Race, Place, and COVID-19: Mapping and Modeling a Spatial Relationship

This post was independently organized by graduate students enrolled in Geography 5103, instructed by Professor Elisabeth Root, during spring semester of 2020. 

By now, most of us are familiar with the risk factors for severe COVID-19 disease: being over 65, male, and suffering from heart problems, diabetes, or hypertension, all seem to contribute to mortality (Harrison, 2020; Rogers, 2020; Wadman, 2020). What hasn’t been discussed as frequently is how the social determinants of health – the neighborhood conditions in which people are born, grow, live, work, and age that affect health (Florida, 2020) – may also impact the risk of contracting and dying from COVID-19. While the health outcomes of a particular individual cannot be predicted solely by the social determinants of health, these measures allow researchers and the interested public to engage with social inequality in public health resources and to devise location-specific improvements.

For COVID-19 this means that these risk factors vary across space. For example, there are some places where the percent of the population that is obese has a strong positive correlation with mortality, while in other places there is no association whatsoever between the two variables. Understanding these place-specific nuances is key in organizing an effective response to COVID-19 outbreaks in a region or community.

As students, we decided to examine whether the relationship between COVID-19 mortality and area-level factors such as race, age, and poverty, varied across counties in the United States. This is a spatial property called “spatial nonstationarity,” meaning some determinant of health (e.g., race) does not have the same effect on an outcome (e.g., COVID-19 mortality) across space. We used a modeling technique called Geographically Weighted Regression (GWR) which we learned about in GEOG 5103, Dr. Elisabeth Root’s class. It is particularly well suited to capturing different effects of potential determinants on mortality across space. We chose the number of deaths per one thousand people as our outcome of interest, and then picked potential explanatory variables based on the COVID-19 research to date (05/14/2020). Among a large set of candidates, seven variables were selected. The variables are rurality, percent of adults reporting to be obese, percent of adults reporting to have asthma, percent of Black or African American, percent of seniors (65 or older), percent of people living in poverty, and the number of confirmed cases. These variables were found to be significant predictors and explain 43% of the total variation of COVID-19 mortality.

The maps below show how the relationship between each variable and COVID mortality varies across the country. So, for example, in places where the map is blue, there is actually a negative relationship between poverty and fatality (Figure 1). However, as the color becomes tan and then red, we begin to see both a stronger relationship and a positive one. These maps also account for the p-value. This means that the shaded areas show us where our results are statistically significant- that is, that the result displays a pattern or strength of the relationship that cannot just be attributed to random noise.

Figure 1. Relationship between the percentage of poverty and COVID-19 deaths: The impact of poverty on mortality is negative in many parts of the country including Ohio, Mississippi, and South Carolina. Exceptionally, positive relationships are clustered in Minnesota and Oklahoma, indicating higher risks in poor neighborhoods.

What you can see in these maps is that while all the variables have geographic variation, the counties where the African American population is a higher percentage of the population show a strong correlation with our outcome variable of fatalities. It’s also important to note the direction of these relationships. For the percentage of seniors, for example, much of the West Coast displays a negative relationship between the variable and the outcome: there, an increase in the percent of seniors was associated with a decrease in COVID-19 mortality (Figure 2). In contrast, large portions of the country show a positive relationship between the percentage of African Americans in a county and COVID-19 mortality (Figure 3). Indeed, of all our variables, this is only the one that displays such a consistently positive relationship across many different locations.

Figure 2. Relationship between the percentage of seniors and COVID-19 deaths: Fewer seniors are related to higher fatality in several parts including the West, Louisiana, and Illinois. On the contrary, the positive relationships are mostly clustered in Florida and parts of Ohio, indicating higher percentages of seniors tend to associate with higher fatality in those regions.

These results do not indicate that African Americans are somehow inherently more susceptible to this disease but instead reflect the structural racism of our country. In the United States, African Americans are more employed in service occupations (U.S. Bureau of Labor Statistics, 2018) and are more likely to live in a neighborhood with higher poverty (Greene et al., 2017). These social determinants of health make them particularly vulnerable to this disease because these risk factors lead to higher rates of exposure and poorer access to health care. Inequalities in health care access are particularly problematic for diseases like COVID-19. The longer a person waits to get necessary care – often because of a lack of insurance or health care resources in their community – the more likely they are to have severe COVID-19 disease. Recognizing the relationship between the social determinants of health and race, and then identifying how those relationships manifest in local contexts are the first steps in addressing the profound inequality presented in these maps.

Figure 3. Relationship between the percentage of African Americans and COVID-19 deaths: While higher percentages of African Americans tend to associate with higher fatality the Southern and Eastern states, Minnesota in particular show the highest correlation. Given the fact that Minnesota has a lower percentage of African Americans compared to Southern states, Black communities in the state seem to be exposed to relatively higher risks.

 

Author Contribution: Dr. Elisabeth Root (Professor, Geography & College of Public Health, Division of Epidemiology) conceived the research and supervised writing; Sohyun Park (Ph.D. candidate, Geography) collected data and supervised analysis; Yun Ye (Ph.D. student, Public Health), Kaiting Lang (Ph.D. student, Public Health), and Junmei Cheng (Ph.D. student, City and Regional Planning) analyzed the data; Anisa Kline (Ph.D. student, Geography) wrote the post; Blake Acton (MA student, Geography) created maps. All the authors contributed to interpreting the results.