Analysis of breast cancer screening disparities in an academic health system

Abstract

Background: Breast cancer screening disparities among Black women have been recognized for decades yet persist. Health systems interested in implementing tailored interventions to improve breast cancer screening disparities must first understand local determinants. Econometrics methods can be employed to evaluate the contribution of factors to the mean difference in outcomes between two groups. Methods: We conducted a cross-sectional analysis including 20,147 individuals who identified as white (n=18,749) or Black (n=1,398) race and were eligible for breast cancer screening within a large academic health system. We evaluated predisposing characteristics (age, previous screening, use of patient portal, number of office visits, smoking status, medical conditions) and enabling resources at insurance and provider (primary care provider (PCP) specialty and training, PCP clinical full time equivalent, clinic location) levels. We conducted logistic regression analyses and a Blinder-Oaxaca (BO) decomposition to evaluate determinants in breast cancer screening disparities. Results: Black and white individuals differed on several factors; those who were Black were younger (mean age 61.5 ± 6.0 years vs. 63.2 ± 6.5 years, standardized mean difference (SMD)= -0.26), had less patient portal use (63.8% vs. 90.4%, SMD=0.67), higher rates of diabetes (29.8% vs. 11.8%, SMD= 0.45), higher rates of tobacco use (13.8% vs. 6.7%, SMD= 0.24), more Medicaid (19.0% vs. 6.2%, SMD= 0.39) insurance, and more often received primary care from a county hospital-based clinic (31.4% vs. 2.3%, SMD=0.82). Breast cancer screening was completed in 64.2% of Black individuals and 71.6% of white individuals (average marginal effect (AME) -0.07, 95% CI -0.10 to -0.05, p < 0.001). In the adjusted logistic regression analysis, Black individuals had a higher estimated likelihood to receive breast cancer screening as compared to white individuals (AME=0.06, 95% CI 0.03-0.09, p < 0.001). In the overall two-fold BO decomposition analysis, observed factors explained 12.7% difference in breast cancer screening. Contributing factors included patient portal use (4.1 percentage points or 32% of total difference), primary care site at a county hospital-based clinic (2.2 percentage points or 17.6% of total difference), diabetes (1.2 percentage points or 9.7% of total difference), and tobacco use (0.9 percentage points or 7.3% of total difference). Conclusion: In a BO decomposition analysis using data from a large academic health system, patient portal use, receiving care from a county-hospital based clinic, diabetes, and tobacco use were the factors that contributed most to the explained difference. These results can help to inform health system efforts to tailor interventions to improve racial disparities in breast cancer screening.