Relationship between emergency care utilization, ambient temperature, and the pollution standard index in Taiwan

Abstract

This study applied a vector error correction model to investigate the effects of ambient temperature (AT) and air quality index values on emergency care utilization (ECU). The Pollution Standards Index (PSI) and total suspended particulates (TSP) were used for analysis. Data were obtained from the National Health Insurance Research Database of the Ministry of Transportation and Communications and Ministry Environmental of Protection Administration of Taiwan. Data from January of 1998 to December of 2012 (180 months) were analyzed. Study results showed that, regardless of long-term equilibrium or short-term dynamics, a 1 °C increase in AT will decrease ECU, showing that AT strongly affects ECU. There were no significant corrections of long-term equilibrium of PSI and TSP on ECU. Only short-term TSP dynamics caused negative effects in the first ECU phase. Emergency care requires special monitoring of AT and TSP to respond to the increased number of high-risk patients consulting emergency departments.

Keywords:

• Climate change
• vector error correction model
• emergency care
• ambient temperature
• pollution standard index
• total suspended particulate

Introduction

The effect of climate change on health is an important topic in the field of environmental health. Many studies consider temperature and atmospheric pollution as risk factors for poor physical and psychological well-being, and most of the existing literature has explored the correlations between high or low temperatures and the incidence of disease and mortality rates (Wang et al. 2012a; Benmarhnia et al. 2014; Zhang et al. 2014). It is well known that atmospheric pollution affects health and that exposure to atmospheric pollutants such as sulfur dioxide (SO₂), nitrogen dioxide (NO₂), ozone (O₃), and carbon monoxide (CO) can aggravate respiratory diseases and increase the incidence of cardiovascular diseases and the associated mortality rates (Atkinson et al. 2012; Brandt et al. 2012; Shah et al. 2013).

Very few studies have discussed the combined effects of temperature and atmospheric pollutants on emergency care utilization (ECU). This is because the relationship between temperature and atmospheric pollution has not yet been clearly defined (Benmarhnia et al. 2014). Recent studies have pointed out that the probability of deaths attributable to moderately hot and cold temperatures was higher than for those caused by extreme temperatures (Gasparrini et al. 2015). This shows that investigations into the effects of moderate temperatures and air pollution on ECU are needed.

Emergency care is defined as the immediate and urgent treatment of patients to save lives, slow the course of disease, salvage extremities, or maintain vital signs. According to the Taiwan National Health Insurance statistics (2013), each hospital serves an average of 39.3 emergency patients per day. However, the annual number of patients who received emergency treatment increased from 5,883,886 in 1999 to 7,101,327 in 2013.

In 2013, the resourced-based points for emergency medical expenditures were 17,978,177.351 thousand points, representing approximately NT$17,978,177,351(US$599,272,578) reimbursed to hospitals (National Health Insurance Administration 2013). Existing literature asserts that an increase in the concentration of suspended particulates will increase the incidence of respiratory diseases (Arbex et al. 2014; Kashima et al. 2014). In addition, an increase in the concentrations of air pollutants such as SO₂, NO₂, CO, and O₃ impacts the incidence of circulatory system diseases and related hospitalization and death rates (Wong et al. 2008; Beard et al. 2012; Kan et al. 2012). In 2013, the reported points for emergency medical expenditures in Taiwan were 1,462,953.929 thousand points (8.14 %) and 1,636,922.431 thousand points (9.11 %) for circulatory and respiratory system diseases, respectively. This indicates that the primary diseases related to emergency treatments were circulatory and respiratory system diseases.

However, the relationships among emergency care, temperature, and atmospheric pollution have rarely been considered and discussed in global literature as well as in Taiwan. Based on the concept of the epidemiologic triangle model, there is interactivity between agent (A), environment (E), and host (H). As such, environmental changes (including temperature and air pollution) may facilitate the spread of agents or the consequent incidence of diseases. In Taiwan, which is situated in a sub-tropical zone, there could be a strong link between the effects of temperature and atmospheric pollutants on ECU. Furthermore, variations in air pollution sources may necessitate different forms of ECU.

Therefore, the main purpose of this study is to explore the relationships among ECU, ambient temperature (AT), and atmospheric pollution measured by the Pollution Standards Index (PSI) or total suspended particulates (TSP) using a vector error correction model (VECM). There are several advantages using the VECM for our analyses. The VECM restricts the long-run behavior of the endogenous variables to converge to their cointegrating relationship while allowing for short-term dynamics. Thus, we can bypass endogeneity problems in our target variables (ECU, AT, and atmospheric pollution), which are theoretically interdependent based on the epidemiologic triangle model. We also can evaluate the mutual effects of the average temperature of subtropical regions, air pollution index, and short- and long-term effects on the number of emergency department patients. We hope that this study can serve as a reference for public health policies and short- and long-term air pollution strategies.

Method

Data and data source

Data pertaining to ECU, calculated using data per 10,000 people per month, were retrieved from the National Health Insurance Research Database. These data spanned 180 months from January of 1998 to December of 2012. AT data were retrieved from the monthly averages of the AT database for the Taiwan region, as provided by the Ministry of Transportation and Communications (Ministry of Transportation and Communications [MOTC] 2015); the data also spanned 180 months from January of 1998 to December of 2012.

The PSI data can indicate the increased risk of chronic obstructive pulmonary disease, heart disease, asthma, and so on (To et al. 2013; Szyszkowicz & Kousha 2014; Zheng et al. 2015). This is particularly true for air pollutants smaller than PM2.5 (PM2.5: fine particles with diameters of 2.5 microns or less), which was found to increase the phenomenon of premature mortality (Lelieveld et al. 2015) or cerebral hemorrhage (Huang et al. 2017) as well as the number of emergency department visits (Fan et al. 2016; Lim et al. 2016). This study will investigate how two air pollution markers, PSI and TSP, affect ECU.

The PSI is an index for monitoring atmospheric pollution adopted for use in Taiwan in 1993. It is based on the concentrations of five pollutants: air suspended particles (PM10: fine particles with diameters of 10 microns or less), SO₂, NO₂, CO, and O₃. These concentrations are converted into a value between 0 and 500, which reflects air quality as follows: good (0–50), average (51–100), bad (101–199), very bad (200–299), and harmful (≥ 300). These data were sourced from the PSI value database provided by the Ministry Environmental of Protection Administration (Ministry of Transportation and Communications (MOTC) 2014) spanning the 180-month period from January of 1998 to December of 2012. The Taiwanese government has adopted a simple and easily comprehensible method for informing the public about air quality through PSI values. The main purpose is to remind those suffering from allergies and circulatory and respiratory system diseases to take precautions as well as for the public to pay attention to air quality when engaging in outdoor activities. In this study, the impact of PSI values on ECU was examined.

TSP is a general term that refers to the number of suspended particles in the air. Particles with sizes above and below PM10 can be absorbed by the human respiratory system and are jointly referred to as TSP. The number of particulates in the air is measured, and a standard exists with respect to permissible levels of pollution. Taiwan has a 24-h mean TSP concentration of 250 μg/m³ and a mean annual value of 130 μg/m³. Data were sourced from the TSP database provided by the Ministry Environmental of Protection Administration spanning the 180-month period from January of 1998 to December of 2012 (Ministry of Transportation and Communications [MOTC] 2015).

Statistical methods

The EViews 9 software was used for VECM analysis. The VECM proposed by Johansen (1995) was employed to observe the dynamic relationships among three factors (AT, PSI, TSP) affecting the ECU for empirical research. The VECM is often used to predict interdependent time series systems and to analyze dynamic impulses from random interference in a system.

VECM is a model derived from vector autoregression (VAR). The ECU, AT, and PSI temporal data are considered non-stationary data (Nelson & Plosser 1982). VAR views all these variables as endogenous variables to avoid the bias caused by the mutual interactions among AT, PSI, and others. However, using VAR will cause errors when setting the model and the omission of important long-term data. Therefore, a VECM with an error correction mechanism can enable observations of the before-and-after short-term relationship between variables. In the long term, this model enables observations of the long-term relationship of errors with the overall variable regression method; feedback effects between variables can be obtained under short- and long-term interactions (Engle & Granger 1987). Recently, researchers have employed VECM statistical methods to study air pollution (Costilla-Esquivel et al. 2014; Norrulashikin et al. 2015). As far as we know, few studies have employed VECM to investigate the relationship between air pollution and health. To ensure rigor, our experimental steps were carried out based on the empirical steps for VECMs recommended by Enders (Enders 2004).

First, a unit root test was used to verify the stability of four-time series data of ECU, AT, PSI, and TSP. These four-time series data were input into the empirical model using natural logarithmic conversion. Therefore, during unit root test operation, verification was carried out on variables that had undergone natural logarithmic conversion. If these four natural logarithmic conversion variables are constant data (with no unit root characteristics), then empirical analysis was directly carried out using the VECM. If these three natural logarithmic variables are non-constant data (with unit root characteristics), then a further step of first-order difference unit root verification was carried out on various natural logarithmic conversion variables to ensure that the orders of integration of these variables were all first-order, and cointegration testing was carried out among these natural logarithmic conversion variables. This study used the augmented Dickey–Fuller test (ADF) test (Dickey & Fuller 1979) for unit root testing to ensure that the unit root characteristics and order of integration of these various natural logarithmic conversion variables were all first-order integrals.

Secondly, if these natural logarithmic conversion variables all have unit root characteristics and are all first-order integrals, then the Johansen trace test and maximum eigenvalue test (Johansen 1995) were employed for cointegration testing. If the cointegration phenomenon occurred among the natural logarithmic conversion variables, then an error correction term must be added to the vector self-regression model when carrying out VECM analysis to form a VECM. The VECM can be written as equation:

where Δ is the first differences operator, and ΔECU_t is the vector matrix of the endogenous variable, i.e. the rate of change of ECU. γ is the intercept. The subscripts t and j represent time and deferred period, respectively. β_i is a coefficient. α is the adjustment vector that represents the adjustment speed of short-term disequilibrium to long-term equilibrium ECM_t−1 is the cointegration vector that represents the long-term equilibrium relationships among ECU, PSI, AT, and TSP. ξ_t is a purely white noise term.

Results

Table 1 shows the results of the descriptive analysis. The mean monthly ECU frequency per 10,000 people was 2114.74 (Standard Deviation [SD]: 301.4); mean AT was 23.52 °C (SD: 4.34 °C); PSI was 3.49 (SD: 2.61); and mean TSP was 93 μg/m3 (SD: 22.21 μg/m3). This shows that the air quality conditions in Taiwan, as indicated by the PSI, were within the national air quality standards (0–50). Figure 1 shows the distribution of ECU, AT, PSI, and TSP from 1998 to 2012.

Table 1. Variables descriptive analysis.

Variables	ECU	AT	PSI	TSP
Mean	2114.74	23.52	3.49	93.00
Median	2101.03	24.27	2.89	93.82
Maximum	3220.83	29.93	14.46	147.12
Minimum	1467.78	14.75	0.11	47.00
Standard Deviation	301.40	4.34	2.61	22.21
Observations	180	180	180	180

Notes:

ECU was emergency care utilization. AT was ambient temperature. PSI was pollution standard index. TSP was total suspended particulate (μg/m3).

Figure 1. Emergency care utilization (ECU), Ambient temperature (AT), pollution standard index (PSI) and Total suspended particulates (TSP) the from 1998 to 2012.

Notes: Emergency care utilization was ECU. Ambient temperature was AT. Pollution standard index was PSI. Total suspended particulates was TSP.

The unit root test was used to verify whether ECU, AT, PSI, or TSP possessed a stationary series. If the null hypothesis was verified in the ADF test, this indicated that the series had a difference stationary process; if the null hypothesis was not rejected, then the series was considered non-stationary. The first-difference unit root test was then used to perform the unit root test to confirm whether the time series belonged to the I(1) series. Once the I(1) series was confirmed, cointegration tests were performed to verify whether a long-term relationship existed between our target variables. It can be seen from the results of the unit root test in Table 2 that the demeaned time series data of ECU, PSI, and TSP demonstrate the unit root property at term level while the detrended time series data of AT, PSI, and TSP illustrated the unit root property at level term. Nevertheless, when we took the first difference of ECU, AT, PSI, and TSP and used these series to conduct the unit root test, we obtained unambiguous results, and the null hypothesis of the existence unit root was rejected. These results suggest ECU, AT, PSI, and TSP belonged to the I(1) series.

Table 2. ADF unit root test.

Variables	Level				1st difference				Order of integration
	Intercept		Trend and intercept		Intercept		Trend and intercept
	t-statistic	p-values	t-statistic	p-values	t-statistic	p-values	t-statistic	p-values
ECU	− 2.014	0.28	− 9.513	<0.01	− 5.82	<0.01	− 5.844	<0.01	I(1)
AT	− 3.003	0.04	− 3.406	0.05	− 10.327	<0.01	− 10.314	<0.01	I(1)
PSI	− 1.396	0.58	− 2.216	0.48	− 9.275	<0.01	− 9.237	<0.01	I(1)
TSP	0.309	0.978	− 6.267	<0.01	− 8.726	5.08E-13	− 8.701	<0.01	I(1)

Notes:

ECU was emergency care utilization. T was ambient temperature. PSI was pollution standard index. TSP was total suspended particulate (μg/m³).

Table 3 shows the results of the Johansen cointegration test, which was used to establish whether the ECU–AT–PSI and ECU–AT–TSP variables had long-running equilibrium relationships. The trace test and maximum eigenvalue test showed that ECU–AT–PSI and ECU–AT–TSP had one cointegration relation at a 5 % significance level. ** denotes rejection of the hypothersis at the 5% significance level.

Table 3. Johansen cointegration test.

Variables	Trace test		Max-eigenvalue test
	Statistic	Prob.**	Statistic	Prob.**
ECU-AT-PSI	72.640	0.00	64.778	0.00
ECU-AT-TSP	110.674	0.00	106.061	0.00

Note:

ECU was emergency care utilization. AT was ambient temperature. PSI was pollution standard index. TSP was total suspended particulate (μg/m3).

** denotes rejection of the hypothersis at the 5% significance level.

Table 4 shows the VECM results using ECU with AT and PSI. From these results, we can represent the long-term equilibrium relationships among ECU, AT, and PSI samples for a 180-month period as ECU = 47715.150 − 1931.06 (AT) + 11.825 (PSI) + residual. For the relationship of ECU with AT and PSI, the adjustment coefficients in long-term equilibrium were 1931.06 and 11.825, respectively. The number of ECU cases per 10,000 people decreased by 1931.06 resource-based points (around NT$1931 or US$64) in relation to a 1 °C increase in mean temperature but increased by 11.825 resource-based points (around NT$12 or US$0.4) in relation to a per-unit increase in PSI. Regarding the long-term equilibrium relationships, AT had the greatest effect on ECU, followed by PSI. The adjustment factor values of AT (− 0.00045) and PSI (0.00024) were not large but had a negative correlation at a 10 % significance level, indicating that the amplitude of the corrected long-term equilibrium value was not large. For the short-run dynamics, the ECU, AT, and PSI for one month were each subjected to short-term significant effects from the preceding four months. ECU was negatively influenced by PSI and positively influenced by AT. Each monthly decrease in ECU cases per 10,000 people was positively affected by the first-, second-, and fourth-week AT average in the same month. ECU was not affected by PSI.

Table 4. VAC model results using ECU with AT and PSI.

	ECU	AT	PSI
Long-term equilibrium relationship
Cointegrating vector	1.000	1931.069 [14.556]	− 11.825 [− 0.110]
Adjustment factor	− 0.012 [− 1.816]	− 0.00045 [− 13.250]	0.00024 [3.095]
Cointegrating Equation	ECU = 47715.150 – 1931.06 × (AT) + 11.825 × (PSI) + residual
Short-run dynamics
D(ECU(-1))	− 0.498 [− 6.243]	0.001 [3.14]	− 0.000177 [− 0.183]
D(ECU(-2))	− 0.357 [− 4.108]	0.001 [2.218]	0.000455 [0.431]
D(ECU(-3))	− 0.298 [− 3.508]	0.001 [1.690]	− 0.000283 [− 0.274]
D(ECU(-4))	− 0.167 [− 2.353]	0.001 [2.200]	0.001 [1.068]
D(AT(-1))	− 35.498 [− 3.960]	0.414 [8.740]	− 0.401 [− 3.693]
D(AT(-2))	15.570 [1.517]	0.642 [11.848]	− 0.075 [− 0.605]
D(AT(-3))	27.095 [1.972]	0.536 [7.385]	− 0.262 [− 1.573]
D(AT(-4))	− 1.992 [− 0.134]	0.334 [4.268]	− 0.334 [− 1.862]
D(PSI(-1))	− 9.835 [− 1.515]	− 0.030 [− 0.870]	− 0.681 [− 8.654]
D(PSI(-2))	− 7.247 [− 0.948]	− 0.073 [− 1.825]	− 0.469 [− 5.061]
D(PSI(-3))	5.056 [0.653]	− 0.109 [− 2.669]	− 0.430 [− 4.590]
D(PSI(-4))	8.658 [1.317]	− 0.064 [− 1.832]	− 0.250 [− 3.144]
Constant term	6.167 [0.403]	− 0.123 [− 1.529]	− 0.017 [− 0.092]

Notes:

ECU was emergency care utilization. AT was ambient temperature. PSI was pollution standard index. [ ] is the z value, black fonts represent that the significance level of 5 % or better.

Table 5 shows the VAC model results using ECU with AT and TSP. The long-term equilibrium relationship of samples between ECU with AT and TSP in a 180-month period was ECU = 42,561.88 − 1698.664 (AT) − 3.221 (TSP) + residual. The number of ECU cases per 10,000 people decreased by 1698.664 in relation to a 1 °C increase in mean temperature and decreased by 3.221 in relation to a 1 μg/m3 increase in TSP. However, the adjustment factor values of AT (− 0.000548) and TSP (0.001621) were not large, indicating that the amplitude of the corrected long-term equilibrium value was not large. For the short-run dynamics, ECU, AT, and TSP for one month were each subjected to short-term significant effects from the preceding four months. ECU was influenced negatively by TSP and positively by AT. Each monthly decrease in ECU cases per 10,000 people was negatively influenced by the first-week average AT and the first-week TSP in the same month.

Table 5. VAC model results using ECU with AT and TSP.

	ECU	AT	TSP
Long-term equilibrium relationship
Cointegration vector	1.000	1698.664 [14.6643]	3.221 [0.33405]
Adjustment vector	− 0.008 [− 1.03]	− 0.000548 [− 13.631]	0.001621 [3.595]
Cointegrating Equation	ECU = 2561.88 – 1698.664 × (T) − 3.221 × (TSP) + residual
Short-run dynamics
D(ECU(-1))	− 0.506 [− 6.439]	0.001 [3.74460]	− 0.001 [− 0.29640]
D(ECU(-2))	− 0.363 [− 4.14369]	0.001 [1.84606]	0.007 [1.44998]
D(ECU(-3))	− 0.268 [− 3.10790]	0.000 [0.81391]	0.003 [0.60357]
D(ECU(-4))	− 0.148 [− 2.12168]	0.001 [1.91383]	0.004 [1.13443]
D(AT(-1))	− 39.795 [− 4.30690]	0.411 [8.86801]	− 4.575 [− 8.79286]
D(AT(-2))	− 0.386 [− 0.03258]	0.627 [10.5507]	− 3.085 [− 4.62878]
D(AT(-3))	15.376 [0.99524]	0.514 [6.62313]	− 1.772 [− 2.03717]
D(AT(-4))	− 1.003 [− 0.06431]	0.266 [3.40092]	− 2.203 [− 2.50830]
D(TSP(-1))	− 4.227 [− 3.05574]	− 0.004 [− 0.55444]	− 0.529 [− 6.78582]
D(TSP(-2))	− 1.900 [− 1.26853]	− 0.007 [− 0.89157]	− 0.347 [− 4.10795]
D(TSP(-3))	0.567 [0.39879]	− 0.034 [− 4.73215]	− 0.448 [− 5.59713]
D(TSP(-4))	0.499 [0.37793]	− 0.026 [− 3.99605]	− 0.211 [− 2.83667]
Constant term	5.685 [0.37799]	− 0.132 [− 1.75142]	− 0.383 [− 0.45190]

Notes:

ECU was emergency care utilization. AT was ambient temperature. TSP was total suspended particulate (μg/m3). [ ] is the z value, black fonts represent that the significance level of 5 % or better.

Discussion

This study confirmed the influence of AT and PSI on changes in the number of patients requiring ECU. It also discussed the influence of AT and TSP on ECU. In terms of AT, both cointegrating equations indicated that an increase in AT by 1 °C reduced the number of people requiring ECU. To illustrate this, we used the mean reported points per 10,000 people who went for emergency consultation in Taiwan in 2013 (710.599 thousand points). Using the proposed impact of AT and PSI on ECU, for a similar change in AT, the starting points per 10,000 people for emergency consultation decreased by 137.22 thousand points. This shows that temperature changes can be used as an important marker for disease prevention (Szyszkowicz 2017). Previous studies investigating U-, V-, and J-shaped temperature changes and mortality rates have shown differences in mortality rates caused by temperature variations in different regions (Huynen et al. 2001; Curriero et al. 2002). In Western European countries, a rise in temperature was followed by an increase in mortality rates or ECU (Lippmann et al. 2013; Tong et al. 2014). In contrast, in subtropical regions, lower temperatures were linked to an increase in ECU because lower temperatures affected the incidence circulatory and respiratory system-related diseases (Wang et al. 2012b; Lin et al. 2013; Yi & Chan 2015). Additionally, previous studies have pointed out that moderately hot and cool temperatures increased mortality rates while the effects of extreme temperatures were lower (Gasparrini et al. 2015). Moreover, with regard to elderly people, exposure to low temperatures increased mortality rates owing to respiratory and circulatory system diseases (Lin et al. 2011). Studies examining mortality rate and temperature have clearly suggested that differences in temperatures may contribute to differences in the mortality rates of various diseases and regional mortality rates.

However, our results can improve our understanding of the effects of subtropical AT on ECU. Results of studies on long-term equilibrium relationship showed that a 1 °C increase in AT can decrease the number of people utilizing emergency care. As adjustment factor values are not large, their effects on long-term equilibrium relationship are few. However, from results of studies on short-run dynamics, we can see that AT has relatively more significant effects on ECU. Our study results are similar to those of studies investigating temperature in subtropical regions and ECU, i.e. temperature increases result in decreased ECU. Results from a study in Shanghai showed that residents of subtropical areas can adapt to hot weather and have self-defense mechanisms, thus decreasing the risk of exposure in high temperatures; therefore, patients with non-severe symptoms may decrease ECU (Sun et al. 2014). A study from Taiwan showed that high temperature and mortality rates are significantly correlated; however, emergency room visits were not significant (Wang et al. 2012c). Low temperatures are a threat in subtropical regions. Studies examining the association between extreme temperatures and emergency room visits have found that low temperatures affected the incidence of diseases such as asthma, respiratory diseases, cerebrovascular disease, and mortality rates (Lin et al. 2011; Wang et al. 2012a; Wang & Lin 2014). This finding is similar to those of previous studies showing that low temperatures increased ECU. Previous studies have used different temperature markers to predict differences in mortality rates and number of outpatient visits (Lin et al. 2012). This study only used average temperatures instead of extreme temperatures for analysis, showing that the study results may have limitations, and subsequent studies should continue to investigate whether different temperature markers can affect ECU. These results show that the use of health alerts is necessary when considering the short-term effects of low temperatures in subtropical regions with respect to AT-related public health initiatives.

The study results indicate that a one-unit increase in PSI will increase ECU. We illustrated this using the mean reported points per 10,000 people who went for emergency consultation in Taiwan in 2013 (710.599 thousand points). Using the proposed impact of AT and PSI on ECU, the reported points per 10,000 people who received emergency consultation increased by 0.842 thousand points when the PSI value increased by one unit. PSI values were related to five air pollutants: PM10, SO₂, NO₂, CO, and O₃. Current research results indicate that PSI concentrations increase the incidence, hospitalization, and mortality rates for cardiovascular and circulatory diseases (Chen et al. 2004; Wong et al. 2008; Kan et al. 2012). Increases in the amounts of SO₂, NO₂, CO, O₃, and PM2.5, determined by the PSI can lead to increased ECU (Cao et al. 2009), especially for female patients over 65 years of age with ischemic heart disease (Lin & Kuo 2013), acute ischemic stroke patients (Chen et al. 2014), and children with asthma (Lavigne et al. 2012). The results of this study are similar to those of existing research. Although an increase in the PSI led to an increase in ECU, the effects were insignificant in relation to the short-run dynamics for ECU.

We investigated the effects of another TSP marker on ECU. With regard to the results of short-term effects, the biggest difference between TSP and PSI is that TSP affected ECU, but PSI did not. The results of short-term effects supported the results of previous studies. Increase in concentrations of suspended particulates induced asthma, respiratory diseases, pneumonia, and so on, and increased the number of emergency department visits (Arbex et al. 2014; Kashima et al. 2014; Michikawa et al. 2015; Lim et al. 2016; Noh et al. 2016). However, with regard to long-term effects, the difference from previous study results was that an increase in TSP concentrations led to a decrease in ECU. Current studies have found that, although an increase in TSP concentrations can aggravate lower respiratory symptoms in adult asthmatic patients, this effect was moderate (Watanabe et al. 2011). Again, this did not result in any significant correction to long-term equilibrium and was only significant in the short term. We illustrated this using the mean reported points per 10,000 people received emergency consultation in Taiwan in 2013 (710.599 thousand points). Using the proposed impact of AT and TSP on ECU, the reported points per 10,000 people who received emergency consultation decreased by only 0.22 thousand points when TSP increased by 1 μg/m3. This indicates that an increase in TSP concentration was not a major factor affecting the reported points for emergency consultation. As the evidence remains uncertain, further research is necessary with regard to the impact of increased TSP concentrations on ECU.

We believe that continuous monitoring of AT, PSI, and TSP is needed to facilitate the development of short- and long-term prediction models as well as to provide corresponding emergency care measures for high-risk patients.

Limitations

This study was limited by data sources, and, therefore, the interactive relationships among specific disease types, air pollution quality, and ECU were not discussed. The study was restricted to discussing the relationship between the incidence of diseases requiring emergency care and air quality in relation to pollution. In addition, the research objective was to discuss the relationship between air quality in Taiwan and ECU, but the regional differences within Taiwan were not explored. Taiwan currently has 77 atmospheric quality monitoring stations, and, as such, the data obtained from these could provide insights into the relationship between air quality differences and ECU. Furthermore, this study did not calculate the effect of cumulative PSI and TSP concentrations on ECU and did not investigate seasonal variations, which limited its ability to provide references for predicting cumulative concentrations in relation to emergency medical care.

Conclusions

This study demonstrated the need to emphasize the impact of low temperatures and TSP values in relation to short-term dynamics. Although an increase in PSI values ultimately led to an increase in ECU, it did not affect ECU in relation to the short-term dynamics. For Taiwan, which is situated in a subtropical zone, we suggest the implementation of air quality monitoring and an emergency alarm system during periods of low temperatures, which have strong short-term implications. Long-term monitoring systems are necessary for PSI values and ECU to ensure emergency care readiness.

Disclosure statement

The authors report no conflicts of interest.

Acknowledgments

We thank Wen-Yi Chen, Ph.D., for assistance with analysis. This study was presented at the 2015 3rd International Conference on Global Economics and Governance Conference, Ming Chuan University, Taiwan, September 3–5, 2015.