Does Geographic Residency Predict Risk of COVID-19?

The extended travel restriction between the U.S. and Canada will expire on May 21, 2020. Although it is up to the two governments to make the decision regarding when to restore the normal border operation, two main arguments against lifting the restriction seem flawed. Both regard residents of the U.S. as linked with a major risk factor for the infection.

The first opinion emphasizes the risk of spreading the virus to Canada by travelers from the U.S. For example, the Global News recently cited a professor at the University of Ottawa as saying that because the U.S. has the biggest epidemic in the world with less adherence to social distancing, removing the restriction is likely to put Canadians at risk. Using the numbers in the report, we can see the infection rates for the two countries: There are about 30 confirmed cases per 10,000 population in the U.S. and about 12 cases per 10,000 population in Canada, respectively.

He assumes that the geographic “group” with the higher infection rate will proportionally carry a higher risk for the other. However, this assumption lacks supporting evidence.

Let’s look at two examples:

1. The Health Canada website reported that during the flu season of 2018-19, there were 12,200 flu-related hospitalizations in Canada. Adjusted for the population, it shows about 3 hospitalizations per 10,000 Canadians. At the same time, the American CDC reported there were 490,600 flu-related hospitalizations, or about 15 cases per 10,000 Americans.

In spite of the gigantic contrast in the infection rate between the two countries, there was no shutdown or restriction at the border during the season. People on both sides traveled freely. There is no evidence showing that travelers from the much higher flu-infected U.S. exacerbate the condition in Canada.

2. Washington State and Oregon are two bordering states. Washington has three times the confirmed COVID-19 cases that of Oregon. The most recent data (April 29, 2020) shows that Washington has 18 cases per 10,000 population, whereas Oregon has less than 6 cases per 10,000 population. There has never been any travel restriction between the two regions. Greater Portland in Oregon and Clark County in Washington are so connected that many people working or studying in Portland reside in Clark County. People in WA drive over Columbia River for shopping, recreation, education, and other activities because of Oregon's clusters of business with no sales tax and numerous opportunities. With the stay-at-home stipulation by the states in the last several weeks, Oregon local media reported the driving has not been reduced substantially. In short, Washington’s higher cases have not aggravated the COVID-19 situation in Oregon via commuting.

The second opinion against lifting the border restriction is based on a recent survey of COVID-19 patients in Ontario, Quebec, Alberta, and British Columbia, about their international travel experiences early this year. As reported in the National Post, those who have traveled to the U.S. account for the largest percentage among the participants. The other top international destinations include the United Kingdom, France, Puerto Rico, Austria, cruise ships, Mexico, Iran, and Italy. The number of cases related to travel from China is negligible.

However, although the largest percentage among the participants involves traveling to the U.S., this fact is insufficient to establish the association between U.S. travel and the infection. This is because the data only suggest the exposure to risk (traveling to the U.S.) and the absence of exposure (not traveling to the U.S.) for the cases, but lack information about the controls for comparison.

In epidemiology, a case-control study can answer the question about the association between U.S. travel and the infection. Researchers need to include the cases and a comparable group of people (e.g., similar in age, health, and gender) as the controls who do not show the outcome—COVID-19. The analysis must integrate information about the exposure (in this case traveling to the U.S.) and the absence of exposure in the controls before calculating the odds ratio between the cases and controls. If a sizable portion of the controls also have a recent trip to the U.S. (that is, a similar exposure), the result is likely to suggest a weak or non-existent association between the exposure and the outcome.

In short, it is inaccurate to assume that individuals in a geographical area have a high or low risk for transmitting an infectious disease just because their geographic “group” has a high or low percentage of the disease cases. The reason is that individuals classified by similar geographic locations may be quite dissimilar in the dimensions related to how the disease is developed and transmitted.