Ethnoracial Disparities Research Articles

Ethical Redress of Racial Inequities in AI: Lessons from Decoupling Machine Learning from Optimization in Medical Appointment Scheduling

An Artificial Intelligence algorithm trained on data that reflect racial biases may yield racially biased outputs, even if the algorithm on its own is unbiased. For example, algorithms used to schedule medical appointments in the USA predict that Black patients are at a higher risk of no-show than non-Black patients, though technically accurate given existing data that prediction results in Black patients being overwhelmingly scheduled in appointment slots that cause longer wait times than non-Black patients. This perpetuates racial inequity, in this case lesser access to medical care. This gives rise to one type of Accuracy-Fairness trade-off: preserve the efficiency offered by using AI to schedule appointments or discard that efficiency in order to avoid perpetuating ethno-racial disparities. Similar trade-offs arise in a range of AI applications including others in medicine, as well as in education, judicial systems, and public security, among others. This article presents a framework for addressing such trade-offs where Machine Learning and Optimization components of the algorithm are decoupled. Applied to medical appointment scheduling, our framework articulates four approaches intervening in different ways on different components of the algorithm. Each yields specific results, in one case preserving accuracy comparable to the current state-of-the-art while eliminating the disparity.

Ethnoracial Disparities in Posttraumatic Stress Disorder Symptoms during the COVID-19 Pandemic: A Brief Report.

Despite the well-identified vulnerability of older adults during the COVID-19 pandemic, it is unclear about their experiences with COVID-related posttraumatic stress disorder symptomology (COVID-PTSD). This study examined ethnoracial disparities in the level of, and factors associated with, COVID-PTSD using a national data set, including 1926 Whites and 488 ethnoracial minorities. Results showed that ethnoracial minorities reported a greater COVID-PTSD than Whites. COVID-related distress was the common risk factor of COVID-PTSD for the both groups. Being a female and greater social support were associated with COVID-PTSD only for Whites, whereas higher education, greater IADL and fewer ADL limitations were associated with COVID-PTSD for ethnoracial minorities. The findings provided preliminary, but generalizable understanding of ethnoracial disparities in COVID-PTSD, among the Medicare beneficiaries aged ≥65.

The Cumulative Prevalence of Congregate Care Placement for U.S. Children by Race/Ethnicity, 2019.

Congregate care placement is among the most consequential forms of foster care placement that youth can experience, as it means a removal from both the family of origin and a family setting more broadly. Unfortunately, little research has estimated how common this intervention is. In this article, we use data from the Adoption and Foster Care Analysis Reporting System (AFCARS) and synthetic cohort life tables to show what proportion of children ever placed in foster care will ever be placed in congregate care, what proportion of children in the entire population will ever be placed in congregate care, and how these proportions vary by state of residence and race/ethnicity. Our results support four main conclusions. First, roughly 15% of all children ever placed in foster care will experience congregate care placement. Second, among children who will ever be placed in foster care, the risk of congregate care placement peaks at age 16. Third, congregate care placement is highly stratified by race/ethnicity. Finally, there are vast geographic differences in both congregate care placement and ethno-racial disparities therein. Taken together, these findings enhance our understanding of the demography of the child welfare system with implications for research, policy, and practice.

Racism and health: the need for urgent action

Racism and health: the need for urgent action

Disparities in Early-Stage Breast Cancer and Survival-Letter.

We commend Primm and colleagues for performing an analysis of ethnoracial disparities in outcomes in women with early-stage breast cancer (1). As the authors note, aggregating race/ethnicity into broad categories may mask important differences between different ethnoracial subgroups. This study was more comprehensive than prior studies in that individuals who identified as Pacific Islander had their own category. In addition, the authors disaggregated those identifying as Asian into four distinct subgroups.With respect to ethnicity, the Hispanic population is an extremely diverse ethnoracial group that represents the fastest growing demographic in the United States. Therefore, while we completely agree with Primm and colleagues about the disaggregation of Asian people we would argue those of Hispanic ethnicity also warrant the same level of ethnoracial granularity. Studies in patients with breast cancer have demonstrated that disaggregating Hispanic ethnicity by race captures ethnoracial (Hispanic White, Hispanic Black, etc.) differences in stage, treatment, and survival not elucidated by aggregating Hispanic individuals into one group (2, 3). Consequently, we encourage the authors to conduct further analysis examining the study endpoints with the disaggregation of the Hispanic population by race. We anticipate this approach will help delineate the influence of race and ethnicity on the trend towards worse survival outcomes seen in Hispanic people with early-stage disease and across subtypes (except triple-negative disease). With respect to triple-negative disease, it would be particularly interesting to see if there are survival differences in Black people based on Hispanic ethnicity. Paradoxically, Hispanic individuals with stage IIIB-IV disease had improved survival relative to non-Hispanic White individuals, but again, it is critical to understand if this disparity is driven by ethnoracial differences in Hispanic patients. Of note, prior studies using SEER have successfully disaggregated Hispanic ethnicity by race (4).In addition to disaggregating ethnicity by race, categorizing Hispanic people by country of origin is also important when possible. SEER does collect birthplace information, but there is a high level of missing data and the veracity of the datapoint is in question (5). While acknowledging that Hispanic people were not differentiated further as a limitation, the authors do not provide contextual information for readers as to why this limitation is important. For example, in patients with pancreatic cancer, investigators have shown that among Hispanic individuals, Dominicans and South/Central Americans had the best survival rates (6).In summary, we agree that contemporary analyses of race and ethnicity needs improvement with as many distinct categories as possible to reflect the diversity of the patient population. We believe that Hispanic patients should be categorized by both race and ethnicity. When feasible, investigators should also report country origin for foreign born individuals of any race and/or ethnicity. Granular ethnoracial categories will help improve our understanding of the independent and synergistic effects of ethnicity and race on cancer outcomes.See the Response, p. 1868No disclosures were reported.

Did ethno-racial disparities in access to transcatheter aortic valve replacement change over time?

ObjectiveIn this study we sought to evaluate whether disparate use of transcatheter aortic valve replacement (TAVR) among non-White patients has decreased over time, and if unequal access to TAVR is driven by unequal access to high-volume hospitals. MethodsFrom 2013 to 2017, we used the State Inpatient Database across 8 states (Ariz, Colo, Fla, Md, NC, NM, Nev, Wash) to identify 51,232 Medicare beneficiaries who underwent TAVR versus surgical aortic valve replacement. Hospitals were categorized as low- (<50 per year), medium- (50-100 per year), or high-volume (>100 per year) according to total valve procedures (TAVR + surgical aortic valve replacement). Multivariable logistic regression models with interactions were performed to determine the effect of race, time, and hospital volume on the utilization of TAVR. ResultsNon-White patients were less likely to receive TAVR than White patients (odds ratio [OR], 0.77; 95% CI, 0.71-0.83). However, utilization of TAVR increased over time (OR, 1.73; 95% CI, 1.73-1.80) for the total population, with non-White patients’ TAVR use growing faster than for White patients (OR, 1.06; 95% CI, 1.00-1.12), time × race interaction, P = .034. Further, an adjusted volume-stratified time trend analysis showed that utilization of TAVR at high volume hospitals increased faster for non-White patients versus White patients by 8.6% per year (OR, 1.09; 95% CI, 1.01-1.16) whereas use at low- and medium-volume hospitals did not contribute to any decreasing utilization gap. ConclusionsThis analysis shows initial low rates of TAVR utilization among non-White patients followed by accelerated use over time, relative to White patients. This narrowing gap was driven by increased TAVR utilization by non-White patients at high-volume hospitals.

Ethnoracial disparities in commute duration and mode among Canadian-born millennials

In the past few decades, ethnoracial diversity has substantially grown in Canada. Both occupational and residential segregation of minorities are well-documented phenomena. Together, they can lead to a “segmented assimilation” of minorities who are born in the host country, whereby they converge to the patterns observed in their foreign-born co-ethnics, rather than the majority population. An important dimension of segmented assimilation regards commute to work, its duration, distance, and mode. Using the Canadian Census of 2016, the present study finds large disparities by race and ethnicity among millennials born in Canada, in commute duration. The commute duration gap is especially large when black millennials are compared with their white counterparts. Further analysis shows that these gaps are driven by the greater reliance of ethnoracial minorities on public transit for their journey to work, compared with whites. Policy implications are discussed.

An Intersectional Approach to Ethnoracial Disparities in Pathways to Care Among Individuals With Psychosis in Coordinated Specialty Care

Intersecting factors of social position including ethnoracial background may provide meaningful ways to understand disparities in pathways to care for people with a first episode of psychosis. To examine differences in pathways to care by ethnoracial groups and by empirically derived clusters combining multiple factors of social and clinical context in an ethnoracially diverse multisite early-intervention service program for first-episode psychosis. This cohort study used data collected on individuals with recent-onset psychosis (<2 years) by clinicians with standardized forms from October 2013 to January 2020 from a network of 21 coordinated specialty care (CSC) programs in New York State providing recovery-oriented, evidence-based psychosocial interventions and medications to young people experiencing early psychosis. Ethnoracial group and other factors of social position (eg, insurance status, living situation, English fluency, geographic region) intersecting with first-contact experiences (ie, type of first service, referral source, and symptoms at referral). Outcome measures were time from onset to first contact, first contact to CSC, and onset to CSC. The total study sample consists of 1726 individuals aged 16 to 30 years and included 452 women (26%), 1263 men (73%), and 11 (<1%) with another gender enrolled in the network of CSC programs. The total sample consisted of 153 Asian (9%), 599 Black (35%), 454 Latinx (26%), and 417 White individuals (24%). White individuals had a significantly shorter time from onset to first contact (median [IQR], 17 [0-80] days) than Asian (median [IQR], 34 [7-94] days) and Black (median [IQR], 30 [1-108] days) individuals but had the longest period from first contact to CSC (median [IQR], 102.5 [45-258] days). Five distinct clusters of individuals emerged that cut across ethnoracial groups. The more disadvantaged clusters in terms of both social position and first-contact experiences had the longest time from onset to first contact, which were longer than for any single ethnoracial group. In this cohort study of individuals with recent-onset psychosis, time-to-treatment outcomes differed by ethnoracial group and by empirically derived clusters combining multiple factors of social and clinical context. The examination of disparities in durations to treatment through an intersectional, ethnoracial lens may improve understanding of the inequities resulting from the various intersecting factors that may compound delays in treatment initiation.

Ethno-Racial Disparities in Adverse Experiences at Mortgage Closing: Evidence From the National Survey of Mortgage Originations

The mortgage industry is central to ethno-racial stratification in access to homeownership. Obtaining a successful mortgage origination requires many steps, points of verification, and interactions between clients, realtors, and mortgage brokers. Ample research has documented ethno-racial inequality in mortgage loan outcomes and during the early stages of the homebuying process, such as minority steering and quality of service. However, it is unclear whether and to what extent ethno-racial inequality is exhibited during the final stage of the homebuying process. Therefore, this article draws on the National Survey of Mortgage Originations, an annual representative sample of new mortgage loan originations, from 2013 to 2017, to assess ethno-racial disparities in adverse loan experiences during the mortgage closing stage. I show that black mortgage holders are more likely to have an adverse experience (a negative experience, an acute lending practice, or an unexpected financial cost) during the closing, compared to Non-Hispanic whites and Asians. Latino and black mortgage holders face similar disparities when it comes to unexpected financial costs (the most severe adverse experience) during the closing appointment. These ethno-racial disparities during the final stage of the mortgage process are extremely troubling as they increase the administrative and economic costs for minority homebuyers. Implications for ethno-racial wealth inequality are discussed.

Ethnoracial Disparities in SARS-CoV-2 Seroprevalence in a Large Cohort of Individuals in Central North Carolina from April to December 2020.

ABSTRACTSevere acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has caused millions of deaths around the world within the past 2 years. Transmission within the United States has been heterogeneously distributed by geography and social factors with little data from North Carolina. Here, we describe results from a weekly cross-sectional study of 12,471 unique hospital remnant samples from 19 April to 26 December 2020 collected by four clinical sites within the University of North Carolina Health system, with a majority of samples from urban, outpatient populations in central North Carolina. We employed a Bayesian inference model to calculate SARS-CoV-2 spike protein immunoglobulin prevalence estimates and conditional odds ratios for seropositivity. Furthermore, we analyzed a subset of these seropositive samples for neutralizing antibodies. We observed an increase in seroprevalence from 2.9 (95% confidence interval [CI], 1.8 to 4.5) to 12.8 (95% CI, 10.6 to 15.2) over the course of the study. Latinx individuals had the highest odds ratio of SARS-CoV-2 exposure at 6.56 (95% CI, 4.66 to 9.44). Our findings aid in quantifying the degree of asymmetric SARS-CoV-2 exposure by ethnoracial grouping. We also find that 49% of a subset of seropositive individuals had detectable neutralizing antibodies, which was skewed toward those with recent respiratory infection symptoms.IMPORTANCE PCR-confirmed SARS-CoV-2 cases underestimate true prevalence. Few robust community-level SARS-CoV-2 ethnoracial and overall prevalence estimates have been published for North Carolina in 2020. Mortality has been concentrated among ethnoracial minorities and may result from a high likelihood of SARS-CoV-2 exposure, which we observe was particularly high among Latinx individuals in North Carolina. Additionally, neutralizing antibody titers are a known correlate of protection. Our observation that development of SARS-CoV-2 neutralizing antibodies may be inconsistent and dependent on severity of symptoms makes vaccination a high priority despite prior exposure.

Ethno-racial stratification in the mortgage market: The role of co-applicants

Unequal access to homeownership is central to ethno-racial stratification. Ample research demonstrates large ethno-racial disparities that exist in access and outcomes throughout the mortgage process at both the individual and neighborhood levels. The underlying assumption in most of these studies is that the couples applying for a mortgage are ethno-racially homogenous. However, the ethno-racial stratification structure is unclear when examining interracial couples in the mortgage market. This paper draws on annual data from the Home Mortgage Disclosure Act (HMDA) from 2010 to 2017 to assess variation in ethno-racial disparities in loan outcomes associated with different ethno-racial couplings. I show that ethno-racial disparities in loan outcomes vary tremendously when factoring the ethno-racial identity of the co-applicant. Interracial couples involving a black or Latino co-applicant are more likely to experience a high-cost loan or be denied a mortgage than mono-racial white couples. The results for Asian co-applicants vary, depending on the adverse loan outcome. When comparing interracial couples to mono-racial couples, the observed lending pattern provides evidence of a tri-racial hierarchy in the mortgage market.

Development of a Practice-based Community Outreach Intervention to Prevent Inequities in COVID-19 Vaccinations.

Despite disproportionately higher rates of morbidity and mortality from COVID-19 among Black and Hispanic adults in the United States, ethnoracial disparities in vaccination rates emerged rapidly. The objective of this quality improvement study was to rapidly develop and implement an equity-focused community outreach intervention that facilitated COVID-19 vaccine appointments. Using the Plan-Do-Study-Act model, this multipronged, primary care-based outreach intervention developed call/recall systems that addressed vaccine hesitancy and facilitated real-time vaccine scheduling. Through 5058 calls to 2794 patients, 1519 patients were successfully reached. Of the 750 patients eligible for vaccine scheduling, 129 (17.2%) had a vaccine appointment scheduled by the caller and 72 (9.6%) indicated a plan to self-schedule. Low confidence in the vaccine was the most cited reason for declining assistance with a vaccine appointment. Primary care practices may wish to consider introducing similar outreach interventions in the future to address ethnoracial inequities in vaccination distribution.

Ethnoracial disparities in intergenerational coresidence among Canadian millennial

Using the Canadian Census of 2016 and restricting the sample to the Canadian-born millennials, this paper is concerned with ethnoracial disparities in adult children's coresidence with their parents. Substantial disparities are found in the prevalence of intergenerational coresidence by race and ethnic origin. Rather surprisingly, Canadian-born millennials of Chinese, Japanese and Korean descent are found the least likely to coreside with their parents than their counterparts in other ethnoracial groups. There is little gender difference in this regard. Further exploration shows that millennials of Chinese, Japanese and Korean descent are much more likely than others to be geographically mobile across Canada, and this greater mobility drives the gaps. In other words, the lower intergenerational coresidence among Canadian millennials of Chinese, Japanese and Korean descent also implies a lower intergenerational proximity among these groups. Future implications of the findings for local labour and housing markets, as well as the healthcare system, are discussed in light of the Confucian concept of filial piety.

Measuring America’s Affordability Problem: Comparing Alternative Measurements of Affordable Housing

Significant scholarly and policy debate has focused on the measurement of affordable housing, with emphasis on what is an appropriate threshold of affordability. However, this threshold is only one component of affordable housing measurement, with accurate and substantively appropriate measurements of income and households also being needed. In this study, I produce a series of estimates of affordable housing among low-income households in the United States under unique combinations of income, providers of income within the household, and thresholds of affordability. I find that these alternative measures yield a broad range of estimates ranging from a majority of households (69.8%) to a low of 20.2%. When examining how individual criteria affect estimates, I find that focusing on wage income alone and using residual income both drastically influence estimates. Ethnoracial disparities are also affected, with alternative measurements often muting—but never completely explaining—disparities between White and non-White households.

Nativity differentials in first births in the United States: Patterns by race and ethnicity.

While recent decades have seen gradual convergence in ethno-racial disparities in completed fertility in the United States, differences in the age pattern of first births remain. The role of nativity has not been fully understood. This paper examines how first births vary by nativity, and how this variation contributes to more significant racial and ethnic differentials. Using data from the National Survey of Family Growth (1997-2017), we jointly estimate the correlates of the timing of first births and childlessness. We assess differences between immigrants and US-born and child-migrant women across ethno-racial groups. The unique first-birth patterns among foreign-born women have a notable impact on Hispanics, reducing differences from Whites in the average age at first birth and contributing to more significant differentials in childlessness. The impact of immigrant women on White and Black first births is more modest in scope. Our work shows the importance of nativity for ethnic/racial disparities in the timing and quantum of fertility in the United States. We demonstrate how the migrant population is more determinant for Hispanic fertility patterns than for Black or White. We conclude by elaborating on the implications of these results for future research as the immigrant population in the United States becomes ethnically and racially more diverse.

Ethnoracial disparities in cognition are associated with multiple socioeconomic status-stress pathways

Systemic racism can have broad impacts on health in ethnoracial minorities. One way is by suppressing socioeconomic status (SES) levels through barriers to achieve higher income, wealth, and educational attainment. Additionally, the weathering hypothesis proposes that the various stressful adversities faced by ethnoracial minorities lead to greater wear and tear on the body, known as allostatic load. In the present study, we extend these ideas to cognitive health in a tri-ethnic sample of young adults—when cognition and brain health is arguably at their peak. Specifically, we tested competing mediation models that might shed light on how two key factors caused by systemic racism—SES and perceived stress—intersect to explain ethnoracial disparities in cognition. We found evidence for partial mediation via a pathway from SES to stress on episodic memory, working memory capacity, and executive function in Black Americans relative to non-Hispanic White Americans. Additionally, we found that stress partially mediated the ethnoracial disparities in working memory updating for lower SES Black and Hispanic Americans relative to non-Hispanic White Americans, showing that higher SES can sometimes reduce the negative effects stress has on these disparities in some cognitive domains. Overall, these findings suggest that multiple pathways exist in which lower SES creates a stressful environment to impact ethnoracial disparities cognition. These pathways differ depending on the specific ethnoracial category and cognitive domain. The present results may offer insight into strategies to help mitigate the late-life risk for neurocognitive disorders in ethnoracial minorities.

Hispanic Ethnicity and Breast Cancer: Disaggregating Surgical Management and Mortality by Race.

Breast cancer is the leading cause of cancer death among Hispanic women. Unfortunately, few studies disaggregate Hispanic patients by race to understand its implications on treatment and clinical outcomes such as mortality. The aim of this study is to examine surgical management and overall mortality among different subgroups of women who self-identify as Hispanic. Hispanic female patients, ages 18-90, stages I-III, diagnosed with breast cancer between 2010 and 2015 from the National Cancer Data Base were identified. The study cohort was divided into three ethnoracial categories: (1) Hispanic White (HW), 2) Hispanic Black (HB), and 3) Hispanic Other (HO). Descriptive statistics and multivariate models were constructed to determine the relationship between sociodemographic factors, clinical variables, surgical management, and mortality when disaggregated by race. There were 56,675 Hispanic women who met the study criteria. Most where HW (n=50,599, 89.3%) and the rest were HB (n=1,334, 2.4%) and HO (n=4,742, 8.3%). There was no difference between the three groups on receipt of breast conservation therapy (P=0.12). HB (48.5%) and HO (46.6%) women were more likely to undergo reconstruction than those who identified as HW (38.7%) (P<0.001). Additionally, HB (38.3%) women were more likely to undergo tissue-based reconstruction than HW (29.0%) and HO women (30%) (P=0.0008). There was no difference between the groups in the utilization of contralateral prophylactic mastectomy (CPM) (P=0.078). On multivariable analysis, there was no difference in mortality between HB and HW patients (HR 1.18, 95%CI 0.92-1.51; Ref HW). However, HO women had a 24% relative risk reduction in mortality (HR 0.76, 95% CI 0.63-0.92; HW ref). Findings from this study suggest there are ethnoracial disparities in reconstruction utilization and mortality among Hispanic women. Future studies should examine how culture, language, healthcare access, and patient preferences contribute to these disparities.

Ethno-racial disparities in COVID-19 morbidity and mortality

The COVID-19 pandemic, by the nature of this crisis and its primarily sanitary, medical and biological management, offers an exceptional opportunity for observations on the scientific debates surrounding the notion of “race”, used and conceptualised at times as a biological reality, at others as a social construct. Based on a synthetic reading of a variety of scientific productions –in medicine, public health, epidemiology, and social sciences– discussing ethno-racial disparities in COVID-19 related morbidity and mortality, this article shows that these two models of interpretation –biological/genetic and socio-economic/socio-demographic– do not stand in opposition to each other: the relationships between race and identified risk factors in COVID-19 related morbidity and mortality, as recognised by researchers are often conjointly conditioned by both biological mechanisms and socio-economic factors. Furthermore, the intersectional approach –articulating various social relations of race, class, gender, age, disability, etc.– proves to be particularly fruitful in grasping the ethno-racial disparities in COVID-19 related morbidity and mortality.

Ethno-racial disparities in COVID-19 morbidity and mortality

The COVID-19 pandemic, by the nature of this crisis and its primarily sanitary, medical and biological management, offers an exceptional opportunity for observations on the scientific debates surrounding the notion of “race”, used and conceptualised at times as a biological reality, at others as a social construct. Based on a synthetic reading of a variety of scientific productions –in medicine, public health, epidemiology, and social sciences– discussing ethno-racial disparities in COVID-19 related morbidity and mortality, this article shows that these two models of interpretation –biological/genetic and socio-economic/socio-demographic– do not stand in opposition to each other: the relationships between race and identified risk factors in COVID-19 related morbidity and mortality, as recognised by researchers are often conjointly conditioned by both biological mechanisms and socio-economic factors. Furthermore, the intersectional approach –articulating various social relations of race, class, gender, age, disability, etc.– proves to be particularly fruitful in grasping the ethno-racial disparities in COVID-19 related morbidity and mortality.

Punitive Social Policy and Vital Inequality.

Geographical inequalities in life and death are among the world's most pronounced in the United States. However, the driving forces behind this macroscopic variation in population health outcomes remain surprisingly understudied, both empirically and theoretically. The present article steps into this breach by assessing a number of theoretically informed hypotheses surrounding the underlying causes of such spatial heterogeneity. Above and beyond a range of usual suspects, such as poverty, unemployment, and ethno-racial disparities, we find that a hitherto neglected explanans is prison incarceration. In particular, through the use of previously unavailable county-level panel data and a compound instrumentation technique suited to isolating exogenous treatment variation, high imprisonment rates are shown to substantially increase the population-wide risk of premature death. Our findings contribute to the political economy of population health by relating the rise of the carceral state to the amplification of geographically anchored unequal life chances.