Resilience of countries to COVID-19 correlated with trust

Abstract

We characterized > 150 countries' resilience to COVID-19 as the nationwide decay rate of daily cases or deaths from peak levels. Resilience to COVID-19 varies by a factor of ~ 40 between countries for cases/capita and ~ 25 for deaths/capita. Trust within society is positively correlated with country-level resilience to COVID-19, as is the adaptive increase in stringency of government interventions when epidemic waves occur. By contrast, countries where governments maintain greater background stringency tend to have lower trust within society and tend to be less resilient. All countries where > 40% agree "most people can be trusted" achieve a near complete reduction of new cases and deaths, but so do several less-trusting societies. As the pandemic progressed, resilience tended to decline, as adaptive increases in stringency also declined. These results add to evidence that trust can improve resilience to epidemics and other unexpected disruptions, of which COVID-19 is unlikely to be the last.

Figure 1

Comparing country-level COVID-19 resilience and reduction results for cases/capita, cases/tests, and deaths/capita. (a) resilience of cases/capita vs cases/tests (n = 100, $\rho$ =0.86, p < 0.0001). (b) resilience of cases/capita vs deaths/capita (n = 150, $\rho$ =0.61, p < 0.0001). (c) reduction of cases/capita vs cases/tests (n = 94, $\rho$ =0.83, p < 0.0001). (d) reduction of cases/capita vs deaths/capita (n = 136, $\rho$ =0.76, p < 0.0001). Linear correlation results tied to the origin are given within each figure panel and correspond to the dotted lines.

Figure 2

Comparing resilience to, and reduction of, COVID-19 across countries. (a) cases/capita (n = 165, $\rho$ =0.70, p < 0.0001), (b) deaths/capita (n = 150, $\rho$ =0.78, p < 0.0001).

Figure 3

World maps of country-level resilience to COVID-19. Decay rate (d⁻¹) from the first peak of: (a) cases/capita. (b) deaths/capita. In both cases countries are coloured where the fit of an exponential decay has r² ≥ 0.8. Countries in grey either have insufficient data or a poorer fit of exponential decay. Country/region boundaries plotted in R using the ‘maps’ package (ver. 3.3.0; https://CRAN.R-project.org/package=maps).

Figure 4

Country-level relationships between timing (day of year) of peak, resilience to COVID-19, and resulting reduction of cases and deaths. (a) cases/capita: relationships between day of year of peak cases and resilience (n = 176, $\rho$ =−0.51, p < 0.0001) and between day of year of peak cases and reduction (n = 164, $\rho$ =−0.54, p < 0.0001). (b) deaths/capita: relationships between day of year of peak deaths and resilience (n = 157, $\rho$ =−0.43, p < 0.0001) and between day of year of peak deaths and reduction (n = 150, $\rho$ =−0.39, p < 0.0001). Cases of complete reduction—i.e. elimination of cases or deaths—are denoted with pale blue. Cases where reduction is incomplete at the end of the time series are denoted with open circles.

Figure 5

Country-level relationships between adaptive stringency or trust, resilience to COVID-19 and resulting reduction of cases and deaths. (a) cases/capita: relationships between adaptive stringency and resilience (n = 167, $\rho$ =0.47, p < 0.0001) and between adaptive stringency and reduction (n = 156, $\rho$ =0.29, p < 0.001). (b) deaths/capita: relationships between adaptive stringency and resilience (n = 153, $\rho$ =0.39, p < 0.0001) and between adaptive stringency and reduction (n = 144, $\rho$ =0.21, p < 0.05). (c) cases/capita: relationship between trust and resilience (n = 77, $\rho$ =0.43, p < 0.0001) and between trust and reduction (n = 72, $\rho$ =0.51, p < 0.0001). (d) deaths/capita: relationship between trust and resilience (n = 75, $\rho$ =0.40, p < 0.001) and between trust and reduction (n = 72, $\rho$ =0.48, p < 0.0001). Note the threshold effect whereby trust > 40% (of population agreeing with the statement “most people can be trusted”) ensures resilience of cases/capita > 0.02 d⁻¹ and deaths/capita > 0.03d⁻¹, which in turn support successful reduction of cases and deaths. Cases of complete reduction—i.e. elimination of cases or deaths—are denoted with pale blue. Cases where reduction is incomplete at the end of the timeseries are denoted with open circles. The trust-reduction relationships are further analysed in Supplementary Fig. 4.

Figure 6

Optimised multiple linear regression models. (a) ln(resilience cases/capita) (n = 71, r² = 0.409; Supplementary Table 4). (b) ln(resilience deaths/capita) (n = 69, r² = 0.508; Supplementary Table 5). (c) cases/capita reduction (n = 66, r² = 0.352; Supplementary Table 6). (d) deaths/capita reduction (n = 66, r² = 0.414; Supplementary Table 7).

Figure 7

Relationships between peak timing, adaptive stringency, and trust: (a) day of year of peak cases/capita versus adaptive stringency (n = 166, $\rho$ =−0.74, p < 0.0001). (b) day of year of peak deaths/capita versus adaptive stringency (n = 154, $\rho$ =−0.72, p < 0.0001). (c) day of year of peak cases/capita versus trust (n = 77, $\rho$ =−0.37, p < 0.001). (d) day of year of peak deaths/capita versus trust (n = 75, $\rho$ =−0.30, p < 0.01).