Spatial heterogeneity and socioeconomic determinants of opioid prescribing in England between 2015 and 2018

Background Opioid overdoses have had a serious impact on the public health systems and socioeconomic welfare of several countries. Within this broader context, we focus our study on primary care opioid prescribing in England from 2015 to 2018, particularly the patterns of spatial variations at the community level and the socioeconomic and environmental factors that drive consumption. Methods Leveraging open data sources, we combine prescription records with aggregated data on patient provenance and build highly granular maps of Oral Morphine Equivalent (OME) prescribing rates for Lower Layer Super Output Areas (LSOA). We quantify the strength of spatial associations by means of the Empirical Bayes Index (EBI) that accounts for geographical variations in population density. We explore the interplay between socioeconomic and environmental determinants and prescribing rates by implementing a multivariate logistic regression model across different temporal snapshots and spatial scales. Results We observe, across time and geographical resolutions, a significant spatial association with the presence of localized hot and cold spots that group neighboring areas with homogeneous prescribing rates (e.g., EBI = 0.727 at LSOA level for 2018). Accounting for spatial dependency effects, we find that LSOA with both higher employment deprivation (OR = 62.6, CI 52.8–74.3) and a higher percentage of ethnically white (OR = 30.1, CI 25.4–35.7) inhabitants correspond to higher prescribing rates. Looking at educational attainment, we find LSOA with the prevalent degree of education being apprenticeship (OR = 2.33, CI 1.96–2.76) a risk factor and those with level 4+ (OR = 0.41, CI 0.35–0.48) a protective factor. Focusing on environmental determinants, housing (OR = 0.18, CI 0.15–0.21) and outdoor environment deprivation (OR = 0.62, CI 0.53–0.72) indices capture the bi-modal behavior observed in the literature concerning rural/urban areas. The results are consistent across time and spatial aggregations. Conclusions Failing to account for local variations in opioid prescribing rates smooths out spatial dependency effects that result in underestimating/overestimating the impact on public health policies at the community level. Our study suggests a novel approach to inform more targeted interventions toward the most vulnerable population strata.