(4) Optimizing Geographic Naloxone Distribution to Reduce Opioid Overdose Deaths

Background & Introduction: Naloxone is highly effective in preventing opioid overdose deaths (OODs), yet decision-makers lack practical tools to allocate naloxone to areas where demand is greatest and limited resources can achieve the greatest impact. To support evidence-based naloxone distribution, we propose a simulation-based decision-support framework to optimize the geographic allocation of naloxone to minimize OODs.

Methods: We developed a framework that combines an existing simulation model (PROFOUND) with a metamodel to optimize naloxone distribution strategies across geographic subregions, with the objective of minimizing OODs in a jurisdiction under a fixed annual number of naloxone kits. The PROFOUND model is an individual-based microsimulation model used evaluate naloxone distribution strategies. In the framework, we first generated a large number of random naloxone allocation schemes across subregions and used the simulation model to project the resulting OODs. Second, we trained a metamodel using simulation results to approximate the relationship between naloxone allocation to each subregion and projected OODs, using an artificial neural network to enable efficient optimization in a high-dimensional setting. Third, we applied a trust-region optimization to the trained metamodel under a fixed total number of naloxone kits to identify allocation strategies minimizing OODs. Last, the optimized solutions were validated through the original simulation model. We applied this framework in a case study of optimizing community-based naloxone distribution across cities/towns in Rhode Island under an annual constraint of 50,000 naloxone kits (level of community-based distribution in 2023).

Results: The metamodel provided excellent approximation to the simulation model while substantially reducing simulation time, enabling efficient optimization. At the state level, the optimized allocation reduced projected OODs from 358.5 under the status quo distribution (following distribution pattern in 2023) to 340.9 in 2026, corresponding to an average reduction of 17 deaths under the same total number of naloxone kits. The optimized strategy resulted in a different recommended geographic distribution of naloxone, with Providence remaining the largest recipient but receiving fewer kits than current allocation, and increased allocations to smaller and more underserved communities such as Coventry and Cumberland.

Conclusion & Discussion: The proposed framework can help reduce OODs by approximately 5% without increasing the number of naloxone kits distributed. This finding highlights the potential of evidence-based optimization to improve the public health impact of existing resources and inform broader resource allocation decisions.

References: Zang, X., Skinner, A., Krieger, M. S., Behrends, C. N., Park, J. N., & Green, T. C. (2024). Evaluation of strategies to enhance community-based naloxone distribution supported by an opioid settlement. JAMA Netw Open. 2024; 7 (5): e2413861.

Zang, X., Bessey, S. E., Krieger, M. S., Hallowell, B. D., Koziol, J. A., Nolen, S., ... & Marshall, B. D. (2022). Comparing projected fatal overdose outcomes and costs of strategies to expand community-based distribution of naloxone in Rhode Island. JAMA network open, 5(11), e2241174-e2241174.

Sugarman, Olivia K., Eric G. Hulsey, and Daliah Heller. "Achieving the potential of naloxone saturation by measuring distribution." JAMA Health Forum. Vol. 4. No. 10. American Medical Association, 2023.

Li, Zongbo, et al. "Meta-Modeling as a Variance-Reduction Technique for Stochastic Model–Based Cost-Effectiveness Analyses." Medical Decision Making 45.8 (2025): 976-986.