Variation in Reporting of the Race and Ethnicity of COVID-19 Cases and Deaths Across US States: April 12, 2020, and November 9, 2020

Auditing the fairness of the US COVID-19 forecast hub's case prediction models

The US COVID-19 Forecast Hub, a repository of COVID-19 forecasts from over 50 independent research groups, is used by the Centers for Disease Control and Prevention (CDC) for their official COVID-19 communications. As such, the Forecast Hub is a critical centralized resource to promote transparent decision making. While the Forecast Hub has provided valuable predictions focused on accuracy, there is an opportunity to evaluate model performance across social determinants such as race and urbanization level that have been known to play a role in the COVID-19 pandemic. In this paper, we carry out a comprehensive fairness analysis of the Forecast Hub model predictions and we show statistically significant diverse predictive performance across social determinants, with minority racial and ethnic groups as well as less urbanized areas often associated with higher prediction errors. We hope this work will encourage COVID-19 modelers and the CDC to report fairness metrics together with accuracy, and to reflect on the potential harms of the models on specific social groups and contexts.

Leveraging Multiple Administrative Data Sources to Reduce Missing Race and Ethnicity Data: A Descriptive Epidemiology Cross-Sectional Study of COVID-19 Case Relative Rates

BACKGROUND

Understanding race and ethnicity (RE) differentials improves health outcomes. However, RE data are consistently missing from electronic laboratory reports, the primary source of COVID-19 case metrics. We addressed the missing RE differentials and compared vaccinated and unvaccinated cases from March 1, 2020, to May 30, 2023, in New York State (NYS), excluding New York City.

METHODS

This descriptive epidemiology cross-sectional study linked the NYS Electronic Clinical Laboratory Reporting System (ECLRS) with NYS Immunization Information System (NYSIIS) to address the missing RE data in the ECLRS system. The primary metric was the COVID-19 case relative risk (RR) for each RE relative to white individuals.

RESULTS

There were 4,212,741 COVID-19 cases with 39% (1,624,818) missing RE data in ECLRS; missing RE data declined to 17% (726,023) after matching with NYSIIS. For those aged 65 years or older (after matching), 42% were missing in 2020, which declined by 17% by 2023. In May 2021, COVID-19 RRs for vaccinated individuals were 1.09 (95% CI 0.90-1.32), 1.11 (95% CI 0.87-1.43), 1.13 (95% CI 0.93-1.39), and 1.89 (95% CI 1.01-3.52), and for unvaccinated individuals were 1.73 (95% CI 1.66-1.82), 0.84 (95% CI 0.78-0.92), 3.10 (95% CI 2.98-3.22), and 3.49 (95% CI 3.05-3.98) respectively for Hispanic, Asian/Pacific Islander, Black people, and American Indian/Alaska Native individuals.

CONCLUSION

Matching case data with vaccine registries reduce missing RE data for COVID-19 cases. Disparity was lower in vaccinated than in unvaccinated individuals indicating that vaccination mitigated RE disparities early in the pandemic. This underscores the value of interoperable systems with automated matching for disparity analyses.

Leveraging Ethnic Backgrounds to Improve Collection of Race, Ethnicity, and Language Data

INTRODUCTION

Healthcare disparities may be exacerbated by upstream incapacity to collect high-quality and accurate race, ethnicity, and language (REaL) data. There are opportunities to remedy these data barriers. We present the Denver Health (DH) REaL initiative, which was implemented in 2021.

METHODS

Denver Health is a large safety net health system. After assessing the state of REaL data at DH, we developed a standard script, implemented training, and adapted our electronic health record to collect this information starting with an individual's ethnic background followed by questions on race, ethnicity, and preferred language. We analyzed the data for completeness after REaL implementation.

RESULTS

A total of 207,490 patients who had at least one in-person registration encounter before and after the DH REaL implementation were included in our analysis. There was a significant decline in missing values for race (7.9%-0.5%, p < .001) and for ethnicity (7.6%-0.3%, p < .001) after implementation. Completely of language data also improved (3%-1.6%, p < .001). A year after our implementation, we knew over 99% of our cohort's self-identified race and ethnicity.

CONCLUSIONS

Our initiative significantly reduced missing data by successfully leveraging ethnic background as the starting point of our REaL data collection.

Racial and Ethnic Disparities in Excess All-Cause Mortality in the First Year of the COVID-19 Pandemic

Research on the COVID-19 pandemic in the United States has consistently found disproportionately high mortality among ethnoracial minorities, but reports differ with respect to the magnitude of mortality disparities and reach different conclusions regarding which groups were most impacted. We suggest that these variations stem from differences in the temporal scope of the mortality data used and difficulties inherent in measuring race and ethnicity. To circumvent these issues, we link Social Security Administration death records for 2010 through 2021 to decennial census and American Community Survey race and ethnicity responses. We use these linked data to estimate excess all-cause mortality for age-, sex-, race-, and ethnicity-specific subgroups and examine ethnoracial variation in excess mortality across states and over the course of the pandemic's first year. Results show that non-Hispanic American Indians and Alaska Natives experienced the highest excess mortality of any ethnoracial group in the first year of the pandemic, followed by Hispanics and non-Hispanic Blacks. Spatiotemporal and age-specific ethnoracial disparities suggest that the socioeconomic determinants driving health disparities prior to the pandemic were amplified and expressed in new ways in the pandemic's first year to disproportionately concentrate excess mortality among racial and ethnic minorities.

Completeness of Race and Ethnicity Reporting in Person-Level COVID-19 Surveillance Data, 50 States, April 2020-December 2021

OBJECTIVES

Black, Indigenous, and People of Color have borne a disproportionate incidence of COVID-19 cases in the United States. However, few studies have documented the completeness of race and ethnicity reporting in national COVID-19 surveillance data. The objective of this study was to describe the completeness of race and ethnicity ascertainment in person-level data received by the Centers for Disease Control and Prevention (CDC) through national COVID-19 case surveillance.

METHODS

We compared COVID-19 cases with "complete" (ie, per Office of Management and Budget 1997 revised criteria) data on race and ethnicity from CDC person-level surveillance data with CDC-reported aggregate counts of COVID-19 from April 5, 2020, through December 1, 2021, in aggregate and by state.

RESULTS

National person-level COVID-19 case surveillance data received by CDC during the study period included 18 881 379 COVID-19 cases with complete ascertainment of race and ethnicity, representing 39.4% of all cases reported to CDC in aggregate (N = 47 898 497). Five states (Georgia, Hawaii, Nebraska, New Jersey, and West Virginia) did not report any COVID-19 person-level cases with multiple racial identities to CDC.

CONCLUSION

Our findings highlight a high degree of missing data on race and ethnicity in national COVID-19 case surveillance, enhancing our understanding of current challenges in using these data to understand the impact of COVID-19 on Black, Indigenous, and People of Color. Streamlining surveillance processes to decrease reporting incidence and align reporting requirements with an Office of Management and Budget-compliant collection of data on race and ethnicity would improve the completeness of data on race and ethnicity for national COVID-19 case surveillance.

Evaluating Demographic Representation in Clinical Trials: Use of the Adaptive Coronavirus Disease 2019 Treatment Trial (ACTT) as a Test Case

Background

Clinical trials initiated during emerging infectious disease outbreaks must quickly enroll participants to identify treatments to reduce morbidity and mortality. This may be at odds with enrolling a representative study population, especially when the population affected is undefined.

Methods

We evaluated the utility of the Centers for Disease Control and Prevention's COVID-19-Associated Hospitalization Surveillance Network (COVID-NET), the COVID-19 Case Surveillance System (CCSS), and 2020 United States (US) Census data to determine demographic representation in the 4 stages of the Adaptive COVID-19 Treatment Trial (ACTT). We compared the cumulative proportion of participants by sex, race, ethnicity, and age enrolled at US ACTT sites, with respective 95% confidence intervals, to the reference data in forest plots.

Results

US ACTT sites enrolled 3509 adults hospitalized with COVID-19. When compared with COVID-NET, ACTT enrolled a similar or higher proportion of Hispanic/Latino and White participants depending on the stage, and a similar proportion of African American participants in all stages. In contrast, ACTT enrolled a higher proportion of these groups when compared with US Census and CCSS. The proportion of participants aged ≥65 years was either similar or lower than COVID-NET and higher than CCSS and the US Census. The proportion of females enrolled in ACTT was lower than the proportion of females in the reference datasets.

Conclusions

Although surveillance data of hospitalized cases may not be available early in an outbreak, they are a better comparator than US Census data and surveillance of all cases, which may not reflect the population affected and at higher risk of severe disease.

The Lack of Race and Ethnicity Data in Australia-A Threat to Achieving Health Equity

Collecting meaningful race and ethnicity data must be part of the national agenda and must be one of its primary objectives in order to achieve public good and support public interests. Yet, Australia does not collect data on race and ethnicity, and prefers the use of collective cultural groups, whose information is not consistently collected and reported at all levels of government and service delivery. This paper examines the current discrepancies in race and ethnicity data collection in Australia. The paper begins with examining the current practices related to collecting race and ethnicity data and then moves on to examine the various implications and public health significance of not collecting data on race and ethnicity in Australia. The evidence suggests that (1) race and ethnicity data matter, are imperative to ensuring proper advocacy and to reducing inequities in health and social determinant factors; (2) that White privilege is constructed as realized or unrealized personal and systemic racism; and (3) the use of non-committal collective terminologies makes visible minorities invisible, leads to the distorted allocation of governmental support, and legitimises and institutionalises racism and othering, hence perpetuating exclusion and the risk of victimisation. There is an urgent need for the collection of customized, culturally competent racial and ethnicity data that can be consistently integrated into all policy interventions, service delivery and research funding across all levels of governance in Australia. Reducing and eliminating racial and ethnic disparities is not only an ethical, social, and economic imperative, but must also be a critical item on the national agenda. Bridging the racial and ethnic disparities will require concerted whole-of-government efforts to collect consistent and reliable data that depict racial and ethnic characteristics beyond collective cultural groupings.

Assessment of Population-Level Disadvantage Indices to Inform Equitable Health Policy

Policy Points The rapid uptake of disadvantage indices during the pandemic highlights investment in implementing tools that address health equity to inform policy. Existing indices differ in their design, including data elements, social determinants of health domains, and geographic unit of analysis. These differences can lead to stark discrepancies in place-based social risk scores depending on the index utilized. Disadvantage indices are useful tools for identifying geographic patterns of social risk; however, indiscriminate use of indices can have varied policy implications and unintentionally worsen equity. Implementers should consider which indices are suitable for specific communities, objectives, potential interventions, and outcomes of interest.

CONTEXT

There has been unprecedented uptake of disadvantage indices such as the Centers for Disease Control and Prevention Social Vulnerability Index (SVI) to identify place-based patterns of social risk and guide equitable health policy during the COVID-19 pandemic. However, limited evidence around data elements, interoperability, and implementation leaves unanswered questions regarding the utility of indices to prioritize health equity.

METHODS

We identified disadvantage indices that were (a) used three or more times from 2018 to 2021, (b) designed using national-level data, and (c) available at the census-tract or block-group level. We used a network visualization to compare social determinants of health (SDOH) domains across indices. We then used geospatial analyses to compare disadvantage profiles across indices and geographic areas.

FINDINGS

We identified 14 indices. All incorporated data from public sources, with half using only American Community Survey data (n = 7) and the other half combining multiple sources (n = 7). Indices differed in geographic granularity, with county level (n = 5) and census-tract level (n = 5) being the most common. Most states used the SVI during the pandemic. The SVI, the Area Deprivation Index (ADI), the COVID-19 Community Vulnerability Index (CCVI), and the Child Opportunity Index (COI) met criteria for further analysis. Selected indices shared five indicators (income, poverty, English proficiency, no high school diploma, unemployment) but varied in other metrics and construction method. While mapping of social risk scores in Durham County, North Carolina; Cook County, Illinois; and Orleans Parish, Louisiana, showed differing patterns within the same locations depending on choice of disadvantage index, risk scores across indices showed moderate to high correlation (rs 0.7-1). However, spatial autocorrelation analyses revealed clustering, with discrepant distributions of social risk scores between different indices.

CONCLUSIONS

Existing disadvantage indices use varied metrics to represent place-based social risk. Within the same geographic area, different indices can provide differences in social risk values and interpretations, potentially leading to varied public health or policy responses.

Missing Race and Ethnicity Data among COVID-19 Cases in Massachusetts

Infectious disease surveillance frequently lacks complete information on race and ethnicity, making it difficult to identify health inequities. Greater awareness of this issue has occurred due to the COVID-19 pandemic, during which inequities in cases, hospitalizations, and deaths were reported but with evidence of substantial missing demographic details. Although the problem of missing race and ethnicity data in COVID-19 cases has been well documented, neither its spatiotemporal variation nor its particular drivers have been characterized. Using individual-level data on confirmed COVID-19 cases in Massachusetts from March 2020 to February 2021, we show how missing race and ethnicity data: (1) varied over time, appearing to increase sharply during two different periods of rapid case growth; (2) differed substantially between towns, indicating a nonrandom distribution; and (3) was associated significantly with several individual- and town-level characteristics in a mixed-effects regression model, suggesting a combination of personal and infrastructural drivers of missing data that persisted despite state and federal data-collection mandates. We discuss how a variety of factors may contribute to persistent missing data but could potentially be mitigated in future contexts.

Racial and Ethnic Disparities in the Observed COVID-19 Case Fatality Rate Among the U.S. Population

Purpose

During the initial 12 months of the pandemic, racial and ethnic disparities in COVID-19 death rates received considerable attention but it has been unclear whether disparities in death rates were due to disparities in case fatality rates (CFRs), incidence rates or both. We examined differences in observed COVID-19 CFRs between U.S. White, Black/African American, and Latinx individuals during this period.

Methods

Using data from the COVID Tracking Project and the Centers for Disease Control and Prevention COVID-19 Case Surveillance Public Use dataset, we calculated CFR ratios comparing Black and Latinx to White individuals, both overall and separately by age group. We also used a model of monthly COVID-19 deaths to estimate CFR ratios, adjusting for age, gender, and differences across states and time.

Results

Overall Black and Latinx individuals had lower CFRs than their White counterparts. However, when adjusting for age, Black and Latinx had higher CFRs than White individuals among those younger than 65. CFRs varied substantially across states and time.

Conclusions

Disparities in COVID-19 case fatality among U.S. Black and Latinx individuals under age 65 were evident during the first year of the pandemic. Understanding racial and ethnic differences in COVID-19 CFRs is challenging due to limitations in available data.

Alternative Methods for Grouping Race and Ethnicity to Monitor COVID-19 Outcomes and Vaccination Coverage

Population-based analyses of COVID-19 data, by race and ethnicity can identify and monitor disparities in COVID-19 outcomes and vaccination coverage. CDC recommends that information about race and ethnicity be collected to identify disparities and ensure equitable access to protective measures such as vaccines; however, this information is often missing in COVID-19 data reported to CDC. Baseline data collection requirements of the Office of Management and Budget's Standards for the Classification of Federal Data on Race and Ethnicity (Statistical Policy Directive No. 15) include two ethnicity categories and a minimum of five race categories (1). Using available COVID-19 case and vaccination data, CDC compared the current method for grouping persons by race and ethnicity, which prioritizes ethnicity (in alignment with the policy directive), with two alternative methods (methods A and B) that used race information when ethnicity information was missing. Method A assumed non-Hispanic ethnicity when ethnicity data were unknown or missing and used the same population groupings (denominators) for rate calculations as the current method (Hispanic persons for the Hispanic group and race category and non-Hispanic persons for the different racial groups). Method B grouped persons into ethnicity and race categories that are not mutually exclusive, unlike the current method and method A. Denominators for rate calculations using method B were Hispanic persons for the Hispanic group and persons of Hispanic or non-Hispanic ethnicity for the different racial groups. Compared with the current method, the alternative methods resulted in higher counts of COVID-19 cases and fully vaccinated persons across race categories (American Indian or Alaska Native [AI/AN], Asian, Black or African American [Black], Native Hawaiian or Other Pacific Islander [NH/PI], and White persons). When method B was used, the largest relative increase in cases (58.5%) was among AI/AN persons and the largest relative increase in the number of those fully vaccinated persons was among NH/PI persons (51.6%). Compared with the current method, method A resulted in higher cumulative incidence and vaccination coverage rates for the five racial groups. Method B resulted in decreasing cumulative incidence rates for two groups (AI/AN and NH/PI persons) and decreasing cumulative vaccination coverage rates for AI/AN persons. The rate ratio for having a case of COVID-19 by racial and ethnic group compared with that for White persons varied by method but was <1 for Asian persons and >1 for other groups across all three methods. The likelihood of being fully vaccinated was highest among NH/PI persons across all three methods. This analysis demonstrates that alternative methods for analyzing race and ethnicity data when data are incomplete can lead to different conclusions about disparities. These methods have limitations, however, and warrant further examination of potential bias and consultation with experts to identify additional methods for analyzing and tracking disparities when race and ethnicity data are incomplete.