Of all Americans who self-identify as Hispanic/Latino/Latinx, 30% had been infected with COVID-19 before May 2021.

Margin of error = ±4.9%

COVID-19 and language

by Sam Brill

Throughout the pandemic, COVID-19 has had a disproportionate effect on both the infection rate and the mortality rate of low-income people and racial and ethnic minorities.

It is important to understand why social class and race/ethnicity are related to the effects of COVID-19 and to isolate specific factors, such as employment and housing conditions or language-use, that may contribute to these disparities. Identifying specific factors can inform better policies and interventions.

Between April 23 and May 3, 2021, Ipsos surveyed a nationally representative sample of 1,449 individuals for the Tufts University Priority Area Research Group on Equity in Health, Wealth, and Civic Engagement. At that time, vaccines for COVID-19 were available to virtually all adults in the USA, and according to both the Tufts survey and the Centers for Disease Control and Prevention (CDC), about 60% of American adults had received the vaccine.

About 21% of survey respondents had either received a positive test for COVID-19 so far or believed they had contracted COVID-19 without a test to confirm the diagnosis. This rate was nearly double the estimated 10% of the population that had tested positive for COVID-19 according to the CDC.

Survey responses indicated strong disparities in rates of COVID-19. For example:

  • About 39% of the survey respondents who primarily spoke Spanish (and 38% of those who chose to take the survey in Spanish) had contracted COVID-19, compared to 20% of those who primarily spoke English. (The Tufts survey only asked about Spanish and English proficiency, because it was offered in those two languages. People who were proficient in neither Spanish nor English would not be able to participate. This is a limitation.)
  • About 31% of people who lived in households of five or more people had contracted COVID-19, compared to 20% of people who lived in households of 1-4 people.
  • About 19% of White, Non-Hispanic, 17% of Black, Non-Hispanic, 30% of Hispanic, 14% of 2+ Races, Non Hispanic, and 17% of other, Non-Hispanic people had contracted COVID-19.

Similarly, according to the CDC’s data tracker, only two ethnic groups exceed their percent of the population in terms of COVID-19 cases (Hispanic/Latino and those who report two or more race/ethnicities and are non-Hispanic). The limitation to the CDC’s data tracker is that race/ethnicity information is only available for 63% of confirmed cases of COVID-19 and does not account for those who suspected they had COVID-19 without receiving a positive test.

One way to assess the relative importance of such factors and to understand how they may interact is to develop a statistical model that predicts the outcome (in this case, contracting COVID-19) based on several variables.

I created a logistic regression model that estimates the odds of contracting COVID-19. The independent variables or predictors in the model were race/ethnicity, employment, the skill level of one’s job, whether one primarily speaks Spanish, the size of one’s household, age, gender, income, and region of residence based on state of residence. In this model, only the Hispanic ethnic category was a statistically significant predictor of contracting COVID-19 (1.7 times more risk for contracting COVID-19).

Race/ethnicity and primarily speaking Spanish could not both be included in the same model because they are too strongly correlated. However, when race/ethnicity was removed and language was used instead, people who were primarily Spanish-proficient had 2.36 times more risk for contracting COVID-19. People who opted to take the survey in Spanish had 2.25 times more risk for contracting COVID-19. None of the other factors were statistically significant predictors in this model.

The survey cannot directly address why language proficiency is related to COVID-19 outcomes. The explanation might have to do with access to information, experiences with the health system, or other unmeasured factors. The model suggests that the explanation is not the nature of jobs or household size, which are controlled for. This result suggests that more attention should be given to outreach and services for people who primarily speak Spanish.