Effects of naloxone distribution to likely bystanders: Results of an agent-based model

Abstract

Background: Opioid overdose deaths in the US rose dramatically in the past 16 years, creating an urgent national health crisis with no signs of immediate relief. In 2017, the President of the US officially declared the opioid epidemic to be a national emergency and called for additional resources to respond to the crisis. Distributing naloxone to community laypersons and people at high risk for opioid overdose can prevent overdose death, but optimal distribution methods have not yet been pinpointed.

Methods: We conducted a sequential exploratory mixed methods design using qualitative data to inform an agent-based model to improve understanding of effective community-based naloxone distribution to laypersons to reverse opioid overdose. The individuals in the model were endowed with cognitive and behavioral variables and accessed naloxone via community sites such as pharmacies, hospitals, and urgent-care centers. We compared overdose deaths over a simulated 6-month period while varying the number of distribution sites (0, 1, and 10) and number of kits given to individuals per visit (1 versus 10). Specifically, we ran thirty simulations for each of thirteen distribution models and report average overdose deaths for each. The baseline comparator was no naloxone distribution. Our simulations explored the effects of distribution through syringe exchange sites with and without secondary distribution, which refers to distribution of naloxone kits by laypersons within their social networks and enables ten additional laypersons to administer naloxone to reverse opioid overdose.

Results: Our baseline model with no naloxone distribution predicted there would be 167.9 deaths in a six month period. A single distribution site, even with 10 kits picked up per visit, decreased overdose deaths by only 8.3% relative to baseline. However, adding secondary distribution through social networks to a single site resulted in 42.5% fewer overdose deaths relative to baseline. That is slightly higher than the 39.9% decrease associated with a tenfold increase in the number of sites, all distributing ten kits but with no secondary distribution. This suggests that, as long as multiple kits are picked up per visit, adding secondary distribution is at least as effective as increasing sites from one to ten. Combining the addition of secondary distribution with an increase in sites from one to ten resulted in a 61.1% drop in deaths relative to the baseline. Adding distribution through a syringe exchange site resulted in a drop of approximately 65% of deaths relative to baseline. In fact, when enabling distribution through a clean-syringe site, the secondary distribution through networks contributed no additional drops in deaths.

Conclusion: Community-based naloxone distribution to reverse opioid overdose may significantly reduce deaths. Optimal distribution methods may include secondary distribution so that the person who picks up naloxone kits can enable others in the community to administer naloxone, as well as targeting naloxone distribution to sites where individuals at high-risk for opioid overdose death are likely to visit, such as syringe-exchange programs. This study design, which paired exploratory qualitative data with agent-based modeling, can be used in other settings seeking to implement and improve naloxone distribution programs.

Keywords: Agent-based modeling; Harm reduction; Naloxone distribution; Opioid overdose.