1AbstractFrom 1999 to 2017, the age-adjusted annual death rate in the United States from opioid overdose has increased several-fold, to about 18.3 per 100,000 population. The federal government has declared opioid overdose a public health emergency. Spatially-aware analyses and surveillance may contribute to control of this growing problem.As naloxone has only one clinical use—the treatment of opioid overdose—its administration by emergency medical services (EMS) personnel can serve as a surveillance indicator for opioid overdose. I previously demonstrated spatial clustering of naloxone-involved EMS calls, over and above that to be expected fromEMS calls in general. To better understand the nature of that clustering, I modelled the spatial distribution of naloxone-involved EMS calls in a three-county region in south-central New York State as an inhomogeneous Poisson process, using as predictors several census-tract-level sociodemographic variables and the point locations of convenience stores (minimarts). In Monte Carlo simulations, I examined how well the model explained the observed clustering.Although it assumes no interaction between event locations, the inhomogeneous Poisson model was nevertheless able to explain much of the observed spatial clustering. The spatial intensity of EMS calls for opioid overdose, in events per square kilometer, was signficantly higher in census tracts with lower rates of owner-occupancy of housing. Neither the proportion of households in poverty nor the proportion of residents in the 20–44 age band were signficant predictors. Spatial intensity of overdose events decreased by about 10% for each kilometer of distance from the nearest minimart, but this was of borderline signfificance at conventional levels. These findings cast some doubt on the utility of real-time surveillance for apparent spatial clusters of opioid overdose and instead favor a longer-term, systemic strategy comprising efforts to improve neighborhood conditions.
