Ambient air pollution, weather changes, and outpatient visits for allergic conjunctivitis: A retrospective registry study

Allergic conjunctivitis is a common problem that significantly impairs patients’ quality of life. Whether air pollution serves as a risk factor for the development of allergic conjunctivitis remains elusive. In this paper, we assess the relationship between air pollutants and weather conditions with outpatient visits for allergic conjunctivitis. By using a time-series analysis based on the largest dataset ever assembled to date, we found that the number of outpatient visits for allergic conjunctivitis was significantly correlated with the levels of NO2, O3, and temperature, while its association with humidity was statistically marginal. No associations between PM10, PM2.5, SO2, or wind velocity and outpatient visits were seen. Subgroup analyses showed that sex seemed to modify the effects of humidity on outpatient visits for allergic conjunctivitis, but not for NO2, O3, or temperature. People younger than 40 were found to be susceptible to changes of all four parameters, while those older than 40 were only consistently affected by NO2 levels. Our findings revealed that higher levels of ambient NO2, O3, and temperature increase the chances of outpatient visits for allergic conjunctivitis. Ambient air pollution and weather changes may contribute to the worsening of allergic conjunctivitis.

the lungs, where air pollutants tend to penetrate deeply and stay, the cleansing system of the ocular surface (i.e. tears) may be capable of alleviating the cellular damage induced by air pollutants. On the other hand, the ocular surface is directly exposed to the environment, which means that it may be more susceptible to weather changes.
Taken together, we hypothesize that ambient air pollution as well as weather changes, are both associated with risk for allergic conjunctivitis. To address this issue, population-wide data is crucial. The current study exemplifies a unique situation whereby population-wide outpatient attendance for allergic conjunctivitis in Shanghai was recorded systemically from 2008 to 2012, as well as the information on ambient air pollutants and weather conditions. The availability of such a large dataset allowed us to assess the effects of ambient air pollutants and weather conditions on the prevalence of allergic conjunctivitis with statistical rigor.

Results
The current study included 3,211,820 outpatient visits by 15,938,870 subjects enrolled in SHIS for allergic conjunctivitis. Table 1 and Fig. 1 show the baseline characteristics and the number of outpatient visits for the five-year period. More than three million outpatient visits occurred during the time period examined. The mean age of the patients' onset was 54 years, and 62 percent were women; the number of outpatient visits was highest in people older than 60 years (Fig. 2). Table 2 lists the concentration of air pollutants and weather conditions, while Figs 3 and 4 further demonstrate the changes in these parameters with time.
We decided to convert the outpatient visits from a daily to a weekly measure, as we noticed that the outpatient visits showed strong periodicity by week due to the working hours of the local population. After the conversion, the average weekly outpatient visits ranged from 423 to 2,927 per week (Fig. 5). Although the outpatient visits may be mostly described as a Poisson event, the large mean of the weekly visit can be approximated by a Gaussian distribution (Supplemental file 1).
The outpatient visits for allergic conjunctivitis were significantly correlated with the levels of NO 2 , O 3 , and temperature (Table 3). Among the pollutants we evaluated, every 10 μg/m 3 increased exposure to NO 2 resulted in 61 more outpatient visits per week, whereas 10 μg/m 3 higher O 3 exposure increased the weekly outpatient visits by 21 per week. As for the weather conditions, temperature was the only factor that was significantly associated with outpatient visits. The association between outpatient visits and humidity was statistically marginal. In spite of these factors, our results show no significant association for PM 10 (particles smaller than 10 μm in aerodynamic diameter), PM 2.5 (particles smaller than 2.5 μm in aerodynamic diameter), SO 2 , or wind velocity.
In order to further investigate the interactions among the pollutants and weather conditions, we tested the interactions between each pollutant and the weather condition, and filtered the interactive factors using the stepwise method (Supplemental file 2). As a result, we found significant interactive effects from NO 2 and humidity, which suggests that changes in the weather may also affect outpatient visits. Table 4 shows that a one-unit increment in exposure to NO 2 , O 3 , temperature, and humidity elevated the number of outpatient visits from different subgroups. Subgroup analyses showed that humidity-but not NO 2 , O 3 , or temperature-influenced the number  of outpatient visits for allergic conjunctivitis, in a sex-dependent manner. People younger than 40 years showed susceptibility to changes in all four parameters, while those older than 40 were only consistently correlated with NO 2 . When the data were stratified according to season, the number of outpatient visits for allergic conjunctivitis was more frequent in the spring and summer than in the winter ( Fig. 6 and Table 5). Sex and age did not modify this effect.

Discussion
During the earliest period of China's economic reforms, lasting from the late 1970 s to 1985, automobiles were scarce and energy consumption was quite low, which greatly limited emissions of ambient air pollutants and     greenhouse gases 8 . From 1985 to 2008, however, the Chinese economy expanded rapidly. Although economic growth associated with industrialization has improved health and various quality of life indicators, environmental issues associated with this process are of increasing concern to the country's citizens and its government. Our study has demonstrated the relationship between allergic conjunctivitis and air pollution and weather changes. We showed that the number of outpatient visits with allergic conjunctivitis increased as the levels of NO 2 , O 3 , temperature, and humidity changed. Currently, there are controversies around the association between respiratory allergic disorders and outdoor NO 2 levels. Hwang et al. found that long-term exposure to outdoor NO 2 increased the subjects' risks of persistent cough and phlegm, as well as "current asthma" (i.e. people who confirm that they have been told by a health practitioner that they have asthma, and who confirm that they still have it) 9,10 . Another earlier study, however, reported that the risk of childhood asthma was not associated with levels of NO 2 among 32,672 Taiwanese schoolchildren 11 . Information on the association of air pollution with the prevalence of allergic conjunctivitis is    found that the rhinoconjunctival tissue is sensitive to irritant stimuli during an ongoing allergic inflammation, and that symptoms of allergic rhinoconjunctivitis might be exacerbated in areas with increased levels of air pollutants 12 . A Japanese hospital-based study also showed that outpatient visits for allergic conjunctivitis had a positive association with air pollutants during a three-month investigation period, although such associations were not statistically significant 4 . Interestingly, among several air pollutants investigated in our study, NO 2 and O 3 were the only potent triggers for allergic conjunctivitis exacerbation. The mechanisms underlying the effects of air pollutants on allergic conjunctivitis are not well understood. The study by Lin et al. showed that PM 10 significantly affects the respiratory system 13 . In the present study, however, no association was found for PM 10 with allergic conjunctivitis. We assume that PM 10 and PM 2.5 do not cause a change in pH (as do other aerosol pollutants after they reach the ocular surface) because they can be cleared out of the ocular surface by the cleansing system of tears 14 , whereas in the respiratory system, fine particulate matter penetrates deeply and stays in the lungs. Furthermore, changes in the lacrimal pH-caused by the acidification of tears exposed to a high-oxidant concentration (NO 2 and SO 2 )-could irritate the ocular surface 15,16 . NO 2 and O 3 pose high oxidative potential, and are able to cause damage to the human nasal mucosa 17 ; the effects are also apparent in ocular mucosa 18,19 . NO 2 and O 3 thus may have a higher capacity to provoke conjunctival inflammation than SO 2 via direct oxidative damage to the ocular surface and acidification of the tears in patients with allergic conjunctivitis. In addition, diesel exhaust pollutants, including NO 2 and O 3 , have been reported to enhance the allergic sensitization 20 . Furthermore, NO 2 and O 3 may induce conjunctival inflammation indirectly via chemical modifications of aeroallergens and subsequent enhanced allergic response.
Human health will likely be adversely affected by accelerating changes in weather patterns 7 . Particularly, allergic diseases may be induced and worsened by weather conditions, because of changes in the type, quantity, distribution, and exposure time of pollens, and because of the interaction between pollens and air pollutants 7 . In the present study, the frequency of outpatient attendance of allergic conjunctivitis was inversely correlated with the humidity, and positively correlated with temperature. The correlation between temperature change and pollen distribution has been well described in the literature. Theoretically, increased temperature stimulates earlier flowering and longer pollen seasons 2 . Results have indicated that if temperature increases under a doubled greenhouse gases scenario by the end of the twenty-first century, pollination seasons will start on average one month earlier,  and airborne pollen concentrations will be 50 percent higher than they are today 21 . Meanwhile, low humidity increases land surface evaporation, thus resulting in an increase in airborne pollen concentrations 22 . Pollen sensitivities were indicated as the most frequent triggers for allergic conjunctivitis 23 . Mimura et al. reported that Cedar pollen-specific IgE were significantly higher in tears of allergic conjunctivitis patients than in control subjects 24 . These weather pattern changes, in combination with air pollution, will further add to the burden of allergic disease in exposed populations. For patients with allergic conjunctivitis, the results of our study mean that both air quality and weather conditions play important roles in minimizing exposure and symptoms. Our study also reported the time-lag effect of air pollutants and weather pattern changes on outpatient visits for allergic conjunctivitis. The finding that outpatient visits reached their highest level after three weeks of exposure to pollutants might be of assistance to the public health systems, so that they may better monitor and prepare for environmental events. In addition, the selective and delayed response in patients implies a certain regulatory mechanism that is still not fully understood.
In summary, our study found that higher levels of ambient NO 2 , O 3 , and temperature and lower humidity lead to an increased chance of outpatient visits for allergic conjunctivitis. Our study indicates that ocular surface health may be impaired by long-term exposure to air pollution. These results suggest that efforts to control emissions of both air pollutants and greenhouse gases are warranted. The relationship of these environmental factors to ocular surface health underscores the importance of having all members of the health care team-including health policy decision-makers, specialists, and primary care providers-send the same message to all patients about the necessity of controlling the exposure of the ocular surface to air pollution and adverse weather conditions, as this should help to prevent the onset of allergic conjunctivitis, and should minimize the symptoms and signs in patients who already suffer from the condition.
Strengths and limitations of the study. While our findings are strengthened by the use of a representative, large-population, registry sample, the present study does have several limitations. First, data on pollen and microbial exposure are not available in our study, which are important outdoor environmental factors associated with allergic conjunctivitis 25 . One may suspect that the overlap time between the concentration of air pollutants and allergens, such as pollens, could bias our findings. Our data, however, did not support this possibility. For example, NO 2 concentration peaks in winter while pollens are more in spring and summer (Fig. 4). Our findings showed that NO 2 and O 3 are independent risk factors that add to the risk posed by the pollens. Second, one may raise the concern about using ambient air pollutant concentrations as exposure surrogates, because people spend more time indoors. However, measurements of indoor air pollutants have a large inter-subject variability. In addition, it seems that the infiltration of ambient pollutants to indoor environments occurs under a high ambient fine particle pollution condition 26 . Real-time ventilation measurements for indoor air pollution are recommended to address this issue. Third, as a registry study, we were not able to take the geographic mobility of the population into account during the follow-up period. Fourth, the SHIS is not solely designated for evaluating the prevalence or clinical characteristics of allergic conjunctivitis. As a result, body mass indexes, sociodemographic situations, education levels, and systemic health conditions were not available for further analysis. This is an inherent limitation of studies that use a pre-existing database. Finally, although we could use the patient's specific diagnostic codes, uncertainties were unavoidable because the exact diagnosis of each case could not be confirmed without individuals' medical records.

Allergic Conjunctivitis Data. Data on outpatient visits between January 1, 2008 and December 31, 2012
were obtained from the database of the Shanghai Health Insurance System (SHIS). More than 96 percent of Shanghai's residents-15.9 million-receive SHIS's compulsory universal health insurance. All hospitals in Shanghai are required to contract with the SHIS and submit standard claim documents for medical expenses on a computerized form that includes the date of visit and discharge, identification number, sex, birthday, and the diagnosis for each patient. Outpatient visits for allergic conjunctivitis were selected in the current study according to the diagnosis codes of the International Classification of Diseases, ninth revision (ICD-9). The following codes were included: 372.05 (acute atopic conjunctivitis), 372.13 (vernal conjunctivitis), and 372.14 (other chronic allergic conjunctivitis). Only outpatient visits with the aforementioned ICD-9 codes as the major diagnosis were included in the study. Finally, a GLS model with correlation structure of autoregressive-moving-average (ARMA; p, q) was used to evaluate the association between environmental factors and outpatient attendance. ARMA models are widely used in hydrology 29 , econometrics 30 , and other fields 31 . The ARMA model consists of two parts, an autoregressive (AR) part, represented by p (the order of the autoregressive part) and a moving average (MA) part, represented by q (the order of the moving average part).
The GLS model is written as: where Y is defined as the outpatient attendance of allergic conjunctivitis, X denotes an environmental factor, ε is the residual error (from which we constructed the ARMA [p, q]), Ω is a known matrix of the conditional variance of the error when given X, and β is an unknown coefficient to be estimated. We called the test P of the regression coefficient significant if it was less than 0.05. We used R software version 3.1 (www.r-project.org) for all of the statistical analyses. The R packages we used in this study include "nlme", "zoo", and "car". All P values were based on two-sided tests. In order to determine the time-lag effects of the pollutants, we used the Ljung-Box test to evaluate the significance of correlation between outpatient visits and the concentration of pollutants in the past with time-lags of 1 week to 5 weeks, respectively (Supplemental file 3). The time-lag was then determined as the first time point where the test P value became insignificant (more than 0.1). Most of the pollutants were found to show significant time-lag effects in week 3, therefore we established the ARMA [p = 3, q = 0] except for SO 2 , which had a time-lag of 1 week. To set our age groups, we compared the results of principal component analysis (< 20, 20-50, and > 50 years, Supplemental file 4) with the Chinese hospital standard.for age classification (childhood and adolescence, < 18 years; adult, 18 to 40 years; middle age, 41 to 60 years; and elderly, > 60 years, Table 4) and found that the two classifications did not match. To facilitate the understanding for clinicians and to have a relatively high resolution, we finally decided to use the latter age classification 32 .