An Association between Air Pollution and Daily Outpatient Visits for Respiratory Disease in a Heavy Industry Area

In this work we used daily outpatient data from the Landseed Hospital in a heavily industrial area in northern Taiwan to study the associations between daily outpatient visits and air pollution in the context of a heavily polluted atmospheric environment in Chung-Li area during the period 2007–2011. We test the normality of each data set, control for the confounding factors, and calculate correlation coefficient between the outpatient visits and air pollution and meteorology, and use multiple linear regression analysis to seek significance of these associations. Our results show that temperature and relative humidity tend to be negatively associated with respiratory diseases. NO and are two main air pollutants that are positively associated with respiratory diseases, followed by , , , CO, and . Young outpatients (age 0–15 years) are most sensitive to changing air pollution and meteorology factors, followed by the eldest (age 66 years) and age 16–65 years of outpatients. Outpatients for COPD diseases are most sensitive to air pollution and meteorology factors, followed by allergic rhinitis, asthma, and pneumonia diseases. In the context of sex difference to air pollution and meteorological factors, male outpatients are more sensitive than female outpatients in the 16–65 age groups, while female outpatients are more sensitive than male outpatients in the young 0–15 age groups and in the eldest age groups. In total, female outpatients are more sensitive to air pollution and meteorological factors than male outpatients.


Introduction
Taiwan contains high density of industrial factories using fossil fuel as a main source of energy to power Taiwan's economy [1]. As such, the air above Taiwan naturally contains heavy emissions from both industrial processes and mobile use every day. The air is polluted [2,3]. The industrial fruit is explicitly counted as the shining figures of significant gross industrial output every year. However, the air pollutants suspended in the air affect people who live in the county and the outcomes of polluted air on public health remains to be quantified. [4]. The cost of air pollution on human health needs to be counted as well so as to give a more balance picture of short-term fossil-fuel burning economy fruit with respect to the long-term impact on public health [5]. After all, the main motivation for the economy growth is for the long-term and sustainable welfare of the society. What will be the human health cost reflected in these waxing and waning of transient industrial figures?
Happily, Taiwan has a very good National Health Insurance (NHI) program for a high density population of 23 million [6,7]. The NHI is characterized by good accessibility (Taiwan citizen can see any doctor without referral, and may also go to any levels of hospitals directly), comprehensive coverage (cheap and abundant care), short waiting times, low cost, high coverage rate (99.6% of Taiwan population is covered by NHI in 2010), and a comprehensive and detailed nationwide NHI database that keeps every detailed record of the use of NHI [7,8]. NHI has contracted 92.1% of all hospitals and clinics in Taiwan in 2010 [9].
The NHI database enabled studies of association between air pollution and hospital admissions for cardiovascular disease [6], chronic obstructive pulmonary disease (COPD) [10], asthma [11], lower respiratory tract illness [12], and pneumonia [13]. In these works, NHI database provides quantified hospital admissions for above mentioned diseases, which were then analyzed against ambient air pollution levels obtained from a network of ambient air monitoring stations operated by Taiwan Environmental Protection Administration (EPA) [2]. People often get very ill to get admissions to hospitals. Mortality rate is another measure frequently used to quantify the impact of air pollution on public health [4]. Most of the time people visit hospitals to seek brief consultation and examination on their health. The results of these hospital outpatient visits are kept and comprising a large set of medical service records in NHI computer database [8]. This big data set reflects the status of the health of a population. As such, the use of outpatient data has become an important tool to understand the association between air pollution and public health [14,15,16,17,18].
Previous studies use NHI database to focus on the association between air pollution and hospital admissions. For example, to associate air pollution and hospital admissions for asthma during the period 1996-2003 [11], and for cardiovascular disease during the period 1997-2001 [6] in Taipei; and to associate air pollution and hospital admissions for pneumonia in Kaohsiung during the period 1996-2004 [13]. A few studies reported association between daily outpatient visits and daily air pollution levels. For example, Hwang and Chan [12] used the NHI database to study effects of air pollution on daily clinic visits for lower respiratory tract illness in 50 townships in Taiwan during 1998. In this work we used outpatient data from a hospital in a heavily industrial area in northern Taiwan to study the association between daily outpatient visits and daily air pollution levels in the context of a heavily polluted atmospheric environment during the period 2007-2011.

Ethics Statement
The statistics of daily hospital visits used in this work is openly published each year by the hospital [19,20,21,22]. Only the frequencies of visits for diseases are used in the study. No details of patient's personal information are involved in this work. As such, no ethics committee approving is needed. No written consent from patients, or from the next of kin, caretakers, or guardians on the behalf of minors/children parents are needed.

(WS)
Model12 OP = b 0 +b 1 (PM 10 )+b 2 (PM 2:5 )+b 3 (O 3 )+b 4 (CO)+b 5 (NO)+b 6 (NO 2 )+b 7 (SO 2 )+ b 8 (TEMP)+b 9 (RAIN)+b 10 (WD)+b 11 (WS)+b 12  air monitoring data, the combination of local community outpatient visit data with air pollution monitoring data makes it ideal to look for the association between ambient air pollution and public health. A list of International Classification of Diseases, 9th Revision (ICD-9), for respiratory system were selected from outpatient visit database for this work. There are four diseases from which data of daily outpatient visits were analyzed and their correlations with daily air pollutants and meteorology were calculated. These diseases are allergic rhinitis (477), asthma (493), pneumonia (480-486), and COPD (490-493).
The hospital contains very detailed records of daily outpatients for each of the disease category. These data extend the period 2007-2011. The daily outpatient data for each disease category are divided into three age groups: ages 0-15, 16-65, and above 66 years old. The divisions of age groups are similar to Hajat et al. [23], Hajat et al. [24], and Jalaludin et al. [18].
Let's denote x ijk as a one-dimensional vector, representing a time series of daily outpatient numbers for each i disease category at each j age group and each k year. Hence, we can write x ijk as following,

Ambient Air Pollution and Meteorological Data
In this work we use ambient air pollution and meteorological data obtained from Chung-Li station (close to the hospital), which is one of the more than 80 ambient air monitoring stations operated by Taiwan EPA [2,25]. This network of ambient air monitoring stations has started its current scale of operation since 1994. Each station reports hourly measurements of air pollutants  and meteorology. A total of 12 observational variables from Chung-Li station are used in this work. These variables contain 7 main air pollutants (PM 10 , PM 2:5 , O 3 , CO, NO, NO 2 , and SO 2 ) and 5 key meteorological observations (temperature, precipitation, wind direction, wind speed, and relative humidity). Let's denote y lk as a vector for these air pollution and meteorological variables. Hence, we can write

Test of Normality for Hospital, Air Pollution, and Meteorological Data
The daily hospital outpatient data is continuous in day, but the hospital does not provide regular outpatient service during weekends (Saturday and Sunday), and other public holidays (Lunar New Year, Moon Festival, Dragon Boat Festival, etc). As such, data that occurred during the public holidays are removed to prevent artificial bias (zero outpatient visits) introduced by these holidays (holiday effect) in the calculation of correlation coefficients. Air pollution and meteorological data during the public holidays are also removed accordingly to avoid holiday effect. We note that if the holiday effect not removed from the time-series data, the correlation coefficients exhibit (not shown here) overwhelmingly associations between the diseases and NO, NO 2 , CO, and SO 2 .
We then compute observed frequency distribution, which is calculated by binning daily data into 15 bin intervals for each year and for each data. The 15-bin intervals is empirically determined to give sufficient resolution of the data when testing data normality with respect to the continuous normal distribution.
Given an observed frequency distribution one can calculate its mean value and standard deviation. With a mean value and its standard deviation, one can calculate an expected frequency distribution of data if the data is normally distributed [26]. Finally, we can access the normality of observed data by testing the observed frequency distribution against its theoretical normal distribution using statistical methods (e.g. the Chi-square test) following Press et al. [26].

Correlation Coefficient Analysis
Given the outpatient vector x n and variable vector y n , we compute correlation coefficient following Press et al. [26]. A measure of significance of this correlation coefficient r is obtained by computing the Student's t probability P following Press et al. [26]. The small values of P indicates a significant correlation.
In addition, meteorological factors that might play a confounding role are separately considered as warm days (temperature above 25 0 C) and cool days (temperature below 25 0 C) in this work, following the method proposed by Chang et al. [6], Cheng et al. [13], and Yang et al. [11]. Hwang and Chan [12] also include temperature as a confounding variable in their analysis. Since warm days are often associated with high relative humidity and cool days are often associated with low relative humidity [27], the stratification of days between warm and cool days also separates the effect of relative humidity from high to low.

Multiple Linear Regression Model Analysis
In this work we use multiple linear regression analysis to account for confounding effects when associating daily hospital visits for diseases with air pollutants and meteorological factors. Multiple linear regression analysis has been widely used to control for confounding factors when predicting dependent variables from independent variables [28,29,30,31,32,33,34].
We have developed multiple linear regression models for association analysis following the methods described in Kleinbaum et al. [35], which contains detailed mathematical formulas for the multiple linear regression models, equations to compute statistical significance of regression results, and a rich example data set that demonstrate the results from various multiple linear regression models. Kleinbaum et al. [35] use SAS package to carry out multiple linear regression computation. However, in order to facilitate the efficiency of processing large data set from medical and meteorological sources, it is necessary to build computational models from the methods described in Kleinbaum et al. [35].
The multiple linear regression models we have developed for this work have been carefully and successfully tested against data from Kleinbaum et al. [35]. We have run our models through data in Table 5.1, which shows observations on systolic blood pressure (SBP) and age for a sample of 30 individuals. We have compared our model results with edited SAS output for Table 5.1 in page 52 of Kleinbaum et al. [35]. We have also run our model through data shown in Table 8.1, containing a list of weight, height, and age of a random sample of 12 nutritionally deficient children. Our models were compared with a list of edited SAS output from 6 multiple regression models (shown in pages 124-126). Our detailed comparisons show consistent results between SAS calculations and our model calculations. These comparisons demonstrate our models are accurate when compared with the SAS package. We then use these multiple models to run through the medical and the meteorological data. Table 2 shows a typical example of our multiple linear regression analysis. A list of PM 10 -based multiple linear regression models that were systematically used to run the medical and meteorological data. The purpose is to find the best model that associate air pollutant PM 10 with outpatient visits for each of respiratory diseases discussed in this work. Following Kleinbaum et al. [35], we use R 2 and statistical significance level (P value) to select the best model for our analysis. Table 2 shows a list of PM 10 -based model analysis. Similarly, we have systematically run twelve multiple linear regression models based on PM 2:5 , O 3 , CO, NO, NO 2 , SO 2 , temperature, rainfall, wind direction, wind speed, and relative humidity, respectively. As such, a total of 144 models were run for each of respiratory disease, in each year of 2007-2011, considering male and female outpatients separately, and for each of three age groups of outpatients (0-15, 16-65, and above 66 years). Hence, we have completed a total of 17,280 model analysis (4 respiratory diseases |144 models/disease |5 years |2 sex groups/year/age group    |3 age groups/sex). The results that are statistically most significant are shown in the following section. Figure 2 shows daily outpatient visits for the 0-15, 16-65, and above 66 year-old age group of people for each of the respiratory system diseases during the 2007-2011 period. For the age 0-15 group of outpatients (Figure 2a), most of the daily visits were for pneumonia disease, followed by allergic rhinitis, asthma, and COPD. Peak visits for pneumonia disease during January-April, while the period with low pneumonia outpatients occurred during summer months of June and July. For allergic rhinitis disease, more daily hospital visits occurred during late February to early May, July-August, and November-December than other months. The first and the last periods are normally associated with poor air quality [36], while the summer months of July August may be associated with high relative humidity.

Daily Outpatients For Respiratory System Diseases During 2007-2011
For the 16-65 age group outpatients (Figure 2b), most of the hospital visits were for COPD disease, followed by asthma and allergic rhinitis. Few visits were for pneumonia. For COPD outpatients, more visits occurred during winter to spring months than the summer months. For the group of outpatients with ages above 66, their visits are mainly associated with COPD and asthma (Figure 2c). Table 3 gives a summary of the total number of outpatient visits for each disease of the respiratory system during 2007-2011. Taiwan is under the strong influence of the southerly Asian summer monsoon flow in the summer season. This monsoon flow brings clean tropical air to the entire county, resulting in a much reduced ambient air pollution in summer [36].    (Figure 4(g)).
The medium values of daily maximum SO 2 are centered about 10 ppbv. Large variabilities of daily maximum SO 2 close to 20 ppbv frequently occurred all year round. Daily maximum temperatures follow the marching of the season, with high temperatures occurred during July-August and low temperatures occurred during January-February (Figure 4(f)). Temperatures in summer months show less variability than those occurred in the winter to spring months.
The observations show that the variations of daily maxima for CO and NO are smaller and the minimum concentrations for both species are higher during June-August months than during the rest of other months. Since the dominating flows during the June-August months are coming from the south directions [36], the persistently high minimum CO and NO levels indicate sustained contribution from local pollution sources. The ability to reach low CO and NO during the rest of other months indicate the effect of large-scale flows from the north directions to bring daily maximum CO and NO to low concentration. This indicates that daily maximum PM 10 in most days of each year are much higher that those WHO guidelines. This is consistent with the time-series analysis shown in Figure 3   The wind directions show three modes of peak frequency distribution centered around 40 0 , 140 0 , and 270 0 directions. The highest frequency distribution of wind direction come from around 140 0 directions, indicating southerly winds. This is an indication of winds mostly occur during summer monsoon season when winds were from the south directions. The wind directions centered around 270 0 representing westerly winds, which normally occur during the day-time period when the sea breeze prevailed over the Taoyuan area (Figure 1(b)). The winds coming from around the 40 0 directions representing the prevailing northerly winds which occur mostly during the late fall, winter, and early spring seasons. period. The outpatient visits shows close resemblance to the expected frequency distribution from normal distribution. Figure 9(a) shows distribution of correlation coefficients when correlating hospital visits for allergic rhinitis with air pollutants and meteorology for three age groups during the 2007-2011 periods. Figure 9(b) shows a measure of significance for the correlation coefficients calculated in Figure 9(a). The small values in Figure 9(b) indicate significant correlation [26].

Test of Normality for Air Pollutant and Meteorology Data
The 0-15 age group of outpatients for allergic rhinitis are mostly and positively correlated with PM 10 and PM 2.5 . Temperature shows significantly negative correlations for this young age group of outpatients. Temperatures do not exhibit a clear association with both the 16-65 and above 66 age groups of outpatients. Air pollutants such as PM 10 , PM 2.5 O 3 , CO, NO, NO 2 , and SO 2 tend to show positive correlations with visits for allergic rhinitis when compared with meteorological factors. These pollutants have been shown to be related to allergic rhinitis [17].
When controlled for temperature effect, both cool (Figure 9(c)) and warm (Figure 9(e)) days also exhibit positive correlation coefficients. However, the correlation coefficients during cool days are more pronounced and higher than those during warm days. Also the significance of correlation occurred during the cool days (Figure 9(f)) are stronger than those during the warm days (Figure 9(d)). This could indicate low temperatures can act to acerbate the symptoms allergic rhinitis. Distribution of correlation coefficients during cool days (Figure 9(e)) are more pronounced than the effects of temperature were not controlled (Figure 9(a)). Hence, there is a confounding effect of temperature when  We note that positive correlations may not exit through all 5 years. For example, allergic rhinitis with PM 10 in 2008 are negative while the rest of other four years are positive for the age 0-15 group of outpatients (Figure 9(a)). The majority of the correlation coefficients are positive.
4.2. Asthma. Frequency distribution of daily outpatient visits for asthma is shown in Figure 10. These daily outpatient distribution shows remarkably close resemblance to the expected normal distribution in ages 16-65 and above 66 groups of people. The largest group of outpatients is from the 16-65 age group of people. Significant increases in the mean values of daily outpatient visits are seen in people with ages 16-65 and above 66.
Air pollutants such as PM 10 , PM 2.5 , and NO 2 exhibit positive correlations with the daily outpatient visits for asthma (Figure 11(a)). These correlations are very significant (Figure 11(b)), especially for outpatients with age above 66 years old. Our findings of strong and significant correlations between asthma and PM 10 and PM 2.5 are consistent with Schwartz et al. [39] and Sheppard et al. [40]. Both studies showed the daily counts of emergency room visits persons under age 65 and non-elderly were significantly associated with PM 10 on the previous day in Seattle. Sunyer et al. [41] found daily admissions for asthma in adults increased significantly with increasing ambient levels of NO 2 in four European cities. McConnell et al. [42] found positive associations between air pollutants (PM 10 and NO 2 ) and bronchitic symptoms in Southern California children with asthma. On the other hand, decrease in particulate air pollution were associated with decrease in asthma and bronchitis admissions of children [43] and mortality rates [44].
Outpatients for asthma disease negatively correlated with temperature and these correlations are significant. This result is consistent with Whittemore and Korn [45], who showed more asthma attacks on cool days in the Los Angeles area. Asthma outpatients also show negative correlations with wind direction. This indicates westerly winds tend to have lower asthma outpatients, while easterly winds tend to have higher number of asthma outpatients. Easterly winds are associated with winds from the ocean directions, while easterly winds are associated with winds from inland directions. More industrial factories and transportation vehicles are located from inland directions than from ocean directions.
For asthma, age 0-15 group of outpatients shows more prevailing positive correlations with PM 10 , PM 2.5 , O 3 , CO, NO, and NO 2 during warm days (Figure 11(c)) than during cold days (Figure 11  pollutants during cool days than during days. For outpatients above age 65, occurrence of positive correlation coefficients during both warm and cold cool days are similar. These results indicate that temperature can play a delicate effect on asthma outpatients. Cool days appear to reduce the effect of air pollutants on asthma symptoms for age 0-15 of outpatients, while effect of air pollutants enhanced during warm days. For older outpatients, age 16-65, effect of air pollutants on asthma symptoms are all enhanced during cool days than during warm days. For the eldest outpatients, age about 66, temperature effects are less clear when comparing cool with warm days. The appearance of highly correlated coefficients also exhibit high levels of significance, Figures 11(d) and 11(f). These results also indicate the confounding effects of temperatures on the association of daily asthma outpatients with air pollution levels.
4.3. Pneumonia. Frequency distribution of outpatients for pneumonia disease are shown in Figure 12 for three age group of people during the 2007-2011 period. Most outpatients for pneumonia concentrate on the 0-15 year-old people. The frequency distribution of these outpatients exhibit a good agreement with the expected normal distribution.
Pneumonia outpatients are positively correlated with PM 10 , PM 2.5 , O 3 , and NO 2 for age 0-15 group of outpatients (Figure 13(a)). High values of the these positive correlation coefficients often associated with low P-values (Figure 13(b)), indicating that the correlations are significant. These results are consistent with previous findings. For example, Penna and Duchiade [46] found a statistically significant association between the average annual levels of particulates and infant mortality from pneumonia. Schwartz [47] found causal association between the increasing particulate air pollution and increase in mortality rates from COPD and pneumonia. In a study of air pollution and infant mortality in Mexico City, Loomis et al. [48] showed that NO 2 and O 3 are also associated with the infant mortality, but not as consistently as with particulate air pollution.
For pneumonia, cool days ( Figure 13(e)) generally exhibit a strong effect on the distribution of air pollutant correlated coefficients than during warm days (Figure 13(c)) and during all days (Figure 13(a)). For the youngest pneumonia outpatients (age 0-15), effects of PM 10 , PM 2.5 , and O 3 show strongly positive correlation coefficients. Effects of CO, NO, NO 2 , and SO 2 correlated coefficients are higher during the cool days than during the warm days. These correlations are significant (Figures 13(d) and 13(f)) and similar during cool and warm days. For the older outpatients (ages 16-65, and above 66), effects of air pollutants are significantly enhanced during the cool days than during the warm days.
Our results show the confounding effects of temperature are quite complicated. Effects of PM 10 , PM 2.5 , and O 3 on pneumonia outpatients are less sensitive to temperature for the youngest group Table 4. List of PM 10 -based Regression Coefficients, F value, R 2 , and P Value. of people but sensitive to other two older groups. Effects of CO, NO, NO 2 , and SO 2 are all sensitive to temperatures for all three age groups of outpatients. 4.4. COPD. Figure 14 shows frequency distribution of outpatient visits for COPD. The observed frequency distribution of outpatient visits are close to theoretical normal distribution for ages 16-65 and above 66 group of outpatients.
Outpatient visits for COPD shows overwhelmingly negative correlation with temperature ( Figure 15(a)), and these correlations are very significant for thee age groups of outpatients. Positive correlations are seen for PM 10 , PM 2.5 , and NO 2 . These correlations are very significant (Figure 15(b)). Positive correlations also seen for O 3 and NO but are not as persistently presented as for particulate air pollution and NO 2 . Previous works have shown positive association between particulate air pollution and COPD [47,49].
For COPD, the youngest outpatients (age 0-15) are more correlated with air pollutants during the warm days (Figure 15(c)) than during the cool days (Figure 15(e)). However, for the other two older group of outpatients (ages 16-65, and above 66), both warm days and cool days show similar distribution of correlation coefficients. COPD are mostly correlated with air pollutants for the oldest outpatients (above 66) and these correlations are significant (Figures 15(d) and 15(f)).

Multiple Linear Regression Model Analysis
As shown in Table 2, a list of twelve multiple linear regression models were used to find best association between outpatient visits and air pollutants and meteorological factors. Table 4 shows a list of typical results when PM 10 -based models were run one by one through the 0-15 age group of male outpatients for allergic rhinitis disease in 2010. These results show that complex models produces statistically significant results (as seen from F value, R 2 , and P value) compared with simple models. Regression results are similar when PM 10 , PM 2:5 , and O 3 are three major independent variables. Inclusion of CO as an additional independent model variable increase R 2 , followed by inclusion of another independent model variable NO 2 . The most significant increase in R 2 occurs when temperature is included as an independent variable in the model. Addition of relative humidity, rainfall, and winds also increase R 2 and the statistical significance of our model analysis. Similar results consistently occur when run multiple linear regression models were run through other disease, age group of outpatients, air pollutants, and meteorological factors. These multiple linear regression results are also consistent with the confounding effects of temperature from correlation coefficient analysis shown in the previous section. Table 5 shows multiple linear regression analysis of daily outpatients for allergic rhinitis. Our 12-parameter multiple linear regression models shows that the 0-15 age group of male outpatient visits are statistically significant in the 2007 (P = 0.0014), 2008 (P = 0.0001), and 2010 (P = 0.0046) data. NO is the main pollutant and wind speeds is the main meteorological factor that are persistently and positively associated with the increase of outpatient visits. Relative humidity is persistently and negatively associated with the outpatient visits. Low relative humidity predicts higher outpatient visits than high relative humidity. For the female outpatients, only the 2010 data showing statistically significant association (P = 0.0002). Positive associations occur with PM 10 , O 3 , NO, NO 2 , and wind speeds. Negative associations occur with temperature and relative humidity.
For the 16-65 age group of outpatients, multiple linear regression model shows only results for male outpatients in the 2007 data are statistically significant (P = 0.0012).In this case, Table 5. Multiple Linear Regression Analysis of Allergic Rhinitis Outpatient Visits: Regression Coefficients, F value, R 2 , and P Value. PM 10 , PM 2:5 , NO, and temperature positively predicts the increase of outpatient visits. Rainfall rates and relative humidity are negatively associated with outpatient visits. For outpatients with ages greater than 66, statistically significant association between outpatient visits and air pollutants and meteorological factors occur mostly in the 2011 data for female outpatients (P = 0.0042). Positive increase in air pollutants PM 10 , O 3 , CO, and NO 2 are associated with increase in outpatient visits. Temperature and relative humidity are negatively associated with the outpatient visits. Table 6 shows multiple linear regression analysis of outpatient visits for asthma. For  Table 6. Multiple Linear Regression Analysis of Asthma Outpatient Visits Regression Coefficients, F value, R 2 , and P Value. visits, while temperature and relative humidity are negatively associated with outpatient visits. Table 7 shows multiple linear regression analysis of outpatient visits for pneumonia.  Table 8 shows multiple linear regression anaysis of associations of outpatient visits for COPD with respect to air pollutants and meteorological factors. Statistically significant association for the age 0-15 group of male outpatients occur only in the 2008 data (P = 0.0013). In this case, PM 2:5 , NO and NO 2 are positively while temperature and relative humidity are negatively associated with outpatient visits. Statistically significant associations for female outpatients also occur only in the 2008 data (P = 0.0007). Female COPD outpatients are positively associated with PM 10 , NO, and NO 2 ; and negatively associated with temperature and relative humidity. For the eldest group of male outpatients (age greater than 66 years old), statistically significant associations occur only in the Table 7. Multiple Linear Regression Analysis of Pneumonia Outpatient Visits Regression Coefficients, F value, R 2 , and P Value.

Summary
Taiwan contains high density of industrial factories, polluted air, and a comprehensive National Health Insurance program that covers population and contracted 92.1% of all hospitals and clinics in Taiwan. In this work we used outpatient data from a hospital in a heavily industrial area in Taiwan to study the association between public health and air pollution in the context of a heavily polluted atmospheric environment during the period 2007-2011. We used hospital visit data from the Landseed Hospital, and air pollution and meteorological data from EPA to study the association between air pollution levels, meteorology, and hospital visit frequency for the period 2007-2011 over Chung-Li area. We test the normality of each data set, control for the confounding factors, and calculate correlation coefficient between the outpatient visits and air pollution and meteorology, and use multiple linear regression analysis to seek significance of these associations. Table 9 shows a summary of list of main pollutants and meteorological factors that are statistically significant in associating with the respiratory diseases from multiple linear regression analysis. Temperature and relative humidity tend to be negatively associated with respiratory diseases. NO and NO 2 are two main air pollutants that are positively associated with respiratory diseases, followed by PM 10 , PM 2:5 , O 3 , CO, and SO 2 . Young Table 8. Multiple Linear Regression Analysis of COPD Outpatient Visits Regression Coefficients, F value, R 2 , and P Value. outpatients (age 0-15 years) are most sensitive to changing air pollution and meteorology factors, followed by the eldest (age §66 years) and age 16-65 years of outpatients. Outpatients for COPD diseases are most sensitive to air pollution and meteorology factors, followed by allergic rhinitis, asthma, and pneumonia diseases. In the context of sex difference to air pollution and meteorological factors, male outpatients are more sensitive than female outpatients in the 16-65 age groups, while female outpatients are more sensitive than male outpatients in the young 0-15 age groups and in the eldest age groups. In total, female outpatients are more sensitive to air pollution and meteorological factors than male outpatients.