What Can We Learn from the Travelers Data in Detecting Disease Outbreaks -- A Case Study of the COVID-19 Epidemic

TL;DR

This study demonstrates how traveler data can be effectively used to detect and monitor COVID-19 outbreaks, providing timely estimates of epidemic growth and informing policy decisions, with a new analytical framework and a user-friendly app.

Contribution

The paper introduces a method to estimate epidemic indicators from traveler data and compares it to domestic data, highlighting its potential for early outbreak detection.

Findings

Traveler data estimates aligned with domestic data when assumptions held.

Critical outbreak detection dates matched policy decision dates.

Recent traveler cases had a larger impact on estimates.

Abstract

Background: Travel is a potent force in the emergence of disease. We discussed how the traveler case reports could aid in a timely detection of a disease outbreak. Methods: Using the traveler data, we estimated a few indicators of the epidemic that affected decision making and policy, including the exponential growth rate, the doubling time, and the probability of severe cases exceeding the hospital capacity, in the initial phase of the COVID-19 epidemic in multiple countries. We imputed the arrival dates when they were missing. We compared the estimates from the traveler data to the ones from domestic data. We quantitatively evaluated the influence of each case report and knowing the arrival date on the estimation. Findings: We estimated the travel origin's daily exponential growth rate and examined the date from which the growth rate was consistently above 0.1 (equivalent to doubling time < 7 days). We found those dates were very close to the dates that critical decisions were made such as city lock-downs and national emergency announcement. Using only the traveler data, if the assumed epidemic start date was relatively accurate and the traveler sample was representative of the general population, the growth rate estimated from the traveler data was consistent with the domestic data. We also discussed situations that the traveler data could lead to biased estimates. From the data influence study, we found more recent travel cases had a larger influence on each day's estimate, and the influence of each case report got smaller as more cases became available. We provided the minimum number of exported cases needed to determine whether the local epidemic growth rate was above a certain level, and developed a user-friendly Shiny App to accommodate various scenarios.