Network Analysis of U.S. Non-Fatal Opioid-Involved Overdose Journeys, 2018-2023

TL;DR

This paper analyzes the spatial dynamics of non-fatal opioid-involved overdoses by constructing a nationwide county-level geospatial network from EMS data spanning 2018–2023. Edges encode journeys from a resident county to the overdose location, enabling cross-county mobility analysis and the identification of import/export patterns. The authors apply network metrics (degree, reciprocity, HITS hub/authority scores) and edge-persistence analyses to characterize distribution and evolution of journeys, revealing that urban fringe counties host key hubs and authorities and that long-distance journeys occur toward these counties. The study finds a median journey length of $47.54$ km and a mean of $105.4$ km, with about 6% discordant journeys and a heavy-tailed distance distribution; temporal persistence is evident with an undirected correlation $\gamma \approx 0.303$, supporting targeted harm-reduction deployment. Limitations include uneven data coverage, county-level geography, and focus on nonfatal events, but the framework offers a scalable approach for monitoring cross-county overdose mobility and guiding public health interventions.

Abstract

We present a nation-wide network analysis of non-fatal opioid-involved overdose journeys in the United States. Leveraging a unique proprietary dataset of Emergency Medical Services incidents, we construct a journey-to-overdose geospatial network capturing nearly half a million opioid-involved overdose events spanning 2018-2023. We analyze the structure and sociological profiles of the nodes, which are counties or their equivalents, characterize the distribution of overdose journey lengths, and investigate changes in the journey network between 2018 and 2023. Our findings include that authority and hub nodes identified by the HITS algorithm tend to be located in urban areas and involved in overdose journeys with particularly long geographical distances.

Figures (11)

• Figure 1: Visualization of the network of overdose journeys laid over a U.S. map. For visual clarity, the edge widths shown are proportional to the square root of the true edge weights. Self-loops and edge directionality are omitted.
• Figure 2: Relationships between the weighted in-degree, weighted out-degree, and population of the node. A) Relationships between the weighted in-degree and the weighted out-degree in the overdose journey network excluding self-loops. B) Relationship between the weighted out-degree and the population. C) Relationships between the weighted in-degree and the population. In-degree and out-degree values have been incremented by one to ensure visibility in log-transformed space. In all three cases, the weighted degrees exhibit slightly sub-linear relationships as functions of population size. The $R^2$ values represent correlations between log-transformed values. The $m$ values represent the slopes of the linear regression in the log-transformed space. Outlier points, i.e., counties for which the magnitude of the residuals between observed values and those predicted by the linear regression exceed three standard deviations, are highlighted in red. The blue arrows indicate the county that is an outlier in both panels B and C.
• Figure 3: Distribution of distances of geographically discordant overdose journeys. A) Empirical distribution of lower-bound distances among the journeys. B) Empirical distribution of upper-bound distances. C) Estimated survival function of the journey distance. D) Average distance of geographically discordant overdose journeys. "All" refers to all journeys. "Top IPC" and "Authority" refer to journeys to the top IPC and authority counties, respectively. "Top OPC" and "Hub" refer to journeys from the top OPC and hub counties, respectively.
• Figure 4: Time courses of the fraction of self-loops and the overdose journey distance. A) Fraction of edge weights owing to self-loops, indicating the portion of overdose events occurring in the home county of the person who experienced an overdose. B) Average and median journey distance for geographically discordant events. The dashed lines indicate the date at which the World Health Organization declared the beginning of the COVID-19 Public Health Emergency of International Concern (PHEIC) npr_covid_2023.
• Figure 5: Persistence of edges in the temporal overdose journey network. A) Average undirected temporal correlation coefficients for all counties, and the top IPC, OPC, authority, and hub counties. B) Average directed (i.e., in- and out-) temporal correlation coefficients for all counties, and the top IPC, OPC, authority, and hub counties. In both the undirected and directed settings, focal importers and exporters have greater-than-average temporal correlation coefficients.