Risk assessment of novel coronavirus COVID-19 outbreaks outside China

Peter Boldog; Tamas Tekeli; Zsolt Vizi; Attila Denes; Ferenc Bartha; Gergely Rost

The output C of the transmission model is used as the pool of potential travellers to abroad, and fed into the online platform EpiRisk [32] . This way, we evaluated the probability that a single infected individual is traveling from the index areas (in our case Chinese provinces other than Hubei) to a specific destination. Using a ten day interval for potential travel after exposure (just as in [15] ), one can find from EpiRisk that in the January-February periods, assuming usual travel volumes, there is a 1/554 probability that a single case will travel abroad and cause an exported case outside China. The dataset for relative importation risks of countries is available as well; thus, one can obtain the probability of an exported case appearing in a specific country. This probability is denoted by θ 0 , and we call it the baseline connectivity of that country with China. The baseline connectivity can be affected by other factors, such as the reduction in travel volume between the index and destination areas, exit screening in China, and the efficacy of entry screening at the destination country. Hence, we have a compound parameter, the actual connectivity θ, which expresses the probability that a case in China outside Hubei will be eventually mixed into the population of the destination country. For example, the relative risk of Japan is 0.13343, meaning that 13.343% of all exportations are expected to appear in Japan. Thus, under normal circumstances, the probability that a case from China eventually ends up in Japan is 0.13343/554 = 2.41 × 10 −4 during the January-February period [32] . Assuming a 20% reduction in travel volume between China and Japan, this baseline connectivity is reduced to a connectivity 0.8 × 2.41 × 10 −4 = 1.928 × 10 −4 . Additionally, assuming a 40% efficacy on entry screening [33] , there is a 0.6 probability that an arriving case passes the screening, and the connectivity parameter is further reduced to 0.6 × 1.928 × 10 −4 = 1.16 × 10 −4 . If we assume interventions at the originating area, for example, exit screening with 25% efficacy, then our actual connectivity parameter is θ = 0.75 × 1.16 × 10 −4 = 8.7 × 10 −5 , which represents the probability that a case in China will eventually mix into the population in Japan. Assuming independence, this θ, together with the cumulative cases C, generates a binomial distribution of importations that enter the population of a given country.