Abstract 10752: Clinical Predictive Models for Cardiovascular Disease Should Adjust for Race, Ethnicity and Social Determinants of Health to Inform More Equitable Interventions: Findings from a Targeted Review

Abstract

Introduction: Clinical models of cardiovascular disease (CVD) that incorporate relevant predictors can provide powerful insights into the complex, inter-related mechanisms of CVD pathophysiology, risk for adverse outcomes, and correlations with social determinants of health (SDoH) to inform practice guidelines and targeted interventions. Despite the growing body of evidence demonstrating the role of race, ethnicity and SDoH on CVD outcomes, limited evidence exists about incorporating these factors into models. Hypothesis: We assessed the hypotheses that clinical predictive studies on CVD risk, outcomes, prognosis and treatment interventions: (1) do not fully examine race, ethnicity and SDoH factors, and (2) may not leverage adequate data sources fully representative of a population of interest or may lack the decisional context to ensure robustness. These are critical limitations that could lead to inaccurate predictions, perpetuating bias and ultimately health and racial inequities. Methods: This targeted literature review yielded 533 potentially relevant PubMed articles (2018-2019) after duplicate removal. Article titles and abstracts were reviewed for eligibility where 375 met the exclusion criteria (age <18 years old, outside North America, no clinical outcomes, language non-English) and were removed. The remaining 158 potentially relevant articles were screened and 116 met the criteria for exclusion (e.g. stroke, heart transplant). After screening, 42 citations were identified for full-text screening, of which 35 articles met criteria for inclusion. Results: Studies employed advanced statistical and machine learning methods to predict CVD risk (10/29%), mortality (19/54%), survival (7/20%), complication (10/29%), disease progression (6/17%), functional outcomes (4/11%) and disposition (2/6%). Most studies incorporated sex (34/97%), co-morbid conditions (32/91%), and behavioral risk factor (28/80%) variables. Race/ethnicity (23/66%) and SDoH variables (e.g. education (3/9%)) were less frequently observed. Conclusions: In conclusion, clinical risk and predictive models often do not adjust for or consider race, ethnicity and SDoH to improve model accuracy to inform more equitable interventions and decision-making.