Catalogue of bias: racial bias

• Published in: BMJ evidence-based medicine Vol. 29; no. 2; pp. 114 - 116
• Main Authors: Naicker, Ramona, Nunan, David
• Format: Journal Article
• Language: English
• Published: England BMJ Publishing Group Ltd 01.04.2024; BMJ Publishing Group LTD
• Subjects: African Americans; Algorithms; Bias; Black people; Breast cancer; Cancer; COVID-19; EBM learning; Ethics; Ethnicity; Evidence-Based Practice; Genetic engineering; Genetic testing; Genomes; Hispanic Americans; Humans; Inclusion; Influence; Kidneys; Medical ethics; Medical research; Medicine; Minority & ethnic groups; Prejudice; Public health; Race; Racial discrimination; Racism; Racism - prevention & control; Systematic review; Validity; White People; Womens health; United States > US; Evidence-Based Practice; Ethics
• ISSN: 2515-446X; 2515-4478; 2515-4478
• DOI: 10.1136/bmjebm-2023-112400

Racial bias in research can take various forms, including the systemic under-representation of ethnic minorities (related to sampling and recruitment bias, impacting external validity), the use of non-validated methods or tools to analyse data from diverse populations (related to measurement bias and impacting construct validity),1 and the inappropriate interpretation of disparities in research findings due to the presentation of the social construct of race as biological and failure to recognise how it influences health outcomes (related to observer bias, confounding bias, confirmation bias, impacting internal validity).2 Racial bias can stem from systemic, institutional, interpersonal or individual forms of prejudice against a race or ethnicity. Prejudice can be explicit or implicit and can affect all stages of a health research project or study, through lack of diverse research teams and recruitment strategies, mistrust, language and cultural barriers,3 simplification of complex issues, convenience,4 or the legacy of incorrect beliefs about biological racial differences, historically used to justify white rule and systems of oppression.5 While the under-representation of racial and ethnic minorities in research studies is a commonly recognised issue, identifying it can be challenging when data on race and/or ethnicity are not reported. In genetic testing studies on race and health outcomes, recognising genetic ancestry as distinct from race is crucial to address its potential confounding influence on observed relationships and avoid mistakenly attributing them to biological race-based differences.6 For example, a study exploring genetic polymorphisms and pharmacotherapeutic responses on the diagnosis and treatment of depression between East and West, incorrectly conflates genetic variation with race and ethnicity as a proxy for genetic ancestry.8 Example In a systematic review of trials that led to Food and Drug Administration (FDA) approvals for oncology drugs, race was not reported in more than one third of the included trials.9 Among trials that documented race, black and Hispanic populations were significantly under-represented compared with estimated US cancer populations. Studies have highlighted the potential dangers of such practices, outlining algorithms that, when traced to their origins, lead to outdated, suspect racial science or biased data.7 These algorithms include: a breast cancer risk calculator which places white women at higher risk, discouraging more rigorous screening for non-white women; a calculator to estimate the risk of death/complications during surgery which place ethnic minorities at higher risk, steering minority patients away from surgery (though authors acknowledge that the mechanism underlying these differences is not known); the Kidney Donor Risk Index tool which uses African American race to predict kidney graft failure, reducing the pool of donations from this population, and therefore, the amount of kidneys available for African American patients waiting for transplants.7 A study by Obermeyer et al highlighted racial bias in a widely used health system algorithm predicting patients’ healthcare costs, not illness severity.11 This flawed process is exacerbated by black patients often receiving lower quality of care than white patients with similar health issues, leading to inaccurate predictions and compromised care.