Objective We aimed to determine which aspects of the COVID-19 national response are independent predictors of COVID-19 mortality and case numbers. Design Comparative observational study between nations using publically available data Setting Worldwide Participants Covid-19 patients Interventions Stringency of 11 lockdown policies recorded by the Blavatnik School of Government database and earliness of each policy relative to first recorded national cases Main outcome measures Association with log10 National deaths (LogD) and log10 National cases (LogC) on the 29th April 2020 corrected for predictive demographic variables Results Early introduction was associated with reduced mortality (n=137) and case numbers (n=150) for every policy aside from testing policy, contact tracing and workplace closure. Maximum policy stringency was only found to be associated with reduced mortality (p=0.003) or case numbers (p=0.010) for international travel restrictions. A multivariate model, generated using demographic parameters (r2=0.72 for LogD and r2=0.74 for LogC), was used to assess the timing of each policy. Early introduction of first measure (significance p=0.048, regression coefficient β=-0.004, 95% confidence interval 0 to -0.008), early international travel restrictions (p=0.042, (β=-0.005, -0.001 to -0.009) and early public information (p=0.021, β=-0.005, -0.001 to -0.009) were associated with reduced LogC. Early introduction of first measure (p=0.003, β=-0.007, -0.003 to -0.011), early international travel restrictions (p=0.003, β=-0.008, -0.004 to-0.012), early public information (p=0.003, β=-0.007, 0.003 to -0.011), early generalised workplace closure (p=0.031, β=-0.012, -0.002 to -0.022) and early generalised school closure (p=0.050, β=-0.012, 0 to -0.024) were associated with reduced LogC. Conclusions At this stage in the pandemic, early institution of public information, international travel restrictions, and workplace closure are associated with reduced COVID-19 mortality and maintaining these policies may help control the pandemic.

Aims To generate a predictive model for the SARS-COV-2 viral reproductive rate, based on government policy and weather parameters. Methods A multivariate model for the log10 of viral reproductive rate was constructed for each country using lockdown stringency (Oxford University tracker), temperature and humidity, for the 1st 110 days of 2020. This was validated by extrapolating to the following 51 days, and comparing the predicted viral rate and cumulative mortality with WHO data. The country models was extrapolated to July 2021 using projected weather forecast for four scenarios; continuing with the 11/6/2020 lockdown policy, 100% lockdown, 20% lockdown and no lockdown. Results From pooled data (40 countries), lockdown stringency had a strong negative correlation with log10viral reproductive rate (-0.648 at 21 days later). Maximum temperature correlated at -0.14, 14 days later and humidity correlated at +0.25, 22 days later. Predictive Models were generated for 11 countries using multivariate regression of these parameters. The R2 correlation for log10R0 ranged from 0.817 to 0.987 for the model generation period. For the validation period, the Pearson's coefficient of correlation for log10R0 ranged from 0.529 to 0.984 and for cumulative mortality from 0.980 to 1.000. Forward extrapolation of these models for 5 nations, demonstrate, that removing the lockdown will result in rapid spread of the disease ranging from as soon as July 2020 for Russia, UK, Italy and India to January 2021 for the USA. The current (11/6/20) lockdown in the USA, Spain, UK, France, Germany, Turkey can control the disease but other nations will need to intensify their lockdowns to prevent future resurgence. Most nations will require more stringent lockdowns in January than in July. Conclusion The viral reproductive rate is highly predicted by a combination of lockdown stringency, temperature and humidity. Country specific predictive models can provide useful forecast of policy requirements.

Introduction In November 2020, a new SARS-COV-2 variant or the Kent variant emerged in the UK, and became the dominant UK SARS-COV-2 variant, demonstrating faster transmission than the original variant, which rapidly died out. However, it is unknown if this actually altered the overall course of the pandemic as genomic analysis was not common place at the outset and other factors such as the climate could alter the viral transmission rate over time. We aimed to test the hypothesis that the overall observed viral transmission was not altered by the emergence of the new variant, by testing a model generated earlier in the pandemic based on lockdown stringency, temperature and humidity. Methods From 1/1/20 to 4/2/21, the daily incidence of SARS-COV-2 deaths and the overall stringency of National Lockdown policy on each day was extracted from the Oxford University Government response tracker. The daily average temperature and humidity for London was extracted from Wunderground.com. The viral reproductive rate was calculated on a daily basis from the daily mortality data for each day. The correlation between log10 of viral reproductive rate and lockdown stringency and weather parameters were compared by Pearson correlation to determine the time lag associated with the greatest correlation. A multivariate model for the log10 of viral reproductive rate was constructed using lockdown stringency, temperature and humidity for the period 1/1/20 to 30/9/20. This model was extrapolated forward from 1/10/20 to 4/2/21 and the predicted viral reproductive rate, daily mortality and cumulative mortality were compared with official data. Results On multivariate linear regression, the optimal model had and R2 0f 0.833 for prediction of log10 viral reproductive rate 13 days later in the model construction period, with (coefficient, probability) lockdown stringency (-0.0109, p=0.0000), humidity (0.0038, p=0.0041) and temperature (-0.0035, p=0.0008). When extrapolated to the validation period (1/10/20 to 4/2/21), the model was highly correlated with daily (Pearson coefficient 0.88, p=0.0000) and cumulated SARS-COV-2 mortality (Pearson coefficient 0.99, p=0.0000). Conclusion The course of the SARS-COV-2 pandemic in the UK seems highly predicted by an earlier model based on the lockdown stringency, humidity and temperature and unaltered by the emergence of newer viral genotype.