Emergency department use and geospatial variation in social determinants of health: a pilot study from South Carolina

Abstract

Background Health systems are increasingly addressing patients’ social determinants of health (SDoH)-related needs and investigating their effects on health resource use. SDoH needs vary geographically; however, little is known about how this geographic variation in SDoH needs impacts the relationship between SDoH needs and health resource use. Methods This study uses data from a SDoH survey administered to a pilot patient population in a single health system and the electronic medical records of the surveyed patients to determine if the impact of SDoH needs on emergency department use varies geospatially at the US Census block group level. A Bayesian zero-inflated negative binomial model was used to determine if emergency department visits after SDoH screening varied across block groups. Additionally, the relationships between the number of emergency department visits and the response to each SDoH screening question was assessed using Bayesian negative binomial hurdle models with spatially varying coefficients following a conditional autoregressive (CAR) model at the census block group level. Results Statistically important differences in emergency department visits after screening were found between block groups. Statistically important spatial variation was found in the association between patient responses to the questions concerning unhealthy home environments (e.g. mold, bugs/rodents, not enough air conditioning/heat) or domestic violence/abuse and the mean number of emergency department visits after the screen. Conclusions Notable spatial variation was found in the relationships between screening positive for unhealthy home environments or domestic violence/abuse and emergency department use. Despite the limitation of a relatively small sample size, sensitivity analyses suggest spatially varying relationships between other SDoH-related needs and emergency department use.

Characterizing Patient Phenotypes and Emerging Trends in Application of Spatial Analysis in Individual-Level Health Data

“…Moreover, hierarchical Bayesian that can incorporate hierarchical structures for modeling interactions in data with multiple levels [76] was used to investigate spatial distributions of patients admitted for drug-related reasons concerning the area deprivation index [24]. Bayesian negative binomial hurdle models that can account for excessive zeros and overdispersion were used by [26] to examine spatial variation between patient responses to the questions concerning unhealthy home environments and the mean number of emergency department visits after screening.…”