Government's policy, citizens' behavior, and COVID-19 pandemic.

To the Editor,

The government's policy responses and citizens' mobility behavior against coronavirus disease 2019 (COVID‐19) may play an important role in avoidance of the pandemic. Their relations, however, have been seldom investigated to date because it is difficult to quantify the government's policy and citizens' behavior. In the present study, the relationships of government's policy, citizens' behavior, and COVID‐19 prevalence and case‐fatality were quantitatively analyzed using a Government Response Stringency Index (GRSI) provided by the “Oxford COVID‐19 Government Response Tracker” (OxCGRT; https://www.bsg.ox.ac.uk/research/research-projects/coronavirus-government-response-tracker) and mobility changes purveyed by the “Google COVID‐19 Community Mobility Reports” (https://www.google.com/covid19/mobility/?hl=en).

For each country, daily GRSI taking a value 0–100 (100 = strictest response; available since 2020/1/1) and mobility changes (%) from the baseline day being the median value during the 5‑week period of 2020/1/3–2/6 (including retail & recreation, grocery & pharmacy, parks, transit stations, workplaces, and residential; procurable since 2020/2/15) were extracted and averaged (during the period of 2020/1/1–6/30 for GRSI and 2020/2/15–6/30 for mobility changes). Total COVID‐19 confirmed cases and deaths on 2020/7/15 (2 weeks after 2020/6/30 on which the latest data of GRSI/mobility‐changes were extracted) were available on the “WHO COVID‐2019 Situation Reports” (https://www.who.int/emergencies/diseases/novel-coronavirus-2019/situation-reports/). The relations between the mean mobility changes and mean GRSI were investigated with simple linear regression using “Microsoft Excel for Mac” version 16.38, and those between mean COVID‐19 prevalence/fatality and GRSI/mobility‐changes were analyzed with inverse‐variance (of prevalence/fatality) weighted linear regression (i.e., random‐effects meta‐regression) using OpenMetaAnalyst (http://www.cebm.brown.edu/openmeta/index.html).

The GRSI and mobility changes were available in 173 and 129 countries, respectively (Table S1). The mobility changes in retail/recreation (Figure S1), grocery/pharmacy (Figure S2), parks (Figure S3), transit stations (Figure S4), and workplaces (Figure 1A) were significantly (p < .001) and negatively (coefficient < 0) associated with the GRSI, and those in residential (Figure 1B) were significantly (p < .001) and positively (coefficient > 0) correlated to the GRSI. The results of the inverse‐variance weighted regression are summarized in Table 1. The COVID‐19 prevalence was significantly (p < .001) and positively associated with the GRSI (Figure 1C) and the changes in residential (Figure S5), and significantly (p < .001) and negatively correlated to the changes in retail/recreation (Figure S6), grocery/pharmacy (Figure S7), transit stations (Figure S8), and workplaces (Figure 1D). The COVID‐19 fatality was significantly (p < .001) and positively associated with the changes in parks (Figure S9), and significantly (p = .030) and negatively correlated to the GRSI (Figure 1E).

Table 1

Results of the inverse‐variance weighted regression

Covariate		COVID‐19 prevalence			COVID‐19 fatality
		Coefficient	p value	Figure	Coefficient	p value	Figure
Mean GRSI		0.080	<.001	1 C	–0.020	.030	1E
Mean community mobility changes (%)	Retail/recreation	–0.071	<.001	S6	0.000	0.986
	Grocery/pharmacy	–0.048	<.001	S7	0.012	0.129
	Parks	–0.001	.918		0.010	< .001	S9
	Transit stations	–0.057	<.001	S8	0.001	0.934
	Workplaces	–0.105	<.001	1D	–0.012	0.133
	Residential	0.134	<.001	S5	–0.028	0.086

Note: Bold values mean statistical significance.

Abbreviation: GRSI, Government Response Stringency Index. Bold values mean statistical significance.

Relations of government's policy, citizens' behavior, and COVID‐19 prevalence and case fatality. (A) Significant (p < .001) and negative (coefficient, –0.772) association (simple regression) of mean changes (%) in workspaces (y‐axis) with mean Government Response Stringency Index (x‐axis); (B) significant (p < .001) and positive (coefficient, 0.360) correlation (simple regression) of mean changes (%) in residential (y‐axis) to mean Government Response Stringency Index (x‐axis); (C) significant (p < .001) and positive (coefficient, 0.080) association (inverse‐variance weighted regression) of COVID‐19 prevalence (y‐axis) with mean Government Response Stringency Index (x‐axis); (D) significant (p < .001) and negative (coefficient, –0.105) correlation (inverse‐variance weighted regression) of COVID‐19 prevalence (y‐axis) to mean changes in workplaces (x‐axis); (E) significant (p = .030) and negative (coefficient, –0.020) association (inverse‐variance weighted regression) of COVID‐19 fatality (y‐axis) to mean Government Response Stringency Index (x‐axis)

The present results suggest that citizens may reduce their mobility and stay at home according to the stringency of government policy, and a stricter policy may contribute to a lower COVID‐19 fatality. The associations of GRSI and mobility changes with COVID‐19 prevalence indicated in the present study, however, should be carefully interpreted. The positive correlation of stricter policy to higher prevalence may not denote that a stricter policy causes higher prevalence. Simply (not causatively), a stricter policy implemented in a country where COVID‐19 cases are increasing may be invalid in evading the further COVID‐19 epidemic and result in a higher prevalence. Fewer citizens' mobility induced by a stricter policy may also ineffective for the avoidance of the outbreak. A stay‐at‐home order, one of the strictest policy responses, however, has been reported to be valid against COVID‐19 infection , and hospitalization. The present findings also suggest that a stricter policy (probably including a stay‐at‐home order) may be associated with lower COVID‐19 fatality. The GRSI and mobility changes used in the present study could be applied to further quantitative investigations of relationships among the government's policy, citizens' behavior, and the COVID‐19 pandemic.

CONFLICT OF INTERESTS

The authors declare that there are no conflict of interests.

Supporting information.

Click here for additional data file.

Supporting information.

Click here for additional data file.