Evaluating the impact of international airline suspensions on the early global spread of COVID-19

Aniruddha Adiga; Srinivasan Venkatramanan; James Schlitt; Akhil Peddireddy; Allan Dickerman; Andrei Bura; Andrew Warren; Brian D Klahn; Chunhong Mao; Dawen Xie; Dustin Machi; Erin Raymond; Fanchao Meng; Golda Barrow; Henning Mortveit; Jiangzhuo Chen; Jim Walke; Joshua Goldstein; Mandy L Wilson; Mark Orr; Przemyslaw Porebski; Pyrros A Telionis; Richard Beckman; Stefan Hoops; Stephen Eubank; Young Yun Baek; Bryan Lewis; Madhav Marathe; Chris Barrett

Using a linear model for the reported time of arrival against the effective distance computed on IATA data, we observe a moderately high coefficient of determination R 2 = 0.78. A scatter plot of COVID-19 ToA for countries with reported cases with respect to their effective distance from China is shown in Figure 1a . It should be noted that in the regression analysis, China has effective distance and time of arrival set to zero. The difference between reported ToA and linear estimator is provided in Figure 1b . The median error in estimation is less than a day with the estimated ToA typically being earlier than the reported ToA. Note that there are a few outliers which have much earlier reported ToA than estimated ToA (e.g., Nepal, Japan, Thailand). For these countries, airline traffic may not be representative of the total connectivity with China due to their spatial proximity (land or sea). In order to assess the risk of direct importation of COVID-19 from China to various countries we computed the estimated ToA for each country, using their effective distance from China (shown in Figure 2 ). The effective distance tree plot 3 clearly show that the early reporting countries indeed had the shortest effective distance to China (at the center). The predicted time of arrival using the linear estimator is shown in the map, excluding countries which do not have direct airline traffic from China. Finally, we plot the estimated times of arrival of all countries against their vulnerability (IDVI), grouped by the United States Combatant Commands, providing a natural geographic grouping. As demonstrated in Figure 2b , the countries with reported cases (shown with plus marker) are mostly to the left. Interestingly, most countries with reported cases also have higher IDVI, which may be due to a combination of (a) high connectivity and air traffic to China, (b) better detection and reporting capabilities for imported cases. In general, the lower left region (low IDVI and low estimated ToA) indicates a regime of high risk, which is relatively empty for this outbreak. However, we notice that several low IDVI countries (especially those in AFRICOM) are estimated to have times of arrival in the first three weeks of February. Equally concerning are countries to the left with circle markers, which have an early estimated time of arrival but haven't officially reported cases yet.