Background The toxic unregulated drug supply in North America continues to produce high rates of drug deaths. In response, several harm reduction interventions have been introduced and/or expanded, including take-home naloxone (THN). Estimating the impact is challenged by a lack of complete reporting data. Objective The aim of this study was to estimate the impact of interventions on drug deaths in British Columbia from January 2019 to September 2024. Methods We extended on a Bayesian hierarchical Markov chain model of drug poisoning events including interventions for overdose prevention sites and opioid agonist treatment. The extended model uses the reported number of THN kits used and distributed and all kits shipped to sites. These data are incorporated into the likelihood to estimate THN kit use during an opioid poisoning event by region and site type. Simulation studies evaluated the model’s performance. Results The estimated probability of THN kit use during an opioid poisoning event was 42.98% (95% credible interval CrI: 41.12–44.84) for kits distributed from community sites and 13.41% (95% CrI: 12.57–14.40) for overdose prevention sites. Correctional centers, pharmacy, and emergency department THN kits all had the lowest probability of use at 0.12% (95% CrI: 0.11–0.13), 1.04% (95% CrI: 0.96–1.13), and 0.65% (95% CrI: 0.60–0.71), respectively. The combined rate of deaths averted was 1,294 (95% CrI: 1,138–1,438) per 100,000 persons who inject drugs, which represents 78% (95% CrI: 76–80) of potential deaths. Conclusion Despite the increasing toxicity of the illegal drug supply, harm reduction interventions including THN have had a large impact on the number of drug deaths. Estimates of the impact of THN based on reported use alone would greatly underestimate the total impact. Highlights We developed a novel Bayesian hierarchical model to estimate take-home naloxone (THN) kit use during opioid poisonings using incomplete but complementary program and surveillance data. The model provides site-specific and regional estimates of kit use, highlighting significant differences by site type and geography. Simulation studies show the model can estimate the probability of THN kit use under realistic data limitations, supporting its use in policy evaluation. Public health decision makers can use this method to better assess and optimize harm reduction programs when direct usage data are scarce.
Irvine et al. (2026) studied this question.