key: cord-0732452-0isi7yd4
authors: Miyawaki, A.; Tabuchi, T.; Tomata, Y.; Tsugawa, Y.
title: Association between Participation in Government Subsidy Program for Domestic Travel and Symptoms Indicative of COVID-19 Infection
date: 2020-12-04
journal: medRxiv : the preprint server for health sciences
DOI: 10.1101/2020.12.03.20243352
sha: cc7f63cc519afec9aef446adfe18ba4272755127
doc_id: 732452
cord_uid: 0isi7yd4

Importance: As countermeasures against the economic downturn caused by the coronavirus 2019 (COVID-19) pandemic, many countries have introduced or considering financial incentives for people to engage in economic activities such as travel and use restaurants. Japan has implemented a large-scale, nationwide government-funded program that subsidizes up to 50% of all travel expenses since July 2020 with the aim of reviving the travel industry. However, it remains unknown as to how such provision of government subsidies for travel impacted the COVID-19 pandemic. Objective: To investigate the association between participation in government subsidies for domestic travel in Japan and the incidence of COVID-19 infections. Design, Setting, and Participants: Using the data from a large internet survey conducted between August 25 and September 30, 2020, in Japan, we examined whether individuals who used subsidies experienced a higher likelihood of symptoms indicative of the COVID-19 infection. Exposure: Participation in the government subsidy program for domestic travel. Main Outcomes and Measures: Five symptoms indicative of the COVID-19 infection (high fever, throat pain, cough, headache, and smell and taste disorder) within the past one month of the survey. Results: Of the 25,482 respondents (50.3% [12,809] women; mean [SD] age, 48.4 [17.4] years), 3,289 (12.9%) participated in the subsidy program at the time of survey. After adjusting for potential confounders, we found that the participants of the subsidy program exhibited higher incidence of high fever (adjusted rate, 4.8% for participants vs. 3.7% for non-participants; adjusted odds ratio [aOR], 1.90; 95%CI, 1.40-2.56; p<0.001), throat pain (20.0% vs. 11.3%; aOR, 2.13; 95%CI, 1.39-3.26; p=0.002), cough (19.2% vs. 11.2%; aOR 1.97; 95%CI, 1.28-3.03; p=0.004), headache (29.4% vs. 25.5%; aOR, 1.26; 95%CI, 1.09=1.46; p=0.005), and smell and taste disorder (2.6% vs. 1.7%; aOR 1.66; 95%CI; 1.16-3.49; p=0.01) compared with the non-participants. Conclusion and Relevance: The participants of government subsidies for domestic travel experienced a higher incidence of symptoms indicative of the COVID-19 infection.

As of the end of November 2020, 62 million people have been infected by the coronavirus disease 2019 (COVID- 19) , and 1.4 million have died of this infection. 1 In order to tackle this unprecedented pandemic, many countries have implemented public health measures -also known as non-pharmaceutical interventions (NPIs) -to control the spread of the virus, including lockdown, movement restrictions, quarantine, and border control. 2 Given that the number of infections and deaths due to COVID-19 has been resurging during this winter, these NPIs are likely to be implemented intermittently, 3 until effective vaccines would be developed and become widely available. While these NPIs have been shown to be effective in reducing the spread of COVID-19 infections, 2,4 they have a substantial negative impact on the economy. 5 To countermeasure the economic downturn due to the NPIs, many countries have introduced or actively considering financial incentives such as government subsidies to engage in economic activities such as to use restaurants or travel domestically. [6] [7] [8] [9] Evidence is limited as to whether the government interventions to financially incentivize economic activities such as using restaurants or traveling impact the COVID-19 infection. For example, the United Kingdom implemented the "Eat out to Help out" campaign, in which the government subsidized up to 50% of the expenses of food and non-alcoholic drinks for immediate consumption at restaurants using a budget of around £500 million throughout August. 9 A recent study using ecological data on COVID-19 infections by region suggested that regions that implemented this campaign experienced 8-17 percentage points higher number of COVID-19 clusters. 10 However, ecological association does not imply that the same association could be observed at the individual level (known as "ecological fallacy"), and therefore, it remains unknown as to whether this policy actually led to an increased number of individuals infected by the COVID-19. Indeed, to our knowledge, no study to date has evaluated the impact of such economic policy on the risk of the COVID-19 infection using individual-level data.

Moreover, it remains unknown as to how similar policies implemented in other countries that incentivize economic activities (e.g., eating out, travel) affected the COVID-19 pandemic.

Japan has implemented large-scale, nationwide government subsidies for domestic travel (called the "Go-To" Domestic Travel Campaign) 8 since July 2020 with the aim of reviving the travel industry, which has been hit hard by a substantial decrease in the number of foreign tourists visiting Japan. This program incentivizes people to travel domestically by subsidizing up to 50% of all travel expenses for those who travel, including transport and accommodation expenses. As of the end of October 2020, more than 200 billion Japanese yen (JPY) (approximately 2 billion US dollars, using an exchange rate of 100 JPY per US dollar) have been used to subsidize a total of 40 million people who traveled domestically. 11 However, as the number of the COVID-19 infected cases has resurged, the Japanese government is facing fierce criticisms speculating that increased mobility and human interactions due to this program may be causing the increase in the number of COVID-19 infections. 8 Yet, empirical evidence is lacking as to whether the introduction of this program is associated with an increased risk of the COVID-19 infection.

Japan's experience from this social experiment provides a unique opportunity to understand the impact of government subsidies for travel on the spread of COVID-19 infections.

In this context, using the data from a large internet survey conducted between August 25 and September 30, 2020, in Japan, we examined whether individuals who used subsidies experienced a higher incidence of symptoms indicative of the COVID-19 infection (COVID-19-like symptoms).

We analyzed the data from the Japan "COVID-19 and Society" Internet Survey (JACSIS) study, a cross-sectional, web-based, self-reported questionnaire survey administered by a large internet research agency (Rakuten Insight, Inc., which had approximately 2.2 million qualified panelists in 2019). 12 This internet research agency has been used in previous studies. 13, 14 The questionnaires were distributed to 224,389 panelists selected by each sex, age, and prefecture category using simple random sampling (it covered all 47 prefectures, the first-tier administrative district in Japan). The panelists who consented to participate in the survey accessed the designated website and responded to the questionnaires, and had the option not to respond or discontinue at any point of the survey. The questionnaires were distributed starting on August 25, 2020, and was completed on September 30, 2020, when the target numbers of respondents for each sex, age, and prefecture category were met (the target numbers for each sex, age, and prefecture category had been determined in advance according to the population distribution in 2019; 28000 respondents; response rate, 12.5%). We excluded 2,518 individuals showing unnatural or inconsistent responses using the algorithm we developed. The final sample size was 25,482 respondents (91.0% of the total survey respondents).

All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted December 4, 2020. ; https://doi.org/10.1101/2020.12.03.20243352 doi: medRxiv preprint

The primary exposure variable was the participation in the subsidy program for domestic travel, which has been effective since July 22, 2020.

Our outcome variable was the incidence of five self-reported COVID-19-like symptoms (high fever, throat pain, cough, headache, and smell and taste disorder) during the one month prior to the time of the survey. 15 Self-reported COVID-19-like symptoms have been reported as a useful measure to monitor the spread of the COVID-19 infections. 16, 17 

We adjusted for the respondents' demographics, 18 socio-economic status (SES), 19 health-related characteristics, 18 and prefecture fixed effects (effectively comparing individuals living in the same prefecture). The demographics included age (categorized as 15-19, 20-29, …, 70-79) and sex. The SES included academic attainment (graduated from college or higher institutions vs.

high school or lower institutions), income level (categorized using the tertiles of household equivalent income ["low" = less than 2.5 million JPY, "medium" = 2.5 to 4.3 million JPY, and "high" = more than 4.3 million JPY], and an indicator for those who refused to respond to this question), household size (number of household members: 1, 2, 3, 4 and 5+), employment status (employer, self-employed, regular employee, non-regular employee, and unemployed), and marital status (married, never married, widowed, and separated). The household equivalized income was calculated as the gross (pre-tax) income in 2019, divided by the square root of the number of household members. Health-related characteristics included smoking status (never, ever, and current smokers), walking disability (whether the person is experiencing difficulties in All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. Body mass index was calculated by dividing self-reported body weight by self-reported body height squared (m 2 ).

First, we compared the demographics, SES, and health-related characteristics of the participants in the subsidy program for domestic travel vs. non-participants. To account for the possibility that those who participated and responded to the internet-based survey may differ from the general population (e.g., a younger population may be more likely to participate and respond to the internet-based survey), we used the weighted regression models (inverse probability weighting [IPW]). 20 The weights (propensity scores) were calculated by fitting a logistic regression model using demographics, SES, and health-related characteristics to adjust for the difference in respondents between the current internet survey and a widely-used nationwide representative survey (i.e., the 2016 Comprehensive Survey of Living Conditions 21, 22 ) (see

Second, we examined the association between participation in the subsidy program for domestic travel and the incidence rates of the COVID-19-like symptoms. For each outcome, we adjusted for the respondents' sociodemographic characteristics, health-related characteristics, and prefecture fixed effects. We used weighted multivariable logistic regression models, with standard errors clustered at the prefecture-level, to account for the potential correlation of the respondents within the same prefecture. To calculate risk-adjusted incidence rates of COVID-19-All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted December 4, 2020. ; https://doi.org/10.1101/2020.12.03.20243352 doi: medRxiv preprint like symptoms, we used marginal standardization (also known as predictive margins or margins of response). 23 For each respondent, we calculated predicted probabilities of the incidence of the COVID-19-like symptom with participation in the subsidy program fixed at each category and then averaged over the distribution of covariates in our sample.

To adjust for multiple comparisons of having five outcome variables using the Holm method, 24 which sequentially compares the i-th smallest P value (for i = 1, …, 5) among the five original P values with progressively less restrictive alpha levels (= .05/(5 − i + 1)). To make the interpretation easier, we calculated the adjusted P value by multiplying the unadjusted P values by (5 -i + 1) times, and considered the adjusted P value < 0.05 to be statistically significant. 25 All analyses were conducted using Stata version 15 (College Station, TX; StataCorp LLC.). This study was approved by the Institutional Review Board of the Osaka International Cancer

We conducted sensitivity analyses. First, the travels to and from Tokyo were ineligible for the subsidy program until September 15, due to a large number of the COVID-19 cases in Tokyo. 8 To assess whether our findings were sensitive to the inclusion of residents of Tokyo (we did not exclude these individuals in our main analyses as they could still use the subsidy program if their companion lived in other prefectures than Tokyo), we reanalyzed the data after excluding the respondents living in Tokyo prefecture. Second, we repeated the analyses without using IPW to examine how the use of this approach affected our findings. Third, in order to test whether the impact of the subsidy program differs by the characteristics of respondents, we conducted All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted December 4, 2020. ; https://doi.org/10.1101/2020.12.03.20243352 doi: medRxiv preprint stratified analyses by age (15-64 years and 65-79 years), the presence of comorbidities (no comorbidities vs. having at least one comorbidity), and sex.

Of the 25,482 respondents, 3,289 (12.9%) had participated in the subsidy program for domestic travel at the time of the survey. The participants in the subsidy program were younger; have higher education and higher income; and more likely to be employed and overweight ( Table 1) .

After adjusting for demographics, SES, health-related characteristics and indicators of prefectures ( 

Our findings were largely unaffected by excluding the respondents living in Tokyo (Supplementary Table 1 ) or using unweighted regression models (Supplementary Table 2) .

The result of the stratified analyses by age showed a higher incidence rate of COVID-19-like All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted December 4, 2020. ; https://doi.org/10.1101/2020.12.03.20243352 doi: medRxiv preprint symptoms were more salient among young respondents (Supplementary Table 3 ). For example, among respondents aged 15-64 years, the adjusted incidence rate of smell and taste disorder was higher for the subsidy program participants compared with the younger non-participants (adjusted rate, 3.4% vs. 2.4%; aOR, 2.03; 95%CI, 1.16-3.56; p=0.01), whereas the incidence rates did not differ between participants and non-participants among those aged 65-79 years 

Using the data from a large cross-sectional internet survey that included more than 25,000 adults in Japan, we found that individuals who participated in the government's subsidy program for domestic travel experienced a higher incidence of COVID-19-like symptoms compared with those who did not participate. This association was also observed for the incidence of smell and taste disorder, which is a highly specific symptom of the COVID-19 infection. 15, 26 This increased incidence of the COVID-19-like-symptoms was clustered among individuals aged <65 years, but not for those aged ≥ 65 years, suggesting that, if any, the non-elderly generation may be contributing to the spread of COVID-19 infection associated with this program. Given that the Japanese government is currently considering to halt this subsidy program with concerns for increased risks of COVID-19 infections, and other countries are also actively considering similar policies to stimulate the economy, our findings should be informative for policymakers to design policies that could increase economic activities without exacerbating the COVID-19 pandemic.

All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted December 4, 2020. ; https://doi.org/10.1101/2020. 12.03.20243352 doi: medRxiv preprint There are several mechanisms through which participation in this subsidy program was associated with a higher incidence of COVID-19-like symptoms. First, traveling itself may have led to a higher risk of incidence of COVID-19 due to increased contact with people in dining and sightseeing at the destination (causal effect). This explanation is supported by a recent genome epidemiological study of SARS-CoV-2 in Japan that found the possibility that the COVID-19 clusters in Tokyo metropolitan areas might have spread throughout Japan after the lifting of movement restrictions. 27 Second, the subsidy program participants might have been more likely to take behaviors that place them at greater risk of contracting the COVID-19 than the nonparticipants (selection effect). However, even if the findings were to be explained by this selection effect, our findings nevertheless indicate that the subsidy program may be incentivizing those who had higher risks of COVID-19 transmission to travel, leading to larger cases of infections. A better policy may be that incentivize individuals with a lower risk of the COVID-19 infection to travel, while those with a high risk to stay at home.

Our study has limitations. First, as with any observational study, we could not fully account for unmeasured confounders, and our study was unable to identify the exact mechanisms of the association between the subsidy program participation and increased incidence rates of COVID-19-like symptoms. Second, given the cross-sectional design of our study, we could not identify the temporal relationship between the subsidy program and the incidence of the COVID-19-like symptoms. Instead of the government subsidy causing the infection of the COVID-19, it was also possible that individuals who had experienced COVID-19-like symptoms were more likely to utilize the program and travel domestically. However, this explanation may be unlikely given All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted December 4, 2020. ; https://doi.org/10.1101/2020.12.03.20243352 doi: medRxiv preprint that travel agents and hotels have been introducing strict protocols to ensure that nobody with the COVID-19-like symptoms to use their services, and individuals who spread the virus are likely to face criticism and stigma in Japan incentivizing people with suspected symptoms to stay at home. 28 Third, it is likely that some individuals who reported five COVID-19-like symptoms had illnesses that were not COVID-19, as we were unable to collect the data on the confirmed diagnosis of COVID-19 infection (e.g., diagnosis using the PCR test). However, smell and taste disorders, one of the outcomes we used, are known to be highly specific (90% specificity) for the COVID-19 diagnosis, 15, 26 suggesting these symptoms would be good proxies of the incidence of COVID-19. Moreover, symptom-based measures would supplement the PCR test-based surveillance to understand a population-level picture of COVID-19 infection, 16, 17 because PCR testing will underestimate the true number of infections because not everyone with symptoms indicative of COVID-19 is tested. Fourth, our findings may be affected by the possibility that individuals who presented with COVID-19-like symptoms might recall and report using the subsidy program for domestic travel (as the cause of their symptoms) compared with those individuals without such symptoms (recall bias). Finally, because our study sample was collected through the web-based survey, our findings may not be generalizable to the population with limited access/literacy to the internet. Nevertheless, we used weighted analysis to minimize the difference in demographics, SES, and health-related characteristics between respondents of the current internet survey and the nationally representative survey, and thus would approximate our estimates to national estimates.

All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted December 4, 2020. ; https://doi.org/10.1101/2020.12.03.20243352 doi: medRxiv preprint Using a large-scale, concurrent, nationwide internet survey in Japan, we found that the participants in the government subsidies for domestic travel in Japan had higher incidence rates of COVID-19-like symptoms compared to the non-participants. Our findings suggest the implementation of the subsidy program for domestic travel might have contributed to the increased cases of the COVID-19 infection. In the midst of economic recession due to the COVID-19 pandemic, economic stimulus policies should incentivize individuals with low-risk of the COVID-19 infection to engage in economic activities while encouraging high-risk individuals to stay at home. 

World Health Organization. WHO Coronavirus Disease (COVID-19) Dashboard

Estimating the effects of non-pharmaceutical interventions on COVID-19 in Europe

Effects of non-pharmaceutical interventions on COVID-19 cases, deaths, and demand for hospital services in the UK: a modelling study. The Lancet Public Health

Physical distancing interventions and incidence of coronavirus disease 2019: natural experiment in 149 countries

The Health and Economic Impacts of Nonpharmaceutical Interventions to Address COVID-19: A Decision Support Tool for State and Local Policymakers. RAND Corporation

These destinations will basically pay you to come visit during the pandemic

NSW government considers restaurant and cafe vouchers to boost Covid-hit industry New South Wales. The Guardian

Japan's GoTo domestic tourism push stalls amid fears of Covid-19 "disaster" Coronavirus. The Guardian

Get a discount with the Eat Out to Help Out Scheme

Subsidizing the Spread of COVID19: Evidence from the UK's Eat-Out-to-Help-Out Scheme

About Us: Rakuten Insight

Awareness and use of electronic cigarettes and heat-not-burn tobacco products in Japan

Internet Survey (JASTIS): a longitudinal internet cohort study of heat-not-burn tobacco products, electronic cigarettes, and conventional tobacco products in Japan

Van den Bruel A. Signs and symptoms to determine if a patient presenting in primary care or hospital outpatient settings has COVID-19 disease

Using symptom-based measures for tracking COVID-19

An assessment of self-reported COVID-19 related symptoms of 227,898 users of a social networking service in Japan: Has the regional risk changed after the declaration of the state of emergency? The Lancet Regional Health -Western Pacific

Characteristics and risk factors for COVID-19 diagnosis and adverse outcomes in Mexico: an analysis of 89,756 laboratory-confirmed COVID-19 cases

Variation in COVID-19 hospitalizations and deaths across New York City boroughs

Selection bias in web surveys and the use of propensity scores

Labour and Welfare. Comprehensive Survey of Living Conditions

Individual income, income distribution, and self rated health in Japan: cross sectional analysis of nationally representative sample

Using the margins command to estimate and interpret adjusted predictions and marginal effects

A simple sequentially rejective multiple test procedure

Prevalence of colorectal neoplasm among patients with newly diagnosed coronary artery disease

Utility of hyposmia and hypogeusia for the diagnosis of COVID-19. The Lancet Infectious Diseases

A genome epidemiological study of SARS-CoV-2 in Japan as of

Japan's "virus vigilantes" take on rule-breakers and invaders. The Japan Times

We examined the association of the participation in the government subsidies for domestic travel in the past 1-2 months with the incidence of the five COVID-19-like symptoms in the past one month. For each outcome, we constructed a weighted multivariable logistic regression model with standard errors clustered at the prefecture-level. We adjusted for the respondents' sociodemographic characteristics, health-related characteristics, and prefecture indicator variables. We weighted the regression models using IPW to account for "being a respondent in an internet survey