The effects of past SARS experience and proximity on declines in numbers of travelers to the Republic of Korea during the 2015 MERS outbreak: A retrospective study

Abstract

Background

The experience of previous sizable outbreaks may affect travelers’ decisions to travel to an area with an ongoing outbreak.

Methods

We estimated changes in monthly numbers of visitors to the Republic of Korea (ROK) in 2015 compared to projected values by selected areas. We tested whether areas’ experience of a previous SARS outbreak of ≥100 cases or distance to the ROK had a significant effect on travel to the ROK during the MERS outbreak using t-tests and regression models.

Results

The percentage changes in visitors from areas with a previous SARS outbreak of ≥100 cases decreased more than the percentage changes in visitors from their counterparts in June (52.4% vs. 23.3%) and July (60.0% vs. 31.4%) during the 2015 MERS outbreak. The percentage changes in visitors from the close and intermediate categories decreased more than the far category. The results from regression models and sensitivity analyses demonstrated that areas with ≥100 SARS cases and closer proximity to the ROK had significantly larger percentage decreases in traveler volumes during the outbreak.

Conclusions

During the 2015 MERS outbreak, areas with a previous sizable SARS outbreak and areas near the ROK showed greater decreases in percentage changes in visitors to the ROK.

1. Background

The 2015 Middle East respiratory syndrome (MERS) outbreak in the Republic of Korea (ROK) was initiated by the arrival of a single infected international traveler who visited the Middle East [1]. The first case was confirmed on May 20, 2015, and the outbreak resulted in 186 cases and 38 deaths before the government of the ROK and the World Health Organization (WHO) officially announced the end of the outbreak on December 23, 2015 [2].

During the 2015 MERS outbreak, the ROK government conducted entry screening and active monitoring of all travelers from the Middle East region to avoid additional importations of MERS cases [2]. In addition, the ROK government restricted international departures by confirmed or suspected MERS patients or their contacts to prevent the global spread of MERS [2]. Although WHO did not recommend any travel restrictions or entry screening for MERS coronavirus (MERS-CoV) [3], the numbers of international travelers visiting the ROK decreased significantly during the 2015 MERS outbreak [4,5].

Fear-induced behavioral changes during novel infectious disease outbreaks have been reported to depend on an individual's perception of risk rather than the actual risk [6]. One factor that could affect an individual's perception of the risk of an infectious disease outbreak is the individual's previous experience of a similar infectious disease outbreak. Governments of countries that had a similar sizable outbreak, as well as individuals from those countries, may have responded differently to the ROK MERS outbreak than governments or individuals who had less exposure to a similar sizable outbreak.

This evaluation estimates the changes in numbers of non-citizen short-term visitor arrivals from selected areas to the ROK during the 2015 MERS outbreak and examines the correlation between travel volume declines and previous experience of the most similar sizable outbreak, the 2003 severe acute respiratory syndrome (SARS) outbreak. We also examined the correlation between decrease in travel volume and proximity to the ROK.

2. Methods

We used monthly international traveler statistics from TOURGO, the Korea tourism knowledge and information system, maintained by the Korea Culture and Tourism Institute [7]. The data include numbers of inbound non-citizen short-term visitor arrivals by area following the international guideline from the United Nations World Tourism Organization [7]. The numbers of arrivals reported in this analysis are limited to the numbers of non-citizen short-term visitor arrivals.

Each area includes one country or political unit. Note that area information in the data set is based on travelers’ passports rather than the areas from which travelers departed. First, we selected areas comprising more than 0.5% of total non-citizen visitor arrivals, i.e. more than 70 000 visitors to the ROK during 2014. Selected areas in descending order of numbers of arrivals to the ROK during 2014 are China (mainland China excluding Hong Kong Special Administrative Region (SAR), Macau, and Taiwan), Japan, United States (USA), Taiwan–China, Hong Kong SAR–China, Thailand, Philippines, Malaysia, Russia, Indonesia, Singapore, India, Canada, Vietnam, Australia, United Kingdom (UK), Germany, and France. They combined to contribute 94% of the total number of arrivals to the ROK during 2014. Monthly actual numbers of arrivals in 2015 by area from TOURGO are shown in Appendix Table 1.

For each area, we compared monthly numbers of arrivals in 2015 to the monthly average numbers of arrivals in 2013 and 2014. The baseline percentage change in numbers of arrivals (y) of each area to the ROK by month was calculated from the projected values (averages of 2013 and 2014) subtracted from the actual values (2015) and then divided by the projected values as shown in the following equation:

$$y_{ij}=\frac{Actual{\left(withtheMERSoutbreak\right)}_{ij}-Projected{\left(iftherewasnoMERSoutbreak\right)}_{ij}}{Projected{\left(iftherewasnoMERSoutbreak\right)}_{ij}}\times 100\%$$

where “i” stands for each area, while “j” stands for each month during the ROK MERS outbreak from June 2015 to December 2015. If the actual value is smaller than the projected value in a given area-month, i.e. decreasing arrivals associated with the MERS outbreak, the negative value $y_{ij}$ corresponds to the estimated percentage decrease in visitor arrivals. The number of travelers for each area in each month were treated as a single observation. We assumed that any changes of numbers of arrivals caused by the MERS outbreak would be observed starting in June because the first MERS case was confirmed on May 20, 2015.

We chose two main variables that could affect individuals’ decisions to travel to an outbreak area. The first variable was whether an area had a sizable outbreak of limited duration of a novel viral respiratory disease before the 2015 MERS outbreak. The 2003 SARS outbreak was chosen because it was sizable including local transmission and its duration was limited [8]. During the 2003 SARS outbreak, more than 8000 individuals were infected worldwide with a 9.6% average case fatality rate [9]. In addition, SARS and MERS are both viral respiratory infections caused by coronaviruses, although the case fatality rates reported for SARS were lower than for MERS [10]. Community transmission of SARS was more commonly reported, while MERS transmission in the ROK occurred primarily in healthcare settings or through close contact, e.g. caring for or living with an infected person [10,11]. The 2009 H1N1 influenza pandemic was also caused by viral respiratory infection; however, country-specific H1N1 attack rates were orders of magnitude higher and its case fatality rate was much lower compared to the limited-duration outbreaks of SARS and MERS [10,12,13].

Using WHO data for reported numbers of probable SARS cases by area during the 2003 outbreak [9], we hypothesized that individuals from areas with ≥100 SARS cases during the 2003 outbreak may have been significantly more or less likely to visit the ROK during the 2015 MERS outbreak. China, Taiwan–China, Singapore, Canada, and Hong Kong SAR–China each had ≥100 SARS cases during the 2003 outbreak, and all of those areas were included in the current analyses. Although the threshold was set at 100 probable SARS cases, each of the five areas experienced at least 238 probable SARS cases, while the country/area with next highest number of probable SARS cases, Vietnam, had 64 cases (Appendix Figure 3) [9]. Thus, our results would be consistent with any threshold set between 65 cases and 237 WHO-reported probable cases.

The second variable we considered was travel distance to the ROK from each area. Since travel time cost is likely to be correlated to travel distance between two areas, the opportunity cost to change travel plans from a more distant place to the ROK would be expected to be higher. Thus, individuals from more distant areas might be less likely to cancel trips to the ROK during the outbreak period. We subdivided selected areas into three groups based on distance to the ROK: (1) close (China, Japan, Taiwan–China, and Hong Kong SAR–China), (2) intermediate (Thailand, Philippines, Malaysia, Indonesia, Singapore, India, and Vietnam), and (3) far (USA, Russia, Canada, Australia, UK, Germany, and France). More details about how we defined distance categories for analyses are shown in Appendix Table 2, Appendix Table 3

For both variables, we performed Welch's t-tests for two samples with unequal variances [14] to examine whether the monthly average percentage decreases in numbers of arrivals from each category differed significantly. For the SARS variable, the null hypothesis is that the simple average percentage change of arrivals from areas with <100 SARS cases (reference) was the same as the percentage change of arrivals for areas with ≥100 SARS cases. The alternative hypothesis is that the percentage decreases for arrivals from areas with ≥100 SARS cases were larger than for areas with <100 SARS cases. For the distance categories, the null hypothesis is that the simple average percentage change of arrivals from the far category (reference) was the same as the percentage change of arrivals from the intermediate or the close category. The alternative hypothesis is that the percentage decreases for arrivals from areas in the far category were smaller than for arrivals from the areas in the intermediate or the close categories. When p-values from Welch's t-tests are <0.05, we reject the null hypotheses in favor of the alternative hypotheses.

We used ordinary least squares (OLS) regression models to examine the impact of each factor on percentage changes in numbers of visitor arrivals for each month during the ROK MERS outbreak (June–December 2015). The following equation was used in an OLS regression model.

$$y_{ij}=\alpha SAR{S}_i+{\beta }_{1}Clos{e}_i+{\beta }_{2}Intermediat{e}_i+{\beta }_0$$

“y _ij” is a dependent variable that shows percentage changes in numbers of arrivals by area-month. “SARS _i” is a dummy variable, which is one an area “i” reported ≥100 SARS cases during the 2003 outbreak, otherwise zero. “Close _i” is a dummy variable, which is one when an area “i” is in the “Close” category, otherwise zero. “Intermediate _i” is an additional distance dummy variable. When an area “i” is in the “Intermediate” category, the variable is one, otherwise zero. In summary, the OLS model included two independent variables: 1) a binary variable to identify areas with ≥100 SARS cases during the 2003 outbreak (reference category: <100 SARS cases during the 2003 outbreak), and 2) a categorical variable summarizing the distance between an area and the ROK that was transformed into two binary variables (reference category: “far”).

We conducted sensitivity analyses using alternative projected values. The baseline model used the average numbers of arrivals in 2013 and 2014 by area to project the expected numbers of arrivals to the ROK in the absence of the outbreak. The baseline model, however, may not account for year-to-year increasing or decreasing trends in the numbers of arrivals to the ROK. That omission could result in underestimated or overestimated projected values in the absence of the MERS outbreak, causing biased estimates from regression models. Thus, we examined two alternative models.

Alternative 1 used projected value based on the monthly average numbers of arrivals from each area in 2014 and 2016; i.e. the predicted value in June 2015 is the average number of arrivals in June 2014 and June 2016. This alternative assumes that neither 2014 nor 2016 travel volumes were affected by the MERS outbreak and that the expected monthly travel volumes in 2015 in the absence of the outbreak could have approximated the average volumes of 2014 and 2016.

For Alternative 2, we again started with the monthly average numbers of arrivals in 2013 and 2014. Since the ROK MERS outbreak was not expected to affect travel volumes during the pre-outbreak period in 2015, we compared the arrivals from January to May 2015 to the average number of arrivals from January to May in 2013 and 2014 to calculate an area-specific adjustment to the projected numbers of arrivals from June to December 2015 in the absence of the outbreak. Alternative 2 used the same method that was used to compare projected and actual numbers of arrivals in Mexico during the 2009 H1N1 pandemic [15].

To avoid bias, we conducted additional sensitivity analyses by excluding outliers. Potential outliers were flagged if any area-month (y _ij) observation was ≥100% or ≤ -100%. We only identified one observation that met these outlier criteria. We excluded the observation for India in August 2015, when an exceptional increase in the number of crewmembers arriving at seaports [16] resulted in an increase of 106.4% relative to the average of arrivals for India in August 2013 and 2014.

3. Results

Decreases in numbers of arrivals to the ROK were observed during June and July 2015 across all examined areas compared to the average numbers of arrivals in 2013 and 2014 (Table 1 ). In June, the number of arrivals decreased by 7.5% from the least affected area and 74.2% from the most affected area. The simple average decrease across 17 areas was 31.4%. In July, the simple average decrease in numbers of arrivals was 39.3% (range 4.8%–82.0% across areas). Although travel volumes declined across all areas during June and July, actual traveler volumes for some areas exceeded projected values starting in August and continuing through December 2015 (Appendix Figures 1 and 2).

Table 1.

Percentage changes between 2015 actual and projected (average of 2013 and 2014) monthly non-citizen arrivals to the Republic of Korea (ROK) by area (%).

Area	Jan	Feb	Mar	Apr	May	Jun	Jul	Aug	Sep	Oct	Nov	Dec
	Before the Middle East respiratory syndrome (MERS) outbreak				First MERS case reported May 20, 2015	Peak months of the MERS outbreak		Non-peak months of the MERS outbreak
China	60.0	79.4	47.1	48.0	55.3	−35.2	−59.5	−26.7	12.9	43.6	37.7	33.5
Japan	−26.2	−29.9	−18.3	−19.2	−11.2	−45.9	−57.0	−37.9	−29.0	−17.1	−18.1	−17.1
USA	2.4	4.1	4.1	18.3	6.5	−13.8	−4.8	3.0	−3.6	3.8	10.6	6.8
Taiwan-China	10.1	16.6	17.3	40.5	16.4	−74.2	−82.0	−54.0	−7.1	−9.5	3.4	6.3
Hong Kong SAR-China	19.4	40.6	47.8	56.5	53.1	−68.9	−81.1	−23.6	53.3	12.5	13.8	18.0
Thailand	19.9	3.9	8.2	0.2	−1.6	−32.0	−55.3	−43.1	−42.1	−21.7	−16.4	4.2
Philippines	−3.2	4.9	2.2	12.0	2.2	−26.9	−42.8	−20.1	7.5	13.8	16.2	13.1
Malaysia	11.1	6.0	20.9	23.2	3.4	−29.1	−42.8	−23.0	−8.7	−7.7	6.8	−3.3
Russia	−9.5	−11.1	−1.4	−2.1	6.4	−18.2	−21.9	−7.0	5.7	3.0	8.0	8.4
Indonesia	8.0	30.0	4.8	13.2	8.4	−26.7	−44.3	−23.6	6.2	15.2	13.0	−4.0
Singapore	11.7	1.9	13.6	15.5	11.1	−59.7	−55.8	−32.5	−19.2	−17.3	−7.9	−19.0
India	−6.1	14.7	−4.1	6.5	22.0	−22.2	−40.6	106.4	5.6	29.3	28.3	11.2
Canada	7.1	17.1	16.2	31.9	14.9	−23.8	−21.6	−5.9	−6.4	3.0	10.1	8.7
Vietnam	46.0	−11.0	67.2	56.5	28.4	−7.5	−26.3	6.2	21.2	49.7	65.6	35.7
Australia	10.9	2.9	−14.8	28.6	30.6	−28.3	−19.0	−5.7	−11.6	11.7	−1.9	9.0
UK	−4.4	1.7	6.6	31.9	2.5	−28.9	−26.9	−8.4	−12.4	4.5	7.4	1.9
Germany	−13.3	−12.6	10.8	6.5	1.1	−13.8	−15.3	0.3	−0.4	13.4	4.8	3.0
France	3.7	19.1	11.0	24.0	21.4	−9.7	−10.5	3.1	4.9	7.7	9.0	4.6
Average	8.2	9.9	13.3	21.8	15.1	−31.4	−39.3	−10.7	−1.3	7.7	10.6	6.7

Notes: Non-citizen arrivals are limited to non-citizen short-term visitor arrivals. All numbers are from the following equation.$\frac{Actual\left(withtheMERSoutbreak\right)-Projected\left(iftherewasnoMERSoutbreak\right)}{Projected\left(iftherewasnoMERSoutbreak\right)}\times 100\%$

Table 2 shows that the percentage decrease in numbers of arrivals from areas with ≥100 SARS cases during the 2003 SARS outbreak was significantly greater than the decrease from areas with <100 SARS cases in June (52.4% vs. 23.3%), July (60.0% vs. 31.4%), and August 2015 (28.5% vs. 3.8%). In comparison, during the pre-outbreak January to May period, the simple averages from both groups showed higher numbers of arrivals than the averages for 2013 and 2014 (Fig. 1 ). The numbers of arrivals from areas with a previous SARS outbreak with ≥100 cases were 21.7%–38.5% higher in January–May 2015 than in January–May averages for 2013 and 2014, while the simple averages from areas with <100 SARS cases were only 1.7%–15.4% higher in January–May 2015 (Table 2).

Table 2.

Average monthly percentage changes between 2015 actual and projected (average of 2013 and 2014) non-citizen arrivals to the Republic of Korea (ROK) by whether the country had 100 or more severe acute respiratory syndrome (SARS) cases during the 2003 outbreak.

Months	The 2015 Middle East respiratory syndrome (MERS) outbreak status	SARS ≥100 cases (n = 5)		SARS <100 cases (n = 13)		p-value
		%	SE	%	SE
Jan	Before the MERS outbreak	21.7	9.8	3.0	4.9	0.934
Feb		31.1	13.6	1.7	4.3	0.954
Mar		28.4	7.8	7.5	5.8	0.973
Apr		38.5	7.0	15.4	5.2	0.988
May	First MERS case reported May 20, 2015	30.2	9.9	9.2	3.5	0.952
Jun	Peak months of the MERS outbreak	−52.4	9.8	−23.3	2.9	0.018
Jul		−60.0	11.0	−31.4	4.7	0.026
Aug	Non-peak months of the MERS outbreak	−28.5	7.8	−3.8	10.2	0.035
Sep		6.7	12.7	−4.4	4.7	0.776
Oct		6.5	10.6	8.1	5.1	0.446
Nov		11.4	7.5	10.3	5.7	0.548
Dec		9.5	8.6	5.7	3.3	0.655

Notes: Non-citizen arrivals are limited to non-citizen short-term visitor arrivals. Monthly percentage changes were calculated by the following equation. $\frac{Actual\left(withtheMERSoutbreak\right)-Projected\left(iftherewasnoMERSoutbreak\right)}{Projected\left(iftherewasnoMERSoutbreak\right)}\times 100\%$

SE stands for standard error. Boldface indicates statistical significance (p < 0.05). Areas in the ≥100 SARS cases category are China, Taiwan–China, Hong Kong SAR–China, Singapore, and Canada.

Fig. 1.

Monthly percentage changes of non-citizen arrivals to the Republic of Korea (ROK) in 2015 compared with the average of monthly non-citizen arrivals in 2013 and 2014; areas are subdivided by whether each had 100 or more severe acute respiratory syndrome (SARS) cases during the 2003 SARS outbreak.

Notes: Non-citizen arrivals are limited to non-citizen short-term visitor arrivals. The solid line shows the average percentage changes in monthly numbers of arrivals among areas with ≥100 SARS cases, while the dotted line is the average percentage changes in numbers of arrivals among areas with <100 SARS cases. The shaded area indicates the period between the confirmation of the first Middle East respiratory syndrome (MERS) case in the ROK (May 20, 2015) and the declaration of the end of the outbreak by the government of the ROK and the World Health Organization (December 23, 2015). The shaded dark gray area indicates the peak period of the MERS outbreak, and the shaded light gray area indicates the non-peak period. Note that increases in arrivals (i.e. positive values) before the outbreak reflect the general trend of increasing numbers of arrivals to the ROK.

The percentage changes in arrivals from areas in the close category decreased more than the percentage changes in arrivals in the far category in June (56.1% vs. 19.5%), July (69.9% vs. 17.1%), and August (35.6% vs. 2.9%) (Fig. 2 and Table 3 ). In addition, the percentage changes in arrivals in the intermediate category decreased more than the percentage changes in arrivals in the far category in July (44.0% vs. 17.1%). Changes in all other months did not show significant differences across distance categories.

Fig. 2.

Monthly average percentage changes between actual and projected non-citizen arrivals to the Republic of Korea (ROK) during 2015 grouped by distance to the ROK.

Notes: Non-citizen arrivals are limited to non-citizen short-term visitor arrivals. The solid line with circles shows the average monthly percentage changes in numbers of arrivals from areas in the close category, while the dotted line with x is the average percentage changes in numbers of arrivals from areas in the intermediate category. The dotted line with triangles shows the average percentage changes in numbers of visitor arrivals from areas in the far category. The shaded area indicates the period between the confirmation of the first Middle East respiratory syndrome (MERS) case in the ROK (May 20, 2015) and the declaration of the end of the outbreak by the government of the ROK and the World Health Organization (December 23, 2015). The shaded dark gray area indicates the peak period of the MERS outbreak, and the shaded light gray area indicates the non-peak period. Note that increases in arrivals (i.e. positive values) before the outbreak reflect the general trend of increasing numbers of arrivals to the ROK.

Table 3.

Average monthly percentage changes between actual (2015) and baseline projected (average of 2013 and 2014) non-citizen arrivals to the Republic of Korea (ROK) by distance from the ROK.

Months	The 2015 Middle East respiratory syndrome (MERS) outbreak status	Far (n = 7)		Intermediate (n = 7)			Close (n = 4)
		%	SE	%	SE	p-value	%	SE	p-value
Jan	Before the MERS outbreak	−0.4	3.4	12.5	6.5	0.946	15.8	17.7	0.787
Feb		3.0	4.6	7.2	4.8	0.730	26.7	22.9	0.811
Mar		4.6	3.9	16.1	9.0	0.863	23.5	15.6	0.843
Apr		19.9	5.0	18.2	6.9	0.422	31.5	17.2	0.723
May	First MERS case reported May 20, 2015	11.9	4.1	10.6	4.1	0.410	28.4	15.9	0.811
Jun	Peak months of the MERS outbreak	−19.5	2.9	−29.2	5.9	0.087	−56.1	9.3	0.010
Jul		−17.1	2.8	−44.0	3.9	<0.001	−69.9	6.7	<0.001
Aug	Non-peak months of the MERS outbreak	−2.9	1.9	−4.2	19.3	0.474	−35.6	6.9	0.006
Sep		−3.4	2.8	−4.2	8.0	0.463	7.5	17.5	0.711
Oct		6.7	1.6	8.8	9.8	0.578	7.4	13.6	0.517
Nov		6.9	1.6	15.1	10.1	0.774	9.2	11.6	0.573
Dec		6.1	1.1	5.4	6.5	0.463	10.2	10.7	0.637

Notes: Non-citizen arrivals are limited to non-citizen short-term visitor arrivals. SE stands for standard error. Boldface indicates statistical significance (p-values < 0.05). The p-values are from Welch's t-tests for two samples with unequal variances. The reference group is far. The p-values in the intermediate column shows p-values to evaluate whether differences between the far and intermediate groups are significant, while p-values in the close column shows p-values to evaluate whether differences between the far and close groups are significant.

Close: China, Japan, Taiwan–China, and Hong Kong Special Administrative Region (SAR)–China; Intermediate: Thailand, Philippines, Malaysia, Indonesia, Singapore, India, and Vietnam; Far: USA, Russia, Canada, Australia, UK, Germany, and France.

The results from baseline OLS regression models also identified statistically significant correlations between the percentage decrease of numbers of arrivals and previous SARS outbreak of ≥100 cases (Table 4 ). Areas with a previous SARS outbreak of ≥100 cases showed significant correlations in June (p-value = 0.032). Also, the correlation between the percentage decrease of numbers of arrivals and distance from the ROK was statistically significant. The significant percentage decreases in numbers of visitor arrivals occurred in areas in the close category in June (p-value = 0.012) and July (p-value≤ 0.001), and in the intermediate category in July (p-value≤ 0.001).

Table 4.

Monthly marginal percentage change between actual (2015) and baseline projected (average of 2013 and 2014) non-citizen arrivals to the Republic of Korea (ROK) associated with areas that experienced ≥100 probable severe acute respiratory syndrome (SARS) cases and distance to the Republic of Korea.

Months	Independent variables		Coef.	p-value	R-squared
Jun	2003 SARS outbreak size (cases)	SARS ≥100	−18.3	0.032	0.68
	Distance	Close	−25.5	0.012
		Intermediate	−9.7	0.155
Jul	2003 SARS outbreak size (cases)	SARS ≥100	−11.8	0.058	0.87
	Distance	Close	−45.6	<0.001
		Intermediate	−26.8	<0.001
Aug	2003 SARS outbreak size (cases)	SARS ≥100	−11.7	0.595	0.41
	Distance	Close	−25.5	0.323
		Intermediate	−1.3	0.944
Sep	2003 SARS outbreak size (cases)	SARS ≥100	7.5	0.593	0.08
	Distance	Close	6.4	0.695
		Intermediate	−0.8	0.945
Oct	2003 SARS outbreak size (cases)	SARS ≥100	−2.2	0.876	<0.01
	Distance	Close	2.0	0.903
		Intermediate	2.0	0.861
Nov	2003 SARS outbreak size (cases)	SARS ≥100	3.1	0.820	0.04
	Distance	Close	0.5	0.975
		Intermediate	8.2	0.470
Dec	2003 SARS outbreak size (cases)	SARS ≥100	2.2	0.819	0.02
	Distance	Close	2.8	0.807
		Intermediate	−0.6	0.937

Notes: Non-citizen arrivals are limited to non-citizen short-term visitor arrivals. Boldface indicates statistical significance (p < 0.05). The results were from baseline ordinary least squares (OLS) regression models using percentage changes between 2015 actual and the average of 2013 and 2014 monthly arrivals as dependent variables. The reference category for the 2003 SARS outbreak areas with <100 probable SARS cases, and the reference for distance is the far category.

Results from sensitivity analyses were consistent with the baseline analysis. For Baseline and Alternatives 1, the percentage changes of arrivals from areas with a previous SARS outbreak of ≥100 cases showed greater decreases than the percentage changes in arrivals from their counterparts in June, July, and August 2015 (Appendix Table 4). For Alternative 2, the differences between areas with and without a SARS outbreak of ≥100 cases were significant in every month from June through December 2015, except for September.

For both Alternatives, declines in numbers of arrivals from areas in the close category were larger than for areas in the far category from June to August 2015 (Appendix Table 5). For Alternative 1, the declines in arrivals in the intermediate category were significantly larger than the declines in arrivals from the far category from April through December 2015, except for August. However, if we instead used Alternative 2, the decline in numbers of arrivals from the intermediate distance category was statistically significantly greater than the far category only in April, June, and July 2015.

The results from sensitivity analyses of regression models were also consistent with the results from baseline models. Using all three projected values, the percentage decrease of arrivals in June and July showed some significant correlations with proximity and previous SARS outbreaks of ≥100 cases (Appendix Table 6). When excluding outliers, the observation for India in August 2015, the difference in the percentage changes in arrivals between the intermediate and far categories was significant; however, when the India, August 2015 observation was included in the analysis (i.e. not excluded), the difference between the intermediate and far categories was no longer significant (Appendix Table 7).

4. Discussion

Results from the baseline and sensitivity analyses found that the 2015 MERS outbreak in the ROK appeared to have a greater effect on travel volumes from areas with ≥100 SARS cases in 2003, especially in June and July 2015. Most of the ROK MERS cases were confirmed during those 2 months, and the ROK government declared a de facto end to the outbreak on July 28, 2015 by which time no individuals were quarantined because of MERS, nor were there any newly confirmed cases [2]. The decrease of arrivals during the 2015 MERS outbreak is underscored when we compare the trends of numbers of arrivals before and during the 2015 MERS outbreak. Positive values in Table 1, Table 2, Table 3 (i.e. increases in arrivals) before the 2015 MERS outbreak reflect the general trend that the number of arrivals to the ROK would be increasing without the outbreak.

Visitors from areas with a previous SARS outbreak of ≥100 cases may have been more likely to change their travel plans because of similarities between SARS and MERS, which are both caused by coronaviruses and capable of causing serious respiratory impairment. In addition, the governments in areas that experienced previous SARS outbreaks may have had stronger restrictions or recommendations about travel to the ROK during the MERS outbreak period. Although WHO did not recommend any travel restrictions to the ROK during the 2015 MERS outbreak [17], WHO noted that raising awareness about MERS among those traveling to affected areas was good public health practice [18]. Some areas issued travel notices to inform travelers about the potential risks and risk mitigation strategies. Although the intent of notices was not to restrict travel, such notices may have raised additional concern among travelers, who may then have decided to cancel their travel plans.

For instance, Hong Kong SAR, which showed the biggest percentage difference between projected and actual arrivals to the ROK, had experienced a sizable SARS outbreak in 2003 with 1755 cases [9]. Hong Kong SAR issued a Red Outbound Travel Alert (OTA) for the ROK on June 9, 2015 [19]. Hong Kong's Red OTA recommends to adjust travel plans or avoid non-essential travel and is the second highest level travel notice [20]. The alert was lifted on August 1, 2015 [21]. In addition, the Travel Industry Council of Hong Kong canceled all tours, excluding cruises, to the ROK that were scheduled to depart from June 9 to the end of June in 2015 [22]. Since only about 2% of travelers from Hong Kong SAR to the ROK arrive via cruise [7], most tourists from Hong Kong SAR were affected by the cancellation. In addition, some local governments in mainland China (Sichuan and Shandong provinces, and Guangzhou, the capital of Guangdong province) issued travel notices to the ROK because of the 2015 MERS outbreak [23]. Also, some flights between China and Jeju international airport in the ROK were cancelled [24].

Taiwan also experienced a sizable SARS outbreak in 2003, and the Taiwan Centers for Disease Control (CDC) issued a level 2 travel notice (Alert) for Seoul and a level 1 travel notice (Watch) for all other areas in the ROK on June 2, 2015 [25]. The level 2 travel notice was extended to the entire ROK on June 9, 2015 [26]. The notice included a recommendation that travelers avoid unnecessary hospital visits in the ROK [26]. The Taiwan CDC lowered its travel notice to level 1 on July 7, 2015 [27], then removed the notice on July 28, 2015 [28]. The Taiwan Ministry of Foreign Affairs also issued a travel alert for MERS to the ROK [29].

The experience with SARS outbreaks in 2003 is only one potential reason that governments with previous SARS outbreak experiences may have issued stronger restrictions or recommendations about travel to the ROK during the MERS outbreak than governments without previous SARS outbreak experiences. For example, a Korean traveler who was experiencing symptoms consistent with MERS pushed ahead with his plan to travel to Guangdong, China with a layover in Hong Kong SAR [17]. When this traveler was diagnosed with MERS after arrival in China, a public health response was required in China [2]. Although there was no further MERS transmission in China or Hong Kong SAR, public awareness of this imported case may have exacerbated fear among individuals in Hong Kong SAR and Guangdong as well as their local governments. In Taiwan, a patient with suspected MERS was reported on May 30, 2015, although this patient was later confirmed to have influenza B virus infection [30]. This suspected case was one of 17 suspected MERS cases reported in Taiwan between September 2012 and May 2015 [30]. The 17th suspected case was in a person who was neither a ROK citizen nor had visited the ROK [30]. However, the patient could have increased public awareness of MERS in Taiwan. Although countries issued varied levels of travel notices, no countries enforced travel bans or strict quarantine requirements for travelers from the ROK as some did for travelers from countries with Ebola during the West African epidemic [31].

Since the 2015 MERS outbreak ended, the ROK government has continuously made efforts to prevent future MERS outbreaks in the ROK. The Korea Centers for Disease Control & Prevention updated its national prevention and control guidelines for MERS in April 2016 [32]. These guidelines include response plans at airports, hospitals, and local health departments to expedite the identification of individuals suspected of MERS-CoV infection after arriving from MERS-affected areas [32]. Following the guidelines, virological testing for MERS-CoV and other respiratory viruses has been implemented to confirm whether individuals with recent travel to MERS-affected areas and symptoms consistent with MERS are infected with MERS-CoV or not [32,33]. Also, the ROK's amended quarantine act specifies MERS as a quarantinable infectious disease [34]. These efforts may decrease the risk of future MERS outbreaks in the ROK.

This analysis has some limitations. First, we assumed that the nationality of an individual was a proxy for the area of departure for that individual. For instance, a Japanese national may live in the United States before visiting the ROK but would be counted as if traveling from Japan (close category) for our analysis. Such discrepancies between citizenship and residency might cause some bias in examining the effect of distance on travel volumes to the ROK during the MERS outbreak.

Next, we were limited in our ability to assess the mechanism by which a previous sizable outbreak experience would affect travel to the ROK. As we discussed, issuance of travel notices may correlate with past outbreak experience in that governments in countries or localities that previously experienced outbreaks may be more likely to issue stronger notices. In addition, stronger travel notices in some countries may have correlated with greater decreases in travel volumes during the MERS outbreak in the ROK. However, we were unable to create categorical variables for issuance of travel notices because each country has its own process and terminology for travel notices. Although many countries have similar structures for level of travel notices, they are not easily comparable. In addition to level, each travel notice contains outbreak-specific recommendations. Recommendations, such as avoiding unnecessary hospital visits while in the ROK, further complicate comparisons of the intensity of travel health notices among countries. Thus, we were unable to quantify differences between stronger and weaker travel notices across countries. Also, there could be other country- or area-specific factors, which could affect travel volumes during a disease outbreak, but were not considered in the current analyses.

Last, we chose only one previous outbreak, the 2003 SARS outbreak as an indicator for previous outbreak experiences. While SARS was also caused by a novel coronavirus resulting in respiratory disease similar to MERS, exposure to other sizable outbreaks, such as the 2009 H1N1 pandemic or the 2013–2015 Ebola epidemic, may also have affected governments' and individuals' perceptions of risk from the 2015 MERS outbreak. Although there was a large difference between the top five areas with at least 238 probable SARS cases for each and the next highest country with 63 probable SARS cases during the 2013 outbreak (Appendix Figure 3), a single threshold of ≥100 SARS cases were used. Also, a threshold of ≥100 SARS cases was used without considering other potentially critical factors, such as population size, the extent of geographic area, the time duration of the SARS outbreak, or extent of unlinked community transmission. For instance, both Canada and Singapore had ≥100 SARS cases in 2003. However, more than 95% of probable SARS cases in Canada (n = 247) were observed in the greater Toronto area [35], a small part of Canada, while the SARS outbreak in Singapore (n = 238) was over the entire country [9]. Thus, Canadian governments' and individuals’ perceptions of risk from the 2015 MERS outbreak associated with their SARS outbreak experience could be different from the perception of risk from the 2015 MERS outbreak in Singapore.

5. Conclusions

During the 2015 MERS outbreak, the numbers of arrivals to the ROK from around the world decreased. Countries with a previous sizable SARS outbreak and countries closer to the ROK had the greatest percentage decreases in travel volumes to the ROK. This is the first analysis that has examined area-specific impacts on travel volume to an outbreak area as we know. Future research using different outbreak situations may improve our understanding of the effects of outbreaks on travel volume.

Disclaimer

The findings and conclusions of this article are those of the authors and do not necessarily represent the official position of the United States Centers for Disease Control and Prevention or the Korea Centers for Disease Control & Prevention.

Appendix B.

Appendix Fig. 1.

Percentage changes between 2015 actual and baseline projected monthly non-citizen arrivals to the Republic of Korea by area (%).

Notes: Non-citizen arrivals refers to non-citizen short-term visitor arrivals. Projected values are estimated from the average of 2013 and 2014 monthly average numbers of non-citizen arrivals. Percentage changes in numbers of non-citizen arrivals are calculated by the following equation.$\frac{Actual\left(withtheMERSoutbreak\right)-Projected\left(iftherewasnoMERSoutbreak\right)}{Projected\left(iftherewasnoMERSoutbreak\right)}\times 100\%$

Appendix Fig. 2.

Maps of percentage changes between 2015 actual and 2015 baseline projected monthly non-citizen arrivals to the Republic of Korea by area (%).

Notes: Non-citizen arrivals are limited to non-citizen short-term visitor arrivals. Projected values are from the average of 2013 and 2014 monthly averages of non-citizen arrivals. Percentage changes in numbers of non-citizen arrivals are calculated using the following equation: $\frac{Actual\left(withtheMERSoutbreak\right)-Projected\left(iftherewasnoMERSoutbreak\right)}{Projected\left(iftherewasnoMERSoutbreak\right)}\times 100\%$Areas colored green represent areas with positive percentage changes from the index. In other words, the actual non-citizen arrivals in 2015 are bigger than the projected values using the monthly average of 2013 and 2014 for the green areas. The red-colored areas indicate that the actual non-citizen arrivals values in 2015 are smaller than the projected values.

Appendix Fig. 3.

Numbers of probable severe acute respiratory syndrome (SARS) cases by area (November 1, 2002–July 31, 2003).

Notes: Numbers of probable SARS cases by area are from the World Health Organization (https://www.who.int/csr/sars/country/table2004_04_21/en/).

Appendix Table 1.

Actual numbers of monthly non-citizen arrivals in the Republic of Korea in 2015 by area

Month	China	Japan	USA	Taiwan–China	Hong Kong SAR–China	Thailand	Philippines	Malaysia	Russia	Indonesia
Jan	394 345	139 632	50 896	49 798	35 407	40 880	24 716	12 396	13 957	12 120
Feb	516 787	142 587	48 728	53 719	50 992	34 861	23 740	14 213	11 734	12 092
Mar	515 130	218 932	66 755	52 304	51 876	46 868	33 763	21 493	15 915	15 667
Apr	641 610	154 402	80 120	60 752	67 465	43 851	45 128	28 330	17 831	21 092
May	618 083	188 420	70 891	59 061	54 095	35 020	41 141	19 463	18 663	20 136
Jun	315 095	101 206	62 550	14 816	13 949	16 749	31 427	10 283	14 368	14 881
Jul	255 632	81 748	62 627	9879	7761	12 130	23 529	5893	13 524	10 418
Aug	513 275	148 632	63 522	25 622	33 752	13 634	32 801	9709	15 936	14 555
Sep	591 242	157 313	63 793	46 353	52 400	15 907	36 605	17 570	15 904	15 393
Oct	650 174	180 348	77 878	48 929	53 648	36 102	43 018	23 767	17 638	20 166
Nov	507 579	164 685	62 273	47 463	47 329	31 627	34 745	28 090	16 472	16 513
Dec	465 218	159 877	57 580	49 494	54 753	44 140	33 009	32 143	16 164	20 557

Month	Singapore	India	Canada	Vietnam	Australia	UK	Germany	France
Jan	8934	7228	9898	7895	9997	9715	6316	5282
Feb	8166	7983	10 537	9809	7010	8051	6046	5978
Mar	15 541	10 206	14 085	18 309	11 005	12 866	10 123	6898
Apr	18 331	11 977	16 447	18 937	15 526	14 556	9371	8816
May	16 747	14 160	14 541	14 485	14 409	10 512	8770	8171
Jun	7926	11 903	8909	10 840	8504	7632	6922	5744
Jul	4439	8551	9496	9172	9017	7921	7008	6467
Aug	6972	31 461	11 881	12 453	8458	10 223	8787	8678
Sep	11 867	14 150	11 870	14 052	12 273	10 163	9525	6889
Oct	15 672	15 476	14 505	18 740	14 002	13 345	12 248	8620
Nov	18 224	11 644	12 345	16 257	9759	10 346	8953	6784
Dec	27 334	8863	11 033	11 816	13 306	7944	6113	5505

Note: Non-citizen arrivals refers to non-citizen short-term visitor arrivals.

Data source: TOURGO (Tourism Knowledge and Information System). Tourism arrival statistics. [cited 2017 Aug. 24]; Available from: http://know.tour.go.kr/stat/tourStatSearchDis.do;jsessionid=8B29C231492BA57E875081E0A6D4C762.

Appendix Table 2.

Categorization of areas based on distance and flight time from capital cities to Seoul (Republic of Korea), and flight frequency

Category	Country or political unit	Capital	Distance (km)	Flight time to Seoul	Numbers of direct flight per day
Close	China	Beijing	952	2hr 5min	12
	Japan	Tokyo	1152	2hr 30min	31
	Taiwan	Taipei	1484	2hr 30min	15
	Hong Kong SAR	Hong Kong	2087	3hr 40min	23
Intermediate	Philippines	Manila	2620	4hr 5min	9
	Vietnam	Hanoi	2740	4hr 15min	12
	Thailand	Bangkok	3720	5hr 30min	15
	Malaysia	Kuala Lumpur	4611	6hr 30min	4
	Singapore	Singapore	4669	6hr 30min	7
	India	New Delhi	4684	6hr 50min	1
	Indonesia	Jakarta	5289	7hr 10min	3
Far	Russia	Moscow	6606	8hr 30min	1
	Germany	Berlin	8122	12hr 5min	0
	UK	London	8852	11hr	3
	France	Paris	8960	11hr	2
	Australia	Canberra	8411	12hr 45min	0
	Canada	Ottawa	10 505	16hr 30min	0
	USA	Washington DC	11 157	14hr 30min	1

Notes: Distance, flight time to Seoul, and numbers of direct flights per day between Seoul and each capital city came from Google search in April 2018.

Appendix Table 3.

Distance category by distance and flight time to Seoul (Republic of Korea)

	Distance to Seoul (km)	Distance to Seoul (miles)	Flight time to Seoul
Close	0 to 2500	0 to 1500	Less than 4 h
Intermediate	2501 to 6000	1501 to 4000	4–8 h
Far	6001 or more	4001 or more	More than 8 h

Appendix Table 4.

Monthly percentage changes between actual and projected non-citizen arrivals to the Republic of Korea (ROK) during 2015 by whether an area had 100 or more severe acute respiratory syndrome (SARS) cases during the 2003 outbreak—sensitivity analyses using different projected values for the number of expected travelers in the absence of a Middle East respiratory syndrome (MERS) outbreak

Months	Baseline (2013 and 2014)					Alternative 1 (2014 and 2016)					Alternative 2 (2013 and 2014 adjustment)
	SARS ≥100 cases (n = 5)		SARS <100 cases (n = 13)		p-value	SARS ≥100 cases (n = 5)		SARS <100 cases (n = 13)		p-value	SARS ≥100 cases (n = 5)		SARS <100 cases (n = 13)		p-value
	%	SE	%	SE		%	SE	%	SE		%	SE	%	SE
Jan	21.7	9.8	3.0	4.9	0.934	−0.04	1.5	−5.7	2.7	0.957	−6.4	3.4	−4.4	2.0	0.314
Feb	31.1	13.6	1.7	4.3	0.954	9.7	4.1	−9.9	2.9	0.999	0.1	3.6	−4.9	3.3	0.833
Mar	28.4	7.8	7.5	5.8	0.973	1.4	2.0	−3.8	2.9	0.921	−1.0	1.8	−0.1	3.0	0.404
Apr	38.5	7.0	15.4	5.2	0.988	8.0	5.3	−1.1	3.3	0.912	7.3	4.0	7.1	2.1	0.510
May	30.2	9.9	9.2	3.5	0.952	−0.1	2.8	−7.9	2.9	0.963	0.02	1.8	2.2	2.4	0.242
Jun	−52.4	9.8	−23.3	2.9	0.018	−60.7	8.8	−32.1	2.6	0.013	−62.9	7.9	−28.3	2.3	0.004
Jul	−60.0	11.0	−31.4	4.7	0.026	−66.7	9.7	−41.0	4.5	0.024	−67.9	9.9	−35.8	4.3	0.012
Aug	−28.5	7.8	−3.8	10.2	0.035	−37.5	7.6	−12.9	7.9	0.020	−44.2	7.2	−10.1	9.4	0.005
Sep	6.7	12.7	−4.4	4.7	0.776	−11.5	10.1	−18.6	3.1	0.737	−18.3	6.5	−10.6	3.9	0.167
Oct	6.5	10.6	8.1	5.1	0.446	−14.1	6.0	−9.5	3.0	0.260	−18.6	3.3	0.9	3.6	0.001
Nov	11.4	7.5	10.3	5.7	0.548	−10.0	4.2	−9.7	2.4	0.477	−14.1	2.4	2.6	3.4	<0.001
Dec	9.5	11.1	5.7	3.3	0.655	−11.7	4.1	−10.1	2.2	0.368	−15.8	2.3	−1.1	2.3	0.002

Notes: Non-citizen arrivals are limited to non-citizen short-term visitor arrivals. SE stands for standard error. Boldface indicates statistical significance (p-value <0.05). Baseline columns showed percentage changes of monthly non-citizen arrivals by area for actual 2015 actual compared to the averages of 2013 and 2014 non-citizen arrivals to project the number of monthly non-citizen arrivals in the absence of a MERS outbreak. Alternative 1 columns used the averages of 2014 and 2016 monthly arrivals to project values. Alternative 2 columns project the expected number of non-citizen arrivals using the average monthly change in visitor arrivals from each area occurring between January and May 2015 (pre-outbreak) compared to the monthly averages from January through May 2013 and 2014. Area-specific projections were made in each of the models.

Appendix Table 5.

Monthly percentage changes of non-citizen arrivals to the Republic of Korea

(ROK) during 2015 based on actual versus projected values by distance to ROK—sensitivity analyses using different projected values to model the number of expected travelers in the absence of a Middle East respiratory syndrome outbreak

	Alternative 1 (2014 and 2016)								Alternative 2 (2013 and 2014 adjustment)
	Far (n = 7)		Intermediate (n = 7)			Close (n = 4)			Far (n = 7)		Intermediate (n = 7)			Close (n = 4)
	%	SE	%	SE	p-value	%	SE	p-value	%	SE	%	SE	p-value	%	SE	p-value
Jan	−4.3	2.5	−4.4	4.8	0.488	−3.4	2.6	0.595	−7.7	1.6	−0.6	3.1	0.963	−7.7	3.7	0.501
Feb	−3.2	3.2	−12.0	4.6	0.072	6.6	8.1	0.842	−4.7	1.8	−3.9	6.0	0.546	−0.7	5.2	0.747
Mar	−1.9	2.4	−4.0	5.2	0.358	−0.3	2.7	0.662	−2.6	4.1	2.1	3.9	0.789	−0.8	2.4	0.644
Apr	7.2	4.9	−6.0	1.8	0.018	4.5	7.7	0.389	11.0	2.5	4.3	2.5	0.039	5.5	4.9	0.179
May	−0.1	3.0	−15.0	2.2	0.001	0.9	3.1	0.593	3.9	2.9	−1.9	3.0	0.092	3.6	3.6	0.471
Jun	−25.7	2.8	−41.4	3.9	0.003	−62.9	9.1	0.009	−24.9	3.8	−37.3	4.6	0.028	−61.9	11.1	0.016
Jul	−26.6	3.3	−55.2	1.7	<0.001	−73.6	7.6	0.001	−22.8	3.3	−50.5	2.4	<0.001	−73.0	9.5	0.003
Aug	−11.0	1.9	−15.6	15.2	0.385	−42.3	7.1	0.008	−9.6	3.0	−14.5	18.3	0.399	−45.9	8.7	0.008
Sep	−12.9	2.7	−25.5	4.2	0.014	−7.8	12.1	0.649	−9.9	4.4	−15.2	6.4	0.249	−13.6	7.8	0.342
Oct	−4.0	3.0	−17.6	4.0	0.009	−10.6	6.5	0.197	−0.5	3.4	−4.0	7.2	0.336	−12.7	6.7	0.081
Nov	−6.2	2.9	−14.0	3.0	0.039	−8.6	4.8	0.338	−0.4	3.1	1.4	6.4	0.598	−11.0	5.2	0.064
Dec	−3.7	1.8	−18.2	1.9	<0.001	−9.1	2.5	0.063	−1.1	2.9	−6.5	4.8	0.177	−10.0	5.1	0.091

Notes: Non-citizen arrivals are limited to non-citizen short-term visitor arrivals. SE stands for standard error. Boldface indicates statistical significance (p-value < 0.05).

Alternative 1 columns used the average of 2014 and 2016 monthly non-citizen arrivals as projected numbers of non-citizen arrivals in the absence of the MERS outbreak. Alternative 2 columns project the expected number of non-citizen arrivals using the average monthly change in visitor arrivals from each area occurring between January and May 2015 (pre-outbreak) compared to the monthly averages from January through May 2013 and 2014. Area-specific projections were made in each of the models. p-values in the intermediate column shows p-values to evaluate significant differences between the far and intermediate groups, while p-values in the close column shows p-values to evaluate significant differences between the far and close groups.

Appendix Table 6.

Monthly marginal percentage change of non-citizen arrivals to the Republic of Korea (ROK) during 2015 from areas associated with ≥100 probable severe acute respiratory syndrome (SARS) cases and by distance to the ROK based on projected values from regression models—sensitivity analyses using different models to project values.

Months	Independent variables	Baseline (2013 and 2014)			Alternative 1 (2014 and 2016)			Alternative 2 (2013 and 2014 adjustment)
		Coef.	p-value	R-squared	Coef.	p-value	R-squared	Coef.	p-value	R-squared
Jun	SARS ≥100	−18.3	0.032	0.68	−19.3	0.003	0.81	−27.1	<0.001	0.82
	Close	−25.5	0.012		−25.5	<0.001		−20.5	0.008
	Intermediate	−9.7	0.155		−15.7	0.004		−12.4	0.022
Jul	SARS ≥100	−11.8	0.058	0.87	−12.5	0.027	0.88	−19.2	0.002	0.90
	Close	−45.6	<0.001		−39.3	<0.001		−38.5	<0.001
	Intermediate	−26.8	<0.001		−28.5	<0.001		−27.7	<0.001
Aug	SARS ≥100	−11.7	0.595	0.41	−13.7	0.433	0.24	−24.0	0.252	0.27
	Close	−25.5	0.323		−23.0	0.261		−21.7	0.366
	Intermediate	−1.3	0.944		−4.7	0.749		−4.9	0.774
Sep	SARS ≥100	7.5	0.593	0.08	1.8	0.847	0.23	−10.4	0.282	0.11
	Close	6.4	0.695		4.0	0.704		2.5	0.820
	Intermediate	−0.8	0.945		−12.6	0.120		−5.4	0.500
Oct	SARS ≥100	−2.2	0.876	<0.01	−6.8	0.311	0.35	−20.5	0.022	0.40
	Close	2.0	0.903		−2.5	0.743		−12.2	0.205
	Intermediate	2.0	0.861		−13.7	0.025		−3.5	0.662
Nov	SARS ≥100	3.1	0.820	0.04	−1.5	0.777	0.39	−15.7	0.051	0.36
	Close	0.5	0.975		−1.5	0.810		−1.0	0.908
	Intermediate	8.2	0.470		−7.9	0.101		1.8	0.773
Dec	SARS ≥100	2.2	0.819	0.02	−3.9	0.233	0.70	−16.8	0.006	0.50
	Close	2.8	0.807		−3.0	0.427		1.4	0.820
	Intermediate	−0.6	0.937		−14.5	<0.001		−5.4	0.235

Notes: Non-citizen arrivals are limited to non-citizen short-term visitor arrivals. Boldface indicates statistical significance (p-value < 0.05). Baseline columns showed percentage changes of monthly non-citizen arrivals by area for actual 2015 actual compared to the averages of 2013 and 2014 monthly non-citizen arrivals to project the number of arrivals in the absence of a MERS outbreak. Alternative 1 columns used the average of 2014 and 2016 monthly non-citizen arrivals to project values. Alternative 2 columns project the expected number of non-citizen arrivals using the average monthly change in non-citizen arrivals from each area occurring between January and May 2015 (pre-outbreak) compared to the monthly averages from January through May 2013 and 2014. Area-specific projections were made in each of the models.

Appendix Table 7.

Sensitivity analyses with and without the area-month observation for India, August 2015

Percentage changes in non-citizen arrivals to the Republic of Korea (ROK) in August 2015 between actual and projected values by whether an area had ≥100 probable severe acute respiratory syndrome (SARS) cases during the 2003 outbreak
	Baseline (2013 and 2014)					Alternative 1 (2014 and 2016)					Alternative 2 (2013 and 2014 adjustment)
	SARS ≥100 cases		SARS <100 cases		p-value	SARS ≥100 cases		SARS <100 cases		p-value	SARS ≥100 cases		SARS <100 cases		p-value
	%	SE	%	SE		%	SE	%	SE		%	SE	%	SE
Without India	−28.5	7.8	−13.0	4.8	0.062	−37.5	7.6	−20.2	3.6	0.040	−44.2	7.2	−18.7	4.0	0.007
With India	−28.5	7.8	−3.8	10.2	0.035	−37.5	7.6	−12.9	7.9	0.020	−44.2	7.2	−10.1	9.4	0.005

	Baseline (2013 and 2014)					Alternative 1 (2014 and 2016)					Alternative 2 (2013 and 2014 adjustment)
	Far		Intermediate		p-value	Far		Intermediate		p-value	Far		Intermediate		p-value
	%	SE	%	SE		%	SE	%	SE		%	SE	%	SE
Without India	−2.9	1.9	−22.7	6.7	0.015	−11.0	1.9	−30.5	8.6	<0.001	−9.6	3.0	−32.5	3.8	<0.001
With India	−2.9	1.9	−4.2	19.3	0.474	−11.0	1.9	−15.6	15.2	0.385	−9.6	3.0	−14.5	18.3	0.399

	Independent variables	Baseline (2013 and 2014)			Alternative 1 (2014 and 2016)			Alternative 2 (2013 and 2014 adjustment)
		Coef.	p-value	R-squared	Coef.	p-value	R-squared	Coef.	p-value	R-squared
Without India	SARS ≥100	−4.3	0.602	0.60	−7.7	0.186	0.75	−16.9	0.010	0.81
	Close	−30.0	0.006		−26.6	0.001		−26.0	0.002
	Intermediate	−19.6	0.014		−19.4	0.001		−22.6	0.001
With India	SARS ≥100	−11.7	0.595	0.41	−13.7	0.433	0.24	−24.0	0.252	0.27
	Close	−25.5	0.323		−23.0	0.261		−21.7	0.366
	Intermediate	−1.3	0.944		−4.7	0.749		−4.9	0.774

Notes: Non-citizen arrivals are limited to non-citizen short-term visitor arrivals. Boldface indicates statistical significance (p-value < 0.05).

Appendix A. Supplementary data

The following is the supplementary data to this article: