JMIR Publications

ABSTRACT

Background:

Naloxone is a life-saving opioid antagonist, but the alignment between real-time public engagement and overdose risk remains unclear across U.S. states. Google search data may offer novel infodemiologic insights to optimize public health responses.

Objective:

To characterize spatiotemporal trends in opioid overdose mortality, naloxone dispensing, and naloxone-related digital search interest across U.S. states and the District of Columbia (2019–2023), examine temporal associations between overdose mortality and naloxone search interest, and assess whether naloxone dispensing modifies this relationship.

Methods:

We conducted a retrospective longitudinal panel study integrating monthly opioid overdose mortality, Google Health Trends (GHT)–derived Naloxone search probability, and annual Naloxone dispensing rates for all 50 states and the District of Columbia from 2019 to 2023. We assessed spatiotemporal trends and temporal associations using Dumitrescu-Hurlin panel Granger causality, as well as statewise and rolling-window Granger analyses. Multivariable fixed-effects panel regression evaluated associations between overdose mortality, Naloxone dispensing, and search probability, including interaction effects.

Results:

From 2019 to 2023, the mean annual opioid overdose mortality rate was 31.2 per 100,000 population (range, 7.8–95.4), and the average annual Naloxone dispensing rate was 456 per 100,000 (range, 100–2,500), both with marked state-level variation. Naloxone search probability remained low overall (mean 66.8 ± 33.8 SD) but showed episodic spikes in several states. Panel Granger causality indicated that increases in overdose mortality significantly preceded increases in Naloxone search interest at 1- and 3-month lags (p < 0.05). Statewise Granger analyses after correcting variable ordering yielded no more significant states than expected by chance at any lag tested. In multivariable panel regression, overdose mortality (β = 0.29, p = 0.31), Naloxone dispensing (β = 0.010, p = 0.13), and their interaction (β = –0.0003, p = 0.37) were not significantly associated with Naloxone search probability.

Conclusions:

Panel-level Granger analysis identified significant temporal coupling between overdose mortality and Naloxone-related search probability at 1- and 3-month lags, though statewise analyses did not yield significant individual-state signals. Fixed-effects regression found no significant contemporaneous associations. These findings suggest that digital search signals capture short-term behavioral dynamics rather than structural access patterns. Integrating digital search surveillance with traditional overdose and dispensing metrics could enhance timely public health responses to opioid overdose risk.