Dr. Jason Goldstick
This project will use the System for Overdose Surveillance, created by the U-M Injury Prevention Center, to conduct a rigorous comparison of several powerful spatiotemporal prediction methods to determine which methods best predict future overdose hotspots within a city and disseminate this information to key community stakeholders. Our overall goal is to apply spatiotemporal modeling techniques to a unique combination of data sources to characterize, and predict, overdose trends in Michigan–a large U.S. state among the hardest hit, with great variability in urbanicity, demographics, and service availability. We will characterize trends by estimating how spatiotemporally proximate factors combine to modulate area-level overdose rates, and we will predict hotspots by comparing several methods of short-term spatial prediction. These aims will leverage the pre-existing System for Overdose Surveillance (SOS) we have been running at the University of Michigan Injury Prevention Center since 2019, and a rich set of place-based features obtained both publicly and via preexisting partnerships.