Predicting fatal and non-fatal overdose in Los Angeles County with Rapid Overdose Surveillance Dashboard to target street-based addiction treatment and harm reduction services

点击翻译按钮获取中文摘要

PROJECT SUMMARY (ABSTRACT) Street medicine teams play a key role in the local overdose response, yet existing data sources lag and are no more granular than the county- or zip code level. Geolocated data exist from various sources that could be used to identify hotspots and inform street-based overdose prevention and addiction treatment, but the data sources are not harmonized or available to the public. One geographic area where these concerns are particularly acute is Los Angeles County (LAC), California, with a population of over 10 million, and which had the largest number of fatal overdoses of any U.S. county in 2020. A rising share of fatal overdoses in LAC occur among unhoused people and involve both stimulants and opioids. The objective of the project is to develop tools and processes to improve timely data acquisition from, rapid processing and integration of diverse data sources, geospatial anal- ysis of overdose hotspots (fatal and non-fatal), and nowcasting of overdose, opioid use disorder (OUD), and injection drug use in LAC. Building on our research team's work in mobile overdose prevention and treatment of OUD in LAC, we will collaborate with five local government agencies with interest and experience in improving local overdose response. Our partners span public health, medicine, emergency medical services (EMS), med- ical examiner and coroner, syringe services programs, and street-based harm reduction. Together, we propose to develop a publicly available Rapid Overdose Surveillance Los Angeles online dashboard that can provide local, granular data and more timely estimates of countywide metrics. To establish the dashboard, we pursue two specific aims. In Aim 1, we will establish data flows to collate geolocated fatal overdose data from the coroner and non-fatal overdose from EMS, adapting natural language processing (NLP) methods to classify free-text data that characterize the specific drugs involved. We then employ geostatistical methods to identify hotspots at the census tract level, providing localization to inform placement of mobile and street-based services. Finally, we will develop nowcasting models to “predict the present” of fatal and non-fatal overdose at the county-level based on incomplete surveillance data. In Aim 2, we will develop further NLP strategies to identify upstream outcomes of overdose (i.e., OUD and injection drug use) in electronic health record data. We will then incorporate metrics from substance use disorder treatment, syringe services, and street medicine to improve our estimates of OUD, injection drug use, and overdose at the county-level. We will visualize these data and nowcasting results through Rapid Overdose Surveillance Los Angeles online dashboard. Findings from these efforts will serve as a model for other jurisdictions to leverage and combine data from diverse stakeholders to improve local situational awareness of overdose. Ultimately, the goal is to produce tools and processes that can speed up the time from data to action to better provide overdose prevention and addiction treatment services.