Impact of climatic, demographic and disease control factors on the transmission dynamics of COVID-19 in large cities worldwide

Abstract

We are now over seven months into a pandemic of COVID-19 caused by the SARS-CoV-2 virus and global incidence continues to rise. In some regions such as the temperate northern hemisphere there are fears of "second waves" of infections over the coming months, while in other, vulnerable regions such as Africa and South America, concerns remain that cases may still rise, further impacting local economies and livelihoods. Despite substantial research efforts to date, it remains unresolved as to whether COVID-19 transmission has the same sensitivity to climate and seasonality observed for other common respiratory viruses such as seasonal influenza. Here we investigate any empirical evidence of seasonality using a robust estimation framework. For 304 large cities across the world, we estimated the basic reproduction number (R0) using logistic growth curves fitted to cumulative case data. We then assessed evidence for association with climatic variables through mixed-effects and ordinary least squares (OLS) regression while adjusting for city-level variation in demographic and disease control factors. We find evidence of association between temperature and R0 during the early phase of the epidemic in China only. During subsequent pandemic spread outside China, we instead find evidence of seasonal change in R0, with greater R0 within cities experiencing shorter daylight hours (direct effect coefficient = -0.247, p = 0.006), after separating out effects of calendar day. The effect of daylight hours may be driven by levels of UV radiation, which is known to have detrimental effects on coronaviruses, including SARS-CoV-2. In the global analysis excluding China, climatic variables had weaker explanatory power compared to demographic or disease control factors. Overall, we find a weak but detectable signal of climate variables on the transmission of COVID-19. As seasonal changes occur later in 2020, it is feasible that the transmission dynamics of COVID-19 may shift in a detectable manner. However, rates of transmission and health burden of the pandemic in the coming months will be ultimately determined by population factors and disease control policies.

The SARS-CoV-2 pandemic: A syndemic perspective

“…The relationship between COVID-19 and environmental factors is complex [ 78 , 115 ]. Recent studies have demonstrated a positive correlation between spread and decay of SARS-CoV-2 and temperature, absolute humidity and population density [ 116 ] and a positive correlation between the transmissibility of the SARS-CoV-2 and lower UV radiation, lower latitudes, lower elevation, and smaller populations with historical air pollution exposure [ 117 ]. No association was detected between deaths and country temperature or precipitation [ 118 ].…”

Is coronavirus disease (COVID-19) seasonal? A critical analysis of empirical and epidemiological studies at global and local scales

“… Relation Country Statistical method used Ref. Negative Global (≥50 countries) Mapping to Köppen's climate ( Bajaj and Arya, 2020 ) Spearman correlation ( Bezabih et al, 2020 ) Simple LR model; Pearson correlation ( Hassan et al, 2020 ) Simple LR model ( Iqbal et al, 2020 ) Multivariate LR model ( Luo et al, 2020 ) Stepwise LR model ( Metelmann et al, 2020 ) Simple LR model ( Notari, 2021 ) Mann-Whitney test; LR model ( Sajadi et al, 2020 ) Pearson correlation ( Sarmadi et al, 2020 ) Gutenberg–Richter law model ( Sil and Kumar, 2020 ) Uni/Multivariate LR model (by panel data strategy) ( Sobral et al, 2020 ) GAM ( Wu et al, 2020b ) Brazil GAM; Polynomial regression model ( Prata et al, 2020 ) Shapiro-Wilk test; Spearman correlation ( Rosario et al, 2020 ) China SEIR model with simple LR model; Pearson correlation ( Guo et al, 2020 ) Multivariate Poisson regression model …”

“…In another simple LR analysis, average high temperature was reciprocally related to both total deaths (β = -0.0377, R 2 = 0.4192) and total cases per million population (β = -0.0037, R 2 = 0.5875) ( Iqbal et al, 2020 ). A stepwise LR model fitted with R 0 displayed a negative relationship between temperature and SARS-CoV-2 infectivity ( Metelmann et al, 2020 ). Fitting a simple LR on the exponential spread rate with the average temperature again exhibited a decreasing slope (β = -0.0035, R 2 = 0.18) ( Notari, 2021 ).…”

“…Several other studies analyzed variables from the Susceptible-Exposed-Infectious-Recovered (SEIR) model with regression analysis and showed that the transmission rate was negatively related to temperature ( Guo et al, 2020 ; Lin et al, 2020b ). Similarly, a stepwise LR model, used to analyze data collected up to May 2020, when the COVID-19 spread was largely under control in China, displayed a negative relationship between the temperature and COVID-19 infectivity ( Metelmann et al, 2020 ).…”

“…A weighted random-effects regression analysis was concluded and showed that the epidemic growth of COVID-19 was reciprocally associated with both RH (ratios of rate ratios [RRR] per 10% = 0.91, 95% CI = [0.85, 0.96]) and AH (RRR per 5 g/m 3 = 0.92, 95% CI = [0.85, 0.96]) ( Jüni et al, 2020 ). Congruently, studies using a LR model revealed a negative correlation between humidity and SARS-CoV-2 transmission ( Metelmann et al, 2020 ; Sajadi et al, 2020 ). Only one study showed no correlation of humidity with global transmission of SARS-CoV-2 at all ( Islam et al, 2020 ) ( Table 2 ) ( Table 3 ).…”