We hope our study can teach us how to detect early signs of COVID-19 and identify who is most likely to become sick. The tools most helpful to us are mobile and wearable devices, like Fitbits or Apple Watches.
For more information, please visit our study’s website at covidentify.org
Biases in Data
The information we get from people’s wearable devices, like their step counts or heart rate, is important data for our study. We use this data to make models and digital biomarkers that might help us find out who is showing early signs of COVID-19. These models can be misleading if the data is not carefully considered.1
It is crucial that the people who provide us with this data are a reflection of our communities. We need representation from people of all ages, genders, and racial/ethnic backgrounds.
Comparing Our Study Population to the U.S. Demographic
In this blog post, we compare our study population to the U.S. population. The figures below only represent our study participants who are located in the United States, and is a snapshot of our study participants as of July 15th, 2020. We also understand the difficulty of categorizing people into specific groups. It might not be easy to clearly define race vs. ethnicity and sex vs. gender. In this blog, the demographic categorizations are based on government data and suggestions from numerous Duke Diversity groups.
Race and Ethnicity
At this time, the number of minority participants in our study does not reflect the general U.S. population. We want the models we create from our study to serve all people and welcome support in improving the diversity of our participants.
Figure 1 below shows the breakdown by race and ethnicity of all CovIdentify participants who live in the U.S.* As you can see, most people in our study are White, non-Hispanic. In the U.S., only 60.1% of Americans are White, non-Hispanic, which is a smaller percentage than what we have in our study.2
People who are Black/African American or Hispanic/Latino have lower representation in the study compared to the general population. Blacks/African Americans make up 13.4% of the U.S. population2, but only 3.6% of our study. Hispanics/Latinos make up 18.5% of the U.S. population2, but only 4% of our study.
We are working with community partners to improve enrollment of minorities into our study. We are also applying for funds to buy wearable devices. People who join the study might be eligible to receive a free wearable device if they don’t already have one.
*We understand that different races and ethnicities (Hispanic or non-Hispanic) may not always be put in the same graph. In this figure, we are following CDC data and placing Hispanics and Latinos as one category among other races.
Sex Assigned at Birth and Gender Identity
Table 1 below shows the breakdown by Sex Assigned at Birth and Gender Identity of our study participants. The Sex Assigned at Birth categories are Female, Male, and Intersex. Gender Identity is how one identifies their gender, regardless to the sex one was assigned at birth. The gender identity categories are Woman/ Cisgender female, Man/ Cisgender male, Non-Binary or Third Gender, and a choice to Self-Describe (Because there are many other terms to describe gender identity, participants had the option to self-describe).
Our study has a much higher percentage of female participants compared to male participants in terms of biological sex. There are similar differences for gender identity. Females make up 50.8% of the U.S. population, but more than 67% of our study. We are working to enroll more males into the study for better balance.Table 1: Breakdown of gender differences in our study and nationally
|Sex Assigned at Birth in Covidentify participants as of July 15th||Gender Identity in Covidentify participants as of July 15th||Sex Assigned at Birth in the U.S.2|
|Female (Woman)||67.75 %||67.53 %||50.8 %|
|Male (Man)||32.20 %||31.97 %||49.2 %**|
|Intersex||0.05 %||(Sex assigned at Birth but not Gender Identity)||(Not recorded)|
|Non Binary/Third Gender||(Gender identity but not Sex Assigned at birth)||0.35 %||(Gender identity but not Sex Assigned at birth)|
|Prefer to self describe||(Gender identity but not Sex Assigned at birth)||0.16 %||(Gender identity but not Sex Assigned at birth)|
**The U.S. Census Bureau only reports the number of males and females in the country. It does not provide data for intersex.
The bar graph in Figure 2 shows participants by their age group and sex. For the time being, our study only enrolls people who are 18 years or older. Most participants are over 40 years old and about 25% are between the ages of 50 and 59. We want to enroll more people in their 20s and 30s. This will help us understand how COVID-19 affects different age groups
Health Disparities in the U.S.
COVID-19 has hit the Black/African American and Hispanic/Latino communities in America the hardest.3 4 Both communities have infection and death rates that are larger than their share of the U.S. population.
There might be many reasons for this disparity. People in these communities often live in neighborhoods that are more crowded. These community members often work in essential jobs that make social distancing difficult or impossible. And health inequities toward minority groups in America remain a persistent problem.4
Figure 3 breaks down COVID-19 positive tests and deaths in the U.S. by race and ethnicity (data from Center of Disease Control and Prevention).5 6 The figure further compares these distributions to our study participants and national demographics. Figure 4 highlights three groups in Figure 3: Black/African American, Hispanic/Latino and White/Caucasian groups. Black/African American and Hispanic/Latino people together are less than a third of the U.S. population (green bars). But the two communities make up more than half of all COVID-19 cases (blue bars) and more than 40% of deaths (orange bars).
Our study participants (red bars) do not adequately represent minorities who are most at risk for COVID-19. To improve health equity and better prevent infection, we are actively reaching out to minority groups to participate in CovIdentify.
Improving the recruitment of diverse study participants from underserved communities.
We are focusing on increasing the diversity of our study population. We need to enroll and engage with more people of color and more young people. The more diverse our study, the more likely we are to make discoveries that benefit everyone.
To reach this goal, we are working closely with the Duke Center for Translational Sciences Institute Community-Engaged Research Initiative and Recruitment Innovation Center, the Duke Latinx Advocacy Team & Interdisciplinary Network for COVID-19, the African Methodist Episcopal Zion Church, and the Duke Mobile App Gateway.
We have translated CovIdentify into seven languages and we are aiming advertising and outreach toward local communities of color. You don’t have to own a wearable device to join our study, but we are distributing them for free to people from the communities hardest hit by COVID-19.
Are you a CovIdentify participant interested in leading or engaging in a Community Advisory Board? Please email us at firstname.lastname@example.org
We also invite interested groups to reach out to us with more ideas on improving our study and community engagement.
To help our team learn more about how we can address this and future pandemics, we encourage you to share this study widely and consider signing up to participate on our website covidentify.org.
This work is not peer-reviewed.
1. Panch, Trishan, et al. “Artificial Intelligence and Algorithmic Bias: Implications for Health Systems.” Journal of Global Health, vol. 9, no. 2, 2019, doi:10.7189/jogh.09.020318.
2. U.S. Census Bureau QuickFacts: United States.” Census Bureau QuickFacts, www.census.gov/quickfacts/fact/dashboard/US/RHI125219
3. Anyane‐Yeboa, A., et al. “Racial Disparities in COVID‐19 Deaths Reveal Harsh Truths about Structural Inequality in America.” Journal of Internal Medicine, 2020, doi:10.1111/joim.13117.
4. Shah, Monica, et al. “COVID-19 and Racial Disparities.” Journal of the American Academy of Dermatology, vol. 83, no. 1, 2020, doi:10.1016/j.jaad.2020.04.046.
5. “COVID-19 Case Surveillance Public Use Data.” Centers for Disease Control and Prevention, Centers for Disease Control and Prevention, data.cdc.gov/Case-Surveillance/COVID-19-Case-Surveillance-Public-Use-Data/vbim-akqf/data.
6. “Deaths Involving Coronavirus Disease 2019 (COVID-19) by Race and Hispanic Origin Group and Age, by State.” Centers for Disease Control and Prevention, Centers for Disease Control and Prevention, data.cdc.gov/NCHS/Deaths-involving-coronavirus-disease-2019-COVID-19/ks3g-spdg.