There are both risks and documented examples of patients receiving differential care as a result of how race is factored into guiding clinical algorithms.[1] Race-corrected algorithms may have unintended negative impacts on patient care and outcomes.[2] The American Heart Association's 2020 Presidential Advisory on Structural Racism asserts our unequivocal commitment to identifying and mitigating structural racism in research and clinical care. The statement specifically cites the need to "reconsider when and how to include race/ethnicity and social determinants of health measures in risk calculators… because its inclusion could have unintended adverse consequences for the care of patients."[3]
In fulfillment of its mission to be a relentless force for a world of longer, healthier lives, the Association's current, multi-year strategic impact goal is to advance cardiovascular health for all by identifying and removing barriers to health care access and quality. This equity-centric commitment builds upon nearly a century of moving cardiovascular and stroke care along a spectrum from one-size-fits-all guidance toward evidence-based, precision medicine, individualized for each patient. Capturing, disaggregating and considering race and ethnicity data are essential steps in this journey, to identify and address equity barriers and outcome disparities. Additional research and rigor are required to ensure appropriate and constructive use of this data while avoiding unintended negative impacts.
The Association has designed a two-year scientific research strategy funded in part by the Doris Duke Foundation to study the complex issue of how race and ethnicity factor into clinical care algorithms and risk prediction tools. The American Heart Association is one of five grantees as part of the Doris Duke Foundation's new initiative, Racial Equity in Clinical Equations. The De-Biasing Clinical Care Algorithms in Cardiovascular Care (DECCA) project purpose is to advance rigorous new scientific evidence, while amplifying awareness and discourse among the stakeholders and audiences in the Association's sphere of influence about the practice of race correction and its implications for equitable patient care.
"Health disparities may emerge as an unintended consequence of developing predictive algorithms from epidemiological studies. This research is an intentional effort to study and mitigate algorithm use cases where the inclusion of race as a descriptive variable in epidemiological studies could have negative equity implications in downstream patient care when converted to a predictive variable," said Jennifer Hall, Ph.D., FAHA, chief of data science for the American Heart Association. "Studying and seeking to constantly improve how race and related factors are incorporated into algorithms – whether internally within the formulas or via unintended contributions to disparate care delivery – is an important scientific endeavor in our relentless pursuit of improved health and health equity for all."
The pursuit to create and optimize clinical decision-making support tools that acknowledge and address racial disparities in the most productive, harm-reducing way is rooted in scientific rigor and guided by several key principles:
- Race and ethnicity data is valuable for identifying and addressing equity barriers and disparities in observational research.
- Race is a social construct, not a biological variable determined by genetic factors.
- The determination of race, whether done by the individual under study or by a researcher, is fraught with potential for bias, imprecise and often inappropriate.
- Race-corrected clinical care algorithms may have unintended negative impacts.
- The use of race and ethnicity data in research needs to be handled carefully and with attention to its purpose and potential pitfalls.
This work is also rooted in, and builds on, related parallel efforts of the Association's equity agenda over many years, including:
- Expanded collection of social determinants of health (SDOH) data points in our clinical registries and other datasets (to mitigate reliance on race as a proxy);
- Funded research on how best to harmonize SDOH variables in diverse health datasets;
- Funded research on approaches to increase diversity in clinical trial participation and resulting datasets;
- Raised awareness on systemic bias in risk prediction; and
- Funded and published the above-mentioned studies focused on debiasing algorithms in our spheres of influence.
The project will leverage significant matched resources, including research funding, expert volunteers, datasets, dialogue within the cardiovascular community at the Association's influential scientific meetings and in its clinical guidance documents including scientific statements, presidential advisories, and guidelines.
