Computationally Efficient Estimation of Localized Treatment Effects for Multi-Level, Multi-Component Interventions to...

Area of Science:

• Public Health
• Computational Modeling
• Health Economics

Background:

• The opioid epidemic is a significant public health crisis in the US.
• Understanding localized treatment effects is vital for effective intervention and resource allocation.
• Exhaustive simulation of all intervention combinations is computationally impractical.

Purpose of the Study:

• To develop an efficient computational framework for estimating localized treatment effects of opioid epidemic interventions.
• To support policymakers in evaluating resource-allocation strategies for mitigating the opioid epidemic.

Main Methods:

• Developed a bi-level metamodel framework with a two-stage sequential sampling design.
• Utilized a response function and Gaussian Process Regression (GPR) to model treatment effects.
• Incorporated spatial and socio-economic covariates for localized effect estimation.

Main Results:

• The framework achieved approximately 5% average relative error in estimating treatment effects.
• The approach required only one-tenth the number of simulation runs compared to exhaustive methods.
• Successfully estimated treatment effects of buprenorphine dispensing and naloxone distribution on overdose mortality in Pennsylvania.

Conclusions:

• The bi-level metamodel framework offers a computationally efficient solution for evaluating opioid intervention strategies.
• This approach enables better-informed policy decisions for localized resource allocation.
• The study provides a valuable tool for mitigating the opioid epidemic at the community level.