SOGOBO_2012_v35n3_234.pdf651.0KB

1. Introduction

 East Asia, including Korea, has often been under the influence of windblown dust storms from the deserts of Mongolia and China mainly in spring, and this is referred to as Asian dust storm (ADS) events. ADS events are often accompanied by ambient dust particles of less than 10㎛ in size, and this is known to cause adverse effects on the health of the general public. Such large-scale movement of clouds of sand primarily influences the eastern Asia regions, but can travel as far as to the western regions of the U.S. and Canada [13]. The Asian dust phenomenon has been recorded in a number of historical records of Korea since the 2^nd  century [3], and ADS lasting over 10 days have occurred every year since the year 2000, except in 2003 and 2004 [8].

 With increasing concerns regarding the influence of Asian dust on human health, there have recently been a large number of studies performed on this issue. According to [4], the effects of particulate matters on health are mainly related to the size and chemical composition (chemistry) of the particle. According to the findings of several studies, there is a dramatic increase in the concentration of coarse particles (between 2.5㎛ and 10㎛ in diameter) compared to fine particles (aero-dynamic diameters equal to or less than 2.5㎛) on the days of ADS events. This is associated with high wind speed, which plays a role in decreasing the concentration of fine particles and other combustion-related pollutants. Thus, the PM¹⁰  levels, i.e., the concentration of particulate matter less than 10㎛, increase greatly above the normal levels caused by the local conditions [6]. Studies considering the nature of Asian dust have investigated the effects of ADS on mortality [2, 9, 11], respiratory symptoms [5, 7, 12, 15], cardiovascular diseases [14] and stroke [16].

 Studies performed to date have analyzed the frequencies of diseases and deaths arising from the occurrence of ADS, but there have not been any studies analyzing the frequency of diseases caused by the accuracy of Asian dust forecasts. Thus, the purpose of this study is to analyze the effects of the accuracy of the ADS event forecast on the asthma attacks, a respiratory disease, in Seoul region from 2005 to 2008. For this purpose, the average numbers of medical treatments provided to asthma patients on the index days and the comparison days were compared. The comparative analysis was performed separately on the hit and miss cases, which were determined by whether the ADS forecast provided 24 hours in advance was correct (hit) or incorrect (miss).

2. Material and Methods

2.1 Data

We obtained two sets of data, which are the number of medical services provided to asthma patients at all hospitals and the information associated with the forecasts and occurrence of ADS in the Seoul region from 2005 to 2008. The data were acquired from ‘Health Insurance Review and Assessment Service’ and ‘Korea Meteorological Admini-stration.’ <Table 1> shows the dates of ADS occurrence observed in the Seoul region for a total of 40 days from 2005 to 2008.  

<Table 1> Asian Dust Days in Seoul, Korea, 2005~2008

2.2 Analysis

 This study employs the concept of index days and comparison days used by [2, 16]. Index days are defined as the corresponding days on which windblown ADS events are actually occurred. However, only the last day of 2 or more consecutive days of ADS was considered as the index day unless there was a significant difference (i.e., 110㎛³ ) between the PM¹⁰ concentrations of each day, so as to cope with the harvesting effect. which is individuals who are vulnerable to asthma caused by ADS will visit the hospital on only one of the consecutive ADS days and thus, the number of visits to the clinics may be unexpectedly decreased on the following days. This phenomenon may deteriorate the correlation analysis between ADS and the related asthma diseases [16]. However, if the ADS forecasts were different for the consecutive ADS event days, all the days were considered as the index days.

Comparison days correspond to the symmetrical days without any windblown dust storms, which are 7 days before and after the corresponding index day. The symmetrical comparison days are adopted to control confounding by the day of the week and seasonal trends [1, 2, 16]. We performed paired t-tests on the numbers of asthma patients on the index days and comparison days. If ADS occurred on one of the two comparison days, only the number of medical services provided on the other comparison day was used as data for the paired t-test. If ADS occurred on both comparison days of an index day, they were excluded from the analysis data. Thus, three days from April 20th to 22nd in 2005 satisfying this criterion were excluded from the data. The index days and comparison days selected by the constraint conditions are shown on <Table 2>.

<Table 2> ADS Days Selected for the Analysis

The effects of ADS on human health tend to appear days later and asthma is not a type of disease that must be treated immediately upon onset like that of stroke or heart diseases. Thus, a patient’s visit to the hospital itself may take place a few days after the ADS event. In order to reflect so-called the lag effect, this study adopted a moving average of the number of daily medical services for asthma patients from the ADS event day to the corresponding lag day in the paired t-test. For example, in order to observe the lag effect between the day of ADS occurrence and three days after the occurrence, a 4-day moving average of the number of visits to clinics on the index day, and the next three days as well as a 4-day moving average of the number of medical services on the comparison day and the next three days were used in the paired t-test for analysis. 

In addition, the index days were categorized into two groups depending on whether the ADS forecast issued 24 hours in advance was correct (hit) or incorrect (miss) and a paired t-test was performed on each group to investigate the effects of the ADS forecasts on the frequency of asthma attacks. Also, in order to analyze the patterns of the number of medical services on the hit group and miss group in detail, each group was further divided into two groups according to the 24-hour average PM¹⁰  concentrations and t-tests were carried out on each of the four groups.    

<Table 3> Number of Asthma Patients on ADS Days and Comparison Days in Hit and Miss Groups by Various Lags

3. Results

 The index days were divided into a group for which the ADS forecast issued 24 hours prior to the ADS event was correct (hit) and a group for which the forecast was incorrect (miss). <Table 3> shows the percentage increase in the number of patients on the index days compared to the comparison days as well as the mean value of the moving average of asthma care according to the day lags of the index days and the comparison days for each group. The p-value of <Table 3> shows the results of one-sided paired t-test on the alternative hypothesis (H1), where the average number of medical services was found to be higher on the index days than the comparison days. The p-values of the paired t-test conducted on the corresponding day lags of each forecast group were represented in a graph form as shown in <Figure 1>.

In terms of the entire group, the increase in the number of asthma patients on the index days became statistically significant after the 4-day lag, and there was a pronounced pattern of increase at 5-day lag and 6-day lag. In the hit group, there was an increase in the number of medical services provided for asthma for all day lags of the index days compared to the comparison days, but it was not statistically significant.  

<Figure 1> p-values of Paired t-test for Miss and Hit Groups by Various Lags

On the other hand, in the miss group, the number of medical services provided on the index days at 5-day lag and 6-day lag increased to the extent that they were statistically significant. Thus, the pattern of the p-values at each day lag of each group shows that there is a difference in the pattern of increase in the number of asthma patients depending on the accuracy of the ADS forecasts. However, there was no significant difference between the number of asthma patients on the ADS event days compared to no ADS days in case of 2-4 day lag and there was actually a decrease in the number of asthma patients on the ADS occurrence day and the following day as their p-values were over 0.5. 

 In order to identify the different patterns of increase in the asthma cases between the miss group and the hit group, the groups was divided into sub-groups according to the concentration of Asian dust, i.e., the PM¹⁰  concentration, for analysis. Thus, the hit and miss groups were further divided into groups with PM¹⁰  concentration of less than 110㎛³  and groups with PM¹⁰  concentration of over 110㎛³ , and the average percentage increase in the number of medical services provided for asthma at each day lag of the index days compared to the comparison days was calculated for each group as shown in <Table 4>.

<Table 4> % Increase on Asthma Patients on ADS Days Compared to Comparison Days According to PM10 Concentration Groups by Various Lags

 The (-) sign indicates that there were more medical services on the comparison days than the index days, whereas the p-values were obtained from the t-test conducted to see whether there were any differences in mean values of the percentage increase between the sub-groups divided according to the PM¹⁰  concentrations for the miss and hit groups. In the hit group with  PM¹⁰ concentration of less than 110㎛³ , the rate of incidence of asthma on the index days was lower than on the comparison days from 2-day lag to 6-day lag, whereas in the case of PM¹⁰concentration of over 110㎛³ , ADS events caused approximately 20~27% increase in the incidence of asthma at all day lags. In the miss group with  PM¹⁰ concentration of less than 110㎛³ , there was an approximately 11~33% increase in the incidence of asthma at all day lags of the index days, whereas in the case of  PM¹⁰ concentration of over 110㎛³ , there was a decrease in the incidence of asthma between 0-day lag and 4-day lag from the ADS events. However, ADS events led to dramatic increases of 21% and 34% at 5-day lag and 6-day, respectively. Five days out of the total 11 days for the miss group showed a decrease in the number of asthma on the index days compared to the comparison days. The 24 hr average  PM¹⁰ concentration of the five days was 133.5㎛³ , whereas the average  PM¹⁰ concentration of the other 6 days was 51.3㎛³ . In other words, even though the forecast did not accurately predict the occurrence of ADS, if the concentration of the ADS was so high that people could perceive the dust with the naked eye, they would have refrained from going outdoors on the day of ADS and for shortly afterwards. Also, asthma patients with minor symptoms would have postponed their regular visits to the hospital. This could explain the reason for the decrease in the number of asthma patients on the ADS event days and the following days and for the significant increase in the hospital visits of asthma patients on 6 days after the occurrence of ADS.

The percentage increases in the number of medical treatments for all the hit/miss groups and  PM¹⁰ concentration groups shown in <Table 4> are demonstrated in the form of a graph in <Figure 2> for a comparative analysis. In the hit group, the percentage increase in the number of asthma patients was higher at all day lags when the 24 hr average  PM¹⁰ concentration was above 110㎛³ compared to when it was below 110㎛³. On the other hand, in the miss group, the patterns of increase were different according to the day lag; a pattern that was similar to that of the hit group was seen at 6-day lag, but from the ADS event day up to the 5-day lag, the percentage increase in the incidence of asthma was higher when the 24 hr average  PM¹⁰ concentration was below 110㎛³ compared to when it was above 110㎛³. Such results of the analysis are thought to have been caused by a combinatorial effect of behavior and psychological factors as well as the nature of the ADS forecast and residential environment. When the ADS forecast released 24 hours in advance was accurate, the 24 hr average  PM¹⁰ concentration was found to be 174.6㎛³ on the day of the ADS occurrence. On the other hand, when the ADS forecast failed to predict the incoming of ADS event, the concentration was 103.5㎛³ on the day of the ADS event. This showed that there were relatively higher frequencies of severe ADS in the hit group compared to the miss group. In particular, in cases of severe ADS events with a 24 hr average  PM¹⁰ concentration of over 200, the ADS forecasts were all accurate except for only one day of ADS event. Thus, in case of severe ADS events, even if people decided to refrain from going outside, pollution of the air inside buildings was unavoidable [10] and this is the reason for the relatively higher percentage increase in the number of asthma patients at higher  PM¹⁰ concentrations of the hit group compared to the lower  PM¹⁰ concentration group. Also, accurate forecast for ADS events with lower  PM¹⁰ concentrations was found to significantly reduce the asthma attacks in patients. For the miss group, the pattern of increase in the number of asthma treatments for  PM¹⁰ concentration of below and above 110㎛³ showed the opposite pattern from the hit group, except at 6-day lag. 

4. Conclusions

The findings from the analysis of the total days of ADS occurrence from <Table 3> and <Figure 1> showed that there was a higher number of asthma patients on the index days than the comparison days and this number was statistically significant between the 4-day lag and 6-day lag, which indicates that the lag effect on the incidence of asthma occurs from 4-day lag and onwards. The effects of ADS on asthma were categorized for analysis depending on the accuracy of the ADS forecast provided 24 hours prior to the actual ADS event day in order to identify the impact of the ADS forecast. For this purpose, the patterns of the p-values were obtained from the paired t-tests for the hit group and miss group according to the day lags shown in <Figure 1>, and a comparison showed in the case of hit group, the numbers of asthma patients on the index days was higher than those on the comparison days but it was not statistically significant for all day lags.  

<Figure 2> Increased Percentage of Asthma Patients for 4 Groups Combined with PM¹⁰ Concentration and Accuracy of Forecast

These analysis results present several implications for the future provision of ADS forecasts and their utilization methods. Accurate ADS forecasts significantly reduce the number of asthma patients visiting hospitals in cases of ADS with low PM¹⁰  concentration and can reduce the percentage increase in the number of asthma patients to a certain extent even in cases of severe ADS. Thus, enhancing the accuracy of ADS forecast will significantly contribute to the improvement of public health. Also, even though the ADS forecast issued 24 hours in advance fails to predict ADS event, if efforts are made to inform the public by media that provides real-time information, such as TV, Radio, DMB and smart phone, it will also effectively reduce the incidence of asthma. The real-time information service by various media will produce a greater positive effect especially in the cases of minor ADS events.