Study population and setting
This study used daily estimates of Rt (the time-varying reproductive number of SARS-CoV-2, or the daily estimate of the number of people each newly infected person is expected to infect) and the dates of implementation of country-level non-pharmaceutical interventions (NPI) policies to assess the impact of the introduction and lifting of NPI policies on changes in SARS-CoV-2 transmission across 131 countries between January 1 and July 20, 2020. Daily estimates of Rt were derived from EpiForecasts, a previously validated epidemiological model, while information on policies was obtained from the Oxford COVID-19 Government Response Tracker. The study used statistical modelling to estimate the impact of individual NPI policies on Rt and the timing of their maximum impact on transmission following introduction or lifting. Briefly, this was done by dividing each country’s timeline into phases (i.e., time periods) in which there was no change in policy: a change in policy marked the start of a new phase. The change in Rt between phases was the primary outcome. Log-linear regression was used to examine the association between policies and the primary outcome as well as interactions between lifting and introduction of combinations of policies.
Summary of Main Findings
Most NPIs were associated with reduced Rt, including school and workplace closures, public events bans, stay at home requirements, and internal movement limits. Bans on public events were associated with the largest reduction in Rt following their introduction (24% by 28 days). With regard to lifting policies, lifting school closures and bans on public gatherings of more than ten people were the only NPIs associated with statistically significant increases in Rt. Introducing an NPI policy was found to have a delay of ~1 week between when the policy was enacted and when 60% of its expected total impact took place. Lifting restrictions was found to have a longer delay of ~3 weeks between lifting and 60% of total impact.
The study used standardized and well-characterized datasets. The policy phase approach was a sensible framework for estimating the impact of individual NPI policies, and has been used previously. Another strength was that lags in policy impact were allowed to occur over the course of four weeks rather than fixed at particular times in contrast to earlier studies. Further, the authors performed a series of sensitivity tests to examine various ways in which their results might change based on different assumptions, improving the plausibility that the effects they estimate represent real impacts rather than artifacts of the models and methods used.
There are several limitations to this study. First, compliance with policies was not assessed and therefore not captured in the models. Relatedly, individuals may change behaviors that impact transmission irrespective of policy changes (e.g., continue to stay home and avoid high risk exposures despite lifting of interventions). Second, multiple policies were often implemented simultaneously or near simultaneously which makes it difficult to plausibly disentangle their individual impacts despite the methods used. Lastly, the biggest impacts tended to be estimated for those policies that were implemented and lifted earlier. The timing of introduction or lifting of policies relative to others may affect the impact of policies implemented at subsequent time points.
This is the first study to estimate the impact of both the introduction and lifting of non-pharmaceutical intervention policies on SARS-CoV-2 transmission.
This review was posted on: 20 November 2020