Estimation of COVID-19 outbreak size in Italy

Abstract

Italy is currently experiencing an epidemic of COVID-19 which emerged in the Lombardy region . During the interval between February 25-29, 2020, we identified 46 cases of COVID-19 reported in 21 countries in Europe, Africa, North America, and South America which were either in individuals with recent travel from Italy, or who had presumed infection by a traveler from Italy 2. In six cases, in four of the affected countries (Switzerland, France, Austria, Croatia), land travel was a likely route of introduction, or was documented to have been the route of introduction. We used air travel volume between Italian cities and cities in other countries as an index of connectedness, using data available from the International Air Transport Association (IATA) for February 2015 (2.61 million total departing international air passengers from Italy). We used the methods of Fraser et al. to estimate the size of the underlying epidemic in Italy necessary in order for these cases to be observed with a reasonable probability. To estimate the time at risk of COVID-19 exposure for travelers departing Italy, we obtained data from the United Nations World Tourism Organization (UNWTO) for the proportion of international travelers that are non-residents of Italy (63%) and the average length of stay of tourists to Italy (3.4 days), and assumed the Italian epidemic began one month prior to February 29, 2020. We also performed sensitivity analyses in which we included outbound travel to all countries regardless of reported case importations, inflated travel volumes by 35%, to account for the relative increase in flight numbers from 2015-2019, and excluded cases in bordering countries and which were documented to have been introduced by overland travel. When all cases were considered we estimated a true outbreak size of 3971 cases (95% CI 2907-5297), as compared to a reported case count of 1128 on February 29, 2020, suggesting non-identification of 72% (61-79%) of cases. In sensitivity analyses, outbreak sizes varied from 1552 to 4533 cases (implying non-identification of 27-75% of cases). We recently used similar methods to estimate a much larger epidemic size in Iran, with a far greater degree of under-reporting, based on many fewer exported cases. The reason for this difference relates to the relatively high volume of travel from Italy, relative to Iran. In summary, we suggest that the numerous COVID-19 case exportations from Italy in recent days suggest an epidemic that is larger than official case counts suggest, and which is approximately on a par with that currently occurring in South Korea, which reports 3526 cases (and fewer deaths) as of February 29, 2020.