The GenderSci Lab releases a US State COVID-19 Report Card
Author: Marion Boulicault
How well is your state reporting on socially relevant COVID data?
For updated Report Card and highlights of how state data reporting has shifted since this July, 2020 blog post, see the GenderSci Lab’s US State Data Report Card page.
Socially relevant variables like gender/sex, age, race and/or ethnicity, and comorbidity status have significant impacts on COVID outcomes. To take one of the starkest examples: Non-Hispanic Black Americans, Native Americans, and Latinos are being hospitalized for COVID at approximately four to five times the rate of Non-Hispanic White individuals. Gender/sex also seems to be important; rates vary widely, but in many locales, more men die of COVID than women. Comorbidities are also critical influencers of COVID outcomes; 93% of deaths attributed to COVID mentioned additional conditions contributing to the fatality, with an average of 2.5 additional comorbidities per death being reported. Comorbidities likely mediate some of the racial and gender/sex disparities above, but also important are demographic factors such as age, and exposure factors such as occupation, transportation availability, neighborhood density, and other social factors.
At the most basic level, to understand the nature and extent of COVID outcome disparities, we must have data on variables like gender/sex, age, race/ethnicity, and comorbidity status. But no variable works independently; socially relevant variables are in constant interplay and must be analyzed in correspondence with one another. We need such interactional data, for example, to understand how apparent sex differences may vary between ethnicities, and how that relates to comorbidity status. Understanding these interactions, in turn, is the foundation of situating COVID disparities in their social context. The reason why some ethnicities have such starkly high rates of COVID is related to histories of structural inequalities, limited access to health care, neighborhood density and shared housing, transportation access, and many other factors. Likewise, simple analysis of higher death rates among “men” creates the illusion that there is a singular “male” story. COVID outcomes, like those of past pandemics, are likely to vary widely by sex in interaction with social class, occupation, and race/ethnicity. Interactional data is critical to understanding COVID outcomes, to crafting effective, tailored public health strategies, and to supporting communities during the pandemic.
Since April, the GenderSci Lab has been monitoring the COVID reporting landscape in the US, aiming to answer the question: what data is available and where?
Today, on the Health Affairs blog, we published the findings of our US State COVID-19 Report Card (Report Card below). To create this Report Card, we gathered information on the availability of surveillance data in each state (and the District of Columbia) using state health department web pages provided by the Centers for Disease Control and Prevention (CDC) and the COVID Tracking Project. We assigned each state a grade based on the adequacy of their reporting. What we found is concerning. US state reporting on socially relevant variables is dangerously inadequate. In what follows, we show how we created the Report Card, and explain what it tells us about US COVID-19 reporting.
Additionally, to both encourage and track ongoing improvements, we will be updating the Report Card on a monthly basis and hosting it on our website. The site also hosts state grades for the month of April and May.
“By illustrating widespread deficiencies as well as highlighting local successes, our Report Card aims to hold states accountable for their data reporting practices.”
Report Card current as of 6/26/20
How we created the Report Card
The Report Card is based on a 0-10 scoring scheme for state surveillance reporting. For case reporting, a state could earn one point for each category of age, gender/sex, race/ethnicity, and comorbidities reported. Any reporting on the interactions of these first four variables also earned a point. Similarly, a state earned a point for each of these variables in their reporting of deaths, plus a point for reporting any interactions. We assigned letter grades based on total scores: F for a score of 5 or below, D for a score of 6, C for a score of 7, B for a score of 8, and A for a score of 9 or 10.
What the Report Card tells us
US state reporting on socially relevant variables is dangerously inadequate (see Map below). As of June 26, 2020:
The average grade across the US is a D (mean score of 6.41 out of 10 (SD=1.50)).
Deficiencies in reporting appear throughout the country and are not limited to a specific geographic region. The mean score for the Midwest is 6.66 (SD=1.80); the Northeast is 6.44 (SD=0.50); the South is 6.59 (SD=1.33), and the West is 5.92 (SD=1.73).
The highest-scoring states are Georgia and Iowa, which each report data on all four socially-relevant variables plus data on at least one interaction for both deaths and cases.
The lowest-scoring state is Hawaii, which reports only age and race/ethnicity data for cases, and no data at all on deaths.
A total of 18 states are reporting any type of interaction between the four socially-relevant variables for cases, fatalities, or both cases and fatalities.
Age was the most frequently reported variable, with 50 states reporting age for cases and 48 reporting age for fatalities.
Comorbidities were the least frequently reported variable, with only 5 states reporting comorbidities among cases, and 11 reporting comorbidities among fatalities.
Map current as of 6/26/20
The importance of interactional data: Examples from our Health Affairs Blog Post
To illustrate the crucial importance of interactional data, it helps to look in detail at particular states. In our Health Affairs piece (hyperlink again), we discuss the case of Georgia, one of the few states reporting data on the interactions between sex/gender and race/ethnicity.
“As of July 8, 2020, the Georgia Department of Public Health reports that 51.5% of the total number of COVID-19 cases occurred among women. However, restricting the analysis to individual race/ethnicity categories shows that the sex/gender distribution of cases varies considerably. While White American women constitute 51.6% of all White American cases, African American females constitute 58.5% of all African American cases. These data suggest a need to explore whether and how social variables linked to race/ethnicity, such as housing density, neighborhood, and occupation, influence sex/gender disparities in COVID-19 outcomes.
California offers another example, this time illustrating the importance of reporting data on interactions between race/ethnicity and age. As of July 8, 2020, California reports case and fatality data disaggregated by both age and race/ethnicity. Overall, 42.3 percent of people who died of COVID-19 California are Latinx. Considering that 38.9 percent of the population in California is Latinx, this data point alone might not present obvious avenues for further investigation. However, a more complicated picture arises when examining COVID-19 fatalities that include interactions between age and race/ethnicity. Despite constituting only 41.5 percent of the Californian population between ages 35 – 49, Latinx individuals make up 77 percent of total deaths in that age category. Because the percentage of Latinx individuals decreases relative to other racial/ethnic groups as the population ages, averaging death rates across all age groups gives the illusion that COVID-19 fatality rates are approximately proportionate for the Latinx population. But disaggregating by both age and ethnicity reveals a dramatic disparity in health outcomes, adding nuance to our understanding of how social factors linked to both age and race/ethnicity shape COVID-19 outcomes.”
A Tool for Accountability and a Call to Action
By illustrating widespread deficiencies as well as highlighting local successes, our Report Card aims to hold states accountable for their data reporting practices. States should, at minimum, report cases and deaths by age, gender/sex, race/ethnicity, comorbidity status, and interactions between these variables. Reporting should be easily accessible and available to the public. Ideally, each state should make de-identified, HIPAA compliant, individual-level data open-access and downloadable, so that the broader research community can contribute to time-sensitive and crucial data analysis. Doing so is essential for understanding and addressing inequities in the COVID-19 pandemic.
These failures aren’t inevitable: states can do better. When we began tracking this data on April 18, only 9 states were reporting on any interactions between variables for either case or fatality data. A little over two months later, that total has risen to 18. This Report Card should serve as an urgent call to action for those states still receiving low grades and encourage state-level cooperation in the development and sharing of best practices.
The GenderSci Lab COVID Project
The Report Card is one piece of the GenderSci Lab’s broader COVID Project. Through this project, the GenderSci Lab aims to contribute to a broad-based effort to elevate the consideration of social variables in COVID-19 outcomes, improve theoretical frameworks and data quality for evaluating gender/sex hypotheses, and identify meaningful interventions to address gender/sex disparities.
For more about how US states vary in COVID cases and deaths, see our US Gender/Sex COVID-19 Data Tracker Page and our accompanying Data Highlights.
We have also written a Communication Guide for journalists and researchers working on gender/sex disparities in COVID-19.
Recommended Citation:
Boulicault, Marion. “The GenderSci Lab releases a US State COVID-19 Report Card ” GenderSci Blog, July, 14, 2020. https://www.genderscilab.org/blog/data-report-card
Contact:
For questions about the Data Report Card, the GenderSci Lab COVID Project, or to collaborate, contact us at genderscilab@fas.harvard.edu.
Statement of Intellectual Labor:
Boulicault drafted the initial blog post and led the writing process. Danielsen, Shattuck-Heidorn, and Tarrant provided revisions and edits.