The opioid epidemic remains a major public health challenge in the United States, requiring a multi-pronged intervention approach to mitigate harms to communities. Given the heterogeneity of the epidemic across the country, it is crucial for policymakers to understand localized treatment effects of different intervention components and utilize limited resources efficiently. While locally calibrated simulation models offer a useful computational tool to project the epidemic outcomes for any given intervention policy, collecting simulation results for all intervention combinations to estimate localized treatment effects for each community is impractical because the number of possible intervention combinations grows exponentially with the number of interventions and levels at which they are applied. To tackle this, we develop a bi-level metamodel framework with a two-stage sequential design for efficient sampling. The metamodel consists of a response function linking health outcomes to each intervention component's treatment effect, and a Gaussian process regression to learn spatial and socio-economic structures of the treatment effects based on locally-contextualized covariates. With two-stage sequential sampling, we leverage spatial correlations and posterior uncertainty to sequentially sample the most informative counties and treatment conditions. We apply this framework to estimate treatment effects of buprenorphine dispensing and naloxone distribution on overdose mortality rates using a calibrated agent-based opioid epidemic model in PA counties. Our approach achieves approximately 5% average relative error using one-tenth the number of runs required for an exhaustive simulation. Our bi-level framework provides a computationally efficient approach to support policymakers, in evaluating resource-allocation strategies to mitigate the opioid epidemic in local communities.

翻译：阿片类药物流行病仍是美国面临的重大公共卫生挑战，需要采取多管齐下的干预策略以减轻对社区的危害。鉴于疫情在全国范围内的异质性，决策者必须了解不同干预成分的局部处理效应，并高效利用有限资源。尽管本地校准仿真模型为预测任意干预政策下的疫情结局提供了有用的计算工具，但采集所有干预组合的仿真结果以估计各社区的局部处理效应在实践中不可行——因为可能的干预组合数量随干预措施及其应用层级呈指数增长。为解决这一问题，我们构建了一个具有两阶段序贯设计的双层元模型框架，以实现高效采样。该元模型包含连接健康结局与各干预成分处理效应的响应函数，以及基于本地情境化协变量学习处理效应空间与社会经济结构的高斯过程回归。通过两阶段序贯采样，我们利用空间相关性及后验不确定性，依次采样信息量最大的县与处理条件。将该框架应用于宾夕法尼亚州各县经校准的基于智能体的阿片类药物流行病模型，以评估丁丙诺啡配发与纳洛酮分布对用药过量死亡率的影响。结果表明：仅需穷举仿真所需运行次数的十分之一，平均相对误差即可达约5%。我们的双层框架为决策者评估缓解本地社区阿片类药物流行的资源分配策略提供了高效计算途径。