Assessing global border controls in response to COVID-19 pandemic using real-world data and target trial emulation

To examine whether more stringent border controls (BCs), including quarantine, ban on regions, and total border closure, are associated with slower initial-stage epidemic progression compared to screening.

Using real-world BCs and COVID-19 epidemic data from 174 countries and regions from 1 January to 31 July 2020, we compared the effectiveness of different stringency BCs in slowing countries from reaching the first infection peak. To account for the immortal time bias due to staggered BC adoptions across countries, the target-trial-emulation and cloning-censoring-weighting approaches were applied. Kaplan-Meier model with inverse probability censoring weights (IPCW) was used to obtain effectiveness estimates for each BC. Country-specific timing and methods of implementing diagnostic testing were not included in the IPCW calculation model due to data unavailability. BC adoption timing was investigated as a potential mechanism for BCs' effectiveness.

Compared with countries adopting basic border screening, those adopting more stringent BCs were not associated with a reduced risk of reaching the first infection peak, including the most stringent total border closure (Hazard Ratio [HR]: 1.07, 95% CI: 0.88-1.30). Countries adopting those stringent BCs experienced significant delays than those adopting screening potentially elucidating the mechanism. Further analysis revealed that BC adoption timing, regardless of stringency, was independently associated with the risk of the first infection peak attainment.

More stringent BCs in response to pandemics did not necessarily lead to slower initial-stage epidemic progression if adoptions were delayed. These findings highlight the importance of considering both timing and stringency-rather than focusing on stringency alone-when adopting BCs against future emerging infectious disease surges.