Mediation of race/ethnicity (R/E) disparities in mental health patient reported outcomes (PROS) among elderly female patients with breast cancer (BC) in the Surveillance, Epidemiology, and end results-medicare health outcomes survey (SEER-MHOS) database.

Abstract

e22527 Background: Variations in mental health PROs between Non-Hispanic Black (NHB) and Hispanic compared to Non-Hispanic White (NHW) elderly women with BC may be attributable to differences in socioeconomic position (SEP), comorbid disease (CD), tumor biology (TB), social support (SS), and treatment factors (TF). We examine the extent to which the R/E disparity in patient reported mental health summary scores (MCS) is transmitted through these domains. Methods: Data is from SEER-MHOS and the analytic sample (n = 17,465) included 12,493 NHW, 2,503 NHB, and 2,019 Hispanic women with newly diagnosed BC between 2010-2017. MCS were calculated using the SF-36/V linear transformation of 0-100 versions of the scale that yield a population mean of 50 and standard deviation of 10 based on 1990 population means. We estimated average controlled direct associations (ACDA) using regression models with predictive margins (risk difference %) to evaluate crude and age/body mass index (BMI)-adjusted MCS by R/E and performed a series of ACDA [with comparison of rescaled coefficients (RC)] controlling for the domains of (1) SEP [household income (< 30K, 30K-80K, > 80K) and education attainment [ < High School (HS), HS, > HS)]; (2) CD [smoking status (current smoker/current non-smoker), diabetes (yes/no), and hypertension (yes/no)]; (3) TB [stages (1-3), grade (well/moderate/poor), and subtype (luminal A/ triple negative)]; (4) SS [domestic partner (yes/no); and (5) TF [systemic therapy (yes/no)]. Results: Overall, NHB and Hispanic women were younger and had higher BMI than NHW women and crude models showed MCS disparity for NHBs (-4.13%; 95% CI: -4.63, -3.63) and Hispanics (-5.91%, 95%CI: -6.49, -5.34) compared with NHWs. In baseline age/BMI adjusted models, NHBs and Hispanics had worse MCS, -3.78% (95% CI: -3.2, -4.35) and -5.7% (95% CI: -5.04, -6.36) respectively, compared with NHWs. Control for TF had minimal effect on the observed disparity [NHB (-3.85%; 95% CI: -4.56, -3.15) and Hispanics (-4.79%; 95% CI: -5.6, -3.97) compared with NHW], but control for SEP largely reduced the disparity [NHB (-1.56%; 95% CI: -0.89, -2.23) and Hispanics (-2.8%; 95% CI: -1.99, -3.6)]. Inclusion of all 5 domains also attenuated the magnitude of the disparity to a degree such that it was still considered a minimally important difference for Hispanics (-2.34%, 95% CI: -3.75, -0.94), but not for NHBs (0.12%; 95% CI: -1.08, 1.32). However, low acculturation in Hispanics appeared to modify (worsen) the disparity. Based on the method of RC, we determined that SEP accounted for the largest fraction of mediated effects. Conclusions: Our findings suggest that observed R/E differences in mental health PROs in elderly BC patients are largely mediated by differences in SEP rather than treatment.