Advancing Spatial Evaluation Methods to Improve Healthcare Efficiency and Quality

描述（由申请人提供）：拟议的研究基于这样一种信念，得到了广泛的理论和经验证据的支持，即需要一种新的、空间明确的分析方法来正确和有力地评估基于地方的卫生政策的有效性、成本和效益。这项工作将直接有助于实现AHRQ的价值组合中概述的研究目标。我们将开发新的空间分析方法，专门用于解决目前可用于计划或政策分析的技术的不足。迄今为止，用于规划和政策评估的统计和计量经济学方法仍然主要以缺乏对空间溢出效应的考虑为特征，即一个社区发生的事情同时影响到其邻居，导致结果变量的非独立性。虽然现有方法可以解释空间相关的解释变量和聚集在共同地点的结果（即空间自相关的外源性来源），但现有方法忽略了空间溢出的同时（内生）动态。大量证据表明，忽视这种空间溢出效应可能导致参数估计有偏差和不一致，对不确定性的量化产生误导，以及模型预测存在缺陷。这对基于地点的干预或政策影响的估计可能产生严重后果，导致高估或低估项目效果的估计，使成本效益模拟实验和未来政策决策产生偏差。我们将解决关键的方法差距，并通过以下目标传播新方法：目标1：开发用于政策评估的新的空间分析方法，并在用户友好的开源软件中实施。我们将开发创新的空间分析方法，明确联合治疗空间依赖性，空间依赖性

DESCRIPTION (provided by applicant): The proposed research is based on the conviction, supported by extensive theoretical and empirical evidence, that a new, spatially explicit analytical methodology is required to properly and robustly assess the effectiveness, costs, and benefits of place-based health policies. The work will contribute directly to attaining the researc objectives outlined in AHRQ's Value Portfolio. We will develop new spatial analytical methods designed specifically to address deficiencies in the techniques currently available for program or policy analysis. To date, the statistical and econometric methods employed in program and policy evaluation are still mostly characterized by a lack of accounting for spatial spillover effects-where what happens in one community simultaneously impacts its neighbors, leading to non-independence in the outcome variable. While existing methods may account for spatially correlated explanatory variables and outcomes clustered within common places (i.e. exogenous sources of spatial autocorrelation), existing methods ignore the simultaneous (endogenous) dynamics of spatial spillovers. Extensive evidence suggests that ignoring such spatial spillover effects can lead to biased and inconsistent parameter estimates, misleading quantification of uncertainty, and flawed model prediction. This has potentially serious consequences for the estimation of place-based intervention or policy impacts, leading to overstated or understated program effect estimates, biasing simulation experiments of cost-effectiveness and future policy decisions. We will address critical methodological gaps and disseminate new methods through the following Aims: Aim 1: To develop new spatial analytical methods for use in policy evaluations and implement them in user-friendly open source software. We will develop innovative spatial analytic methods for the explicit joint treatment of spatial dependence, spatial heterogeneity, and selectivity in panel data models and implement them as additions to our well-established software development and dissemination efforts. Aim 2: To conduct spatially explicit evaluation analysis and disseminate the findings and applied methods. We will assess the effects of particular Medicare health policy changes that were implemented in 2006 in a natural experimental (pre-post) space-time research design, to explore changes in disparities in the utilization of colorectal cancer (CRC) screening and the geographic diffusion of CRC screening technology over time. Selection bias is prevalent, as the elderly selectively enroll in managed care plans, which were significantly impacted by Medicare reforms, and in a geographically disparate fashion. While this policy application is important, the methods to be developed are broadly applicable to many policy evaluation contexts, where the combination of spatial spillover effects, other forms of spatial autocorrelation, various sources of selection bia, and inappropriately or un-modeled spatial heterogeneity may critically affect the measured policy impact.