A Bayesian Mixture Model Approach to Examining Neighborhood Social Determinants of Health Disparities in Endometrial Cancer Care in Massachusetts

Many studies have examined social determinants of health (SDoH) factors independently, overlooking their interconnected and intersectional nature. Our study takes a multifactorial approach to construct a neighborhood level measure of SDoH and explores how neighborhood residency impacts care received by endometrial cancer patients in Massachusetts. We used a Bayesian multivariate Bernoulli mixture model to create and characterize neighborhood SDoH (NSDoH) profiles using the 2015-2019 American Community Survey at the census tract level (n=1478), incorporating 18 variables across four domains: housing conditions and resources, economic security, educational attainment, and social and community context. We linked these profiles to Massachusetts Cancer Registry data to estimate the odds of receiving optimal care for endometrial cancer using Bayesian multivariate logistic regression. The model identified eight NSDoH profiles. Profiles 1 and 2 accounted for 27% and 25% of census tracts, respectively. Profile 1 featured neighborhoods with high homeownership, above median incomes, and high education, while Profile 2 showed higher probabilities of limited English proficiency, renters, lower education, and working class jobs. After adjusting for sociodemographic and clinical characteristics, we found no statistically significant association between NSDoH profiles and receipt of optimal care. However, compared to patients in NSDoH Profile 1, those in Profile 2 had lower odds of receiving optimal care, OR = 0.77, 95% CI (0.56, 1.07). Our results demonstrate the interconnected and multidimensional nature of NSDoH, underscoring the importance of modeling them accordingly. This study also highlights the need for targeted interventions at the neighborhood level to address underlying drivers of health disparities, ensure equitable healthcare delivery, and foster better outcomes for all patients.