for ischemic stroke in Beijing, China.

Methods
Data collection. Data on daily admissions for ischemic stroke from January 1, 2010, through June 30, 2012, was obtained from Beijing Medical Claim Data for Employees (BMCDE). Medical claim data for all working or retired employees who are covered by basic medical insurance in Beijing were recorded in BMCDE. The medical information recorded in BMCDE includes basic demographics (age and sex), admission date, medication use, clinical diagnosis in Chinese and corresponding International Classification of Diseases, 10th Revision (ICD-10) codes, and reimbursement information. The information on brain computed tomography or magnetic resonance imaging use was also recorded in the database. Hospital admissions for ischemic stroke were identified according to the principal diagnosis, using the ICD-10 code of I63. Specifically, subjects who had been hospitalized for stroke (I60-I64) during the five years preceding the index event were excluded, and we focused on the first admissions for ischemic stroke in this study. We also used the corresponding Chinese diagnoses to check the identified cases, ensuring that only the first admissions could be included in this study.
The air pollution data, hourly PM 2.5 concentration from January 1, 2010, through June 30, 2012, was obtained from a web platform (http://www.stateair.net/web/historical/1/1.html) run by the US embassy, which established an ambient air quality monitoring station on the rooftop of embassy building located in Chaoyang district, Beijing. A prior study has indicated that the PM 2.5 concentrations obtained from the monitoring station exhibited approximately the same trend as citywide PM 2.5 observations 21 . Furthermore, 79.2% of Beijing's total population resided within a 40-km radius of the U.S. embassy ambient monitoring station. All areas of high population density (>5000 people/km 2 ), 97.8% (44/45) of the tertiary hospitals and 79.3% (69/87) of the secondary hospitals in Beijing that admit ischemic stroke cases located within a 40-km radius of the monitoring station. It has been suggested that the monitoring data could be used as a good proxy for personal exposure among individuals residing <40 km from the monitoring station 8,14 . In addition, only urban residents were included in this study to further reduce exposure misclassification. To allow for the effects of weather conditions, meteorological data on daily 24-hour average temperature (°C) and relative humidity (%) was obtained from the Chinese Meteorological Bureau over the same period.
Statistical Analysis. Daily average PM 2.5 concentration, daily admissions for ischemic stroke, and weather conditions were linked by date and, thus could be analyzed with a time-series design. Because daily hospital admission for ischemic stroke was rare, a generalized additive Poisson model was applied to explore the association between PM 2.5 and ischemic stroke, after adjusting for day of week (DOW), calendar time, public holiday, and daily average temperature and relative humidity. The penalized spline (ps) function of calendar time with 10 degrees of freedom (df) was used to adjust for seasonality and long-term trends 8 . We also used the ps functions of daily average temperature (df = 3) and relative humidity (df = 3) to allow for the potential nonlinear confounding effects of weather conditions 22,23 . The DOW and public holiday were also incorporated in the model to control for the difference in the baseline hospital admission rates for each day. After the basic model was established, the variable of PM 2.5 concentration was introduced. The final model was described below: where, E(Y t ) is the expected number of admissions for ischemic stroke on day t; α is the model intercept; PM2.5 t−i is the mean PM 2.5 concentration on day t, and i is the day lag; β represents regression coefficient; DOW is the day of the week; public holiday is a binary variable indicating a public holiday on day t (coded as 0 indicates no holiday, and 1 indicates a holiday); ps() indicates penalized spline function; Temp 0 and RH 0 indicate the daily mean temperature and relative humidity on the current day, respectively. The results are presented as the percentage change and 95% confidence interval (CI) in the daily ischemic stroke admissions per 10 μg/m 3 increase in PM 2.5 concentration.
Smoothing function was used to graphically analyze the exposure-response relationship between the log-relative risk of ischemic stroke and PM 2.5 concentration.
To examine the temporal association of PM 2.5 concentration with ischemic stroke admission, we fitted the models with different lag structures from the current day (lag0) up to 4 lag days (lag4). Considering that single-day lag models may underestimate the effect of pollutant 24 , we also evaluated the cumulative effects using 3-day (lag0-2) and 5-day (lag0-4) moving averages of PM 2.5 concentrations.
Stratified analyses were conducted to examine whether the association differed by age (≥65 years and <65 years) and gender. The Z-test was applied to test the statistical significance of differences by age or gender 25 . We also conducted sensitivity analyses to examine the robustness of the results in terms of the df in the smooth function of time trend (8)(9)(10)(11)(12), daily mean temperature (2-6) and daily relative humidity (2-6).
All statistical analyses were carried out using R Programming Language (V.3.2.2, R Development Core Team) with the "mgcv" and "nlme" packages. All statistical tests were two-sided, and tests of statistical significance were set at P < 0.05. Table 1 summarizes the basic characteristics for our study. A total of 63,956 first hospital admissions for ischemic stroke were identified from BMCDE database between January 1, 2010 and June 30, 2012. There were 66.5% male patients, and 56.7% patients were ≥65 years old. The mean (standard deviation, SD) age of the ischemic stroke patients was 66.4 (12.1) years.

Results
The summary statistics for daily ischemic stroke admissions, air pollution and weather conditions from January 1, 2010 to June 30, 2012 are shown in Table 2. The mean daily count for ischemic stroke admission was 70 during the study period. The overall mean daily PM 2.5 concentration was 99.5 μg/m 3 , with a range from 7.2 to 492.8 μg/m 3 . According to the CAAQS Grade II standards for daily mean concentrations of PM 2.5 (75 μg/ m 3 ), 45.4% (414 days) of the daily PM 2.5 concentrations were up to the standard. However, in terms of the WHO Air Quality Guidelines for 24-hour average PM 2.5 concentration (25 μg/m 3 ), only 13.6% (124 days) days met the standard. The means (SD) of temperature and relative humidity were 12.6 °C (11.6 °C) and 48.6% (20.3%), respectively.
Stratified analyses showed that the association between PM 2.5 and ischemic stroke was not different between patients ≥65 years and <65 years old or between males and females (P > 0.05) ( Table 4).

Variable
No.   Table 5 shows the percentage changes in ischemic stroke admissions in relation to a 10 μg/m 3 increase in PM 2.5 , under different df for calendar time, temperature and relative humidity. The risk estimates were stable when df for calendar time was 10, while different df for temperature and relative humidity had little effect on the estimates, suggesting that the results on the association of PM 2.5 concentration with ischemic stroke admissions was robust in this study.

Discussion
The present study provides strong evidence of the association between PM 2.5 and ischemic stroke in Beijing. The short-term exposure to PM 2.5 was significantly associated with the first admission for ischemic stroke accounting for temperature, relative humidity, day of week, public holiday, long-term trends and seasonality of stroke events. To the best of our knowledge, this is the first city-level investigation of the relationship between PM 2.5 and first admission for ischemic stroke in a real and severe air pollution environment. Although the magnitude of increased risk of ischemic stroke due to PM 2.5 exposure was relatively small, the number of ischemic stroke events attributable to PM 2.5 may be high due to the high incidence of ischemic stroke and the fact that the overwhelming majority of the public is exposed to ambient fine particulate matter, suggesting potentially large public health implications. Figure 1. The smoothed exposure-response curves of daily average PM 2.5 concentrations at current-day against ischemic stroke hospital admission. Note: The X-axis is the current-day (lag 0 day) fine particulate matter (PM 2.5 ) concentrations (mg/m 3 ). Y-axis is the predicted log (relative risk (RR), after adjusting for calendar time, day of the week, public holiday, current-day temperature, and relative humidity, is shown by the solid line, and the dotted lines represent the 95% CI.    There has been numerous epidemiological and experiments research suggesting a possible link between air pollution and higher risk of ischemic stroke 26 . Some pathophysiological hypotheses have been proposed to explain the possible association between short-term effects of air pollution and ischemic stroke. Researchers have suggested that exposure to air pollution had adverse effects on vascular endothelial function, increased activity of the sympathetic nervous system and systemic inflammation, resulting in vasoconstriction, increased plasma viscosity, and increased risk of blood clotting and thrombosis [27][28][29] . These pathophysiological responses could potentially place individuals at higher risk of ischemic stroke. However, direct evidence on the association between PM 2.5 and ischemic stroke was limited and the findings remain equivocal. A comprehensive review identified only six studies published before January 2014 that specifically examined the association between short-term exposure to PM 2.5 and the risk of ischemic stroke hospitalization, and the pooled results suggested a non-significant increased risk of hospital admission for ischemic stroke per 10 μg/m 3 increase in PM 2.5 30 . Wellenius et al. investigated association between short-term ambient air pollution and risk of ischemic stroke in 1,705 patients admitted from Boston, USA. They found that PM 2.5 was positively associated with ischemic stroke for the <24 hours lag period 31 . However, a study conducted in Canada, including 9,202 admissions for acute ischemic stroke, found no association between short-term increases in PM 2.5 and ischemic stroke risk 32 . The heterogeneity of results across studies may due to differences in demographic characteristics of study population, meteorological patterns, and PM 2.5 levels.
In this study, the estimate scaled to a 10 μg/m 3 increase in PM 2.5 was relatively lower when compared to previous reports. A meta-analysis that combined the risk estimates derived from previous studies reported a 1.0% increase in the daily ischemic stroke admissions per 10 μg/m 3 increase in the level of PM 2. 5 30 . The differences in the definition of health outcomes across studies may be responsible for the weaker effect of PM 2.5 observed in this study. Considerable attention has been focused on overall hospitalizations including admissions for both first-ever and recurrent stroke in previous studies [32][33][34] , possibly combining effects with different sensitivities to PM 2.5 . In contrast, only the first hospital admissions for ischemic stroke were included in this study. Epidemiological evidence indicated that individuals with a history of cardiovascular disease were more susceptible to air pollution. Patients with myocardial infarction 35 , congestive heart disease 36, 37 , or chronic obstructive pulmonary disease 38 were demonstrated to have higher risk of death on days with heavy air pollution. Stroke survivors were at higher risk of recurrent ischemic stroke in relation to increased level of PM 10 39 . Therefore, it is reasonable to suppose that PM 2.5 may exert greater adverse effect on recurrent stroke than first-ever stroke, which may partly explain the relatively weaker effect observed in this study. Differences in characteristics of study population and pollutant, meteorological pattern, and exposure assessment strategy are also likely to contribute to the heterogeneity across studies. Given that this is the first study to evaluate the association between PM 2.5 and first admissions for ischemic stroke in a badly polluted environment, the findings should be interpreted with caution and future studies are warranted to confirm the current results for PM 2.5 .
Identifying the dose-response relationship for ischemic stroke in relation to PM 2.5 concentration is of public health and regulatory interest. Nevertheless, as prior studies were primarily conducted in Western countries with slight PM 2.5 pollution, the exposure-response relationship in a real and severe air pollution environment remains unclear. In this study, we conducted a dose-response analysis to explore the pattern and scope of the adverse response. We observed an approximately linear exposure-response relationship, which is consistent with a time-stratified case-crossover study conducted in USA that indicated a linear relationship between PM 2.5 and risk of acute ischemic stroke 31 . It is worth noting that, in that study, even in concentration ranges well below the present US National Ambient Air Quality Standards (20 μg/m 3 ), PM 2.5 was also significantly associated with increased risk of acute ischemic stroke, which is contrast to a relatively shallower response at low concentrations  Table 5. Percentage change with 95% CI in ischemic stroke admissions associated with a 10 μg/m 3 increase in fine particulate matter (PM 2.5 ) concentration on the same day, by different degree of freedom (df) for calendar time, temperature, and relative humidity. *The df value used in this study model. in our study. A recent time-series study conducted in China evaluated the exposure-response relationship for respiratory emergency visits in relation to fine particulate air pollution 40 . The exposure-response curve was virtually flat at low levels of PM 2.5 (<200 μg/m 3 ) and became sharp at high levels, which is consistent with our results. Our findings were also supported by a study involving 369,469 ischemic heart disease cases suggesting that PM 2.5 at levels below 75 μg/m 3 would not significantly increase the risk of ischemic heart disease 8 . Based on these findings, we hypothesized that there might be a threshold concentration at which PM 2.5 become harmful enough to impose an adverse impact on the development and progression of ischemic stroke. This hypothesis may be able to partly explain the non-significant association between short-term exposure to PM 2.5 and ischemic stroke observed in other areas where the PM 2.5 concentrations were too low to have significant effects on ischemic stroke 32,34 . Future studies are needed to clarify this quite important issue. Possibly because of the higher PM 2.5 concentration in Beijing, a stronger temporal association was noted in this study when compared to prior reports of a 1-2 day lag 31,33 . Nevertheless, it is important to note that the pollution exposure assessment was based on the date of hospital admission rather than the time of stroke onset in this study, resulting in a substantial exposure misclassification and underestimates of pollutants effects 41 . A study conducted in USA, including 1,101 hospital admissions for acute ischemic stroke, found that hospital admission occurred at a median of one calendar day after onset of symptoms, and this delay could cause an underestimation of the association between ambient particulate matter exposure and stroke 42 . Therefore, the lag effects of PM 2.5 exposure and ischemic stroke should be interpreted cautiously.
Examining the modification effects of individual characteristics will help to identify susceptible population and develop specific interventions for subgroups. We did not observe significant effect modification of risk by age, which is in consistent with a study indicating that air pollution risk estimates did not differ substantially between patients ≤65 years and >65 years for both ischemic and hemorrhagic stroke 43 . However, the American Heart Association made an updated scientific statement on the association between particulate matter air pollution and cardiovascular disease, suggesting that the elderly are more sensitive to the elevated PM 2.5 level; however, the evidence is limited 9 . Elderly people are very likely to spend a greater proportion of time indoors or wear a face mask outdoors when PM 2.5 pollution is severe, thus decreasing personal exposure 44 , which may cover up the real age-specific effects of PM 2.5 . The gender subgroup analysis showed that the associations with PM 2.5 were not different between men and women, which are in line with previous reports 43,45 .
Compared with prior studies that focused on the association in developed countries where PM 2.5 pollution was mild, our study was conducted in a heavily polluted city. Given the high level of pollution, we were able to evaluate the exposure-response relationship in a wide range of PM 2.5 concentration and have an opportunity to reveal a more complete picture of the association. Additionally, we specifically used the first admissions as the health outcome, providing a novel insight into the underlying mechanism of the association.
This study was subject to certain limitations. First, the use of air pollution data deriving entirely from one single monitoring station is expected to cause exposure measurement error, leading to underestimate the pollutant effects 46 . Second, we had no access to monitoring data on other air pollutants, such as sulfur dioxide, nitrogen dioxide, carbon monoxide and ozone, thus limiting our ability to explore the independent effect of PM 2.5 . Therefore, our findings should be interpreted with caution, and additional studies are warranted to examine the independent effect of PM 2.5 on first hospital admission for ischemic stroke. Another limitation is our inability to differentiate the ischemic stroke subtypes, because that information was not available in our database. Future studies are needed to examine whether the acute effects of PM 2.5 differed across strata defined by ischemic stroke etiology. Finally, the retrospective data collection may bring about bias from diagnostic and coding inaccuracy. However, both ICD-10 codes and corresponding Chinese diagnoses were used to identify eligible ischemic stroke hospitalizations, which significantly reduced bias from coding inaccuracy 47,48 . Given the robustness of the association in all lag models, stratified analyses and sensitivity analyses, and large sample size in the present study, these potential limitations are unlikely to have compromised our results.
In conclusions, our results indicated that short term elevation of PM 2.5 was associated with increased risk of ischemic stroke among populations in Beijing. The present study contributed to the limited scientific literature about the short-term effects of particulate matter air pollution on ischemic stroke in developing countries and additional research on this topic is warranted.