SELECTION AND RISK ADJUSTMENT IN PRIVATE EMPLOYER PLANS

描述：这个项目利用自然实验的优势， 为福特汽车公司的雇员提供的计划。在头两年， 他们没有选择的计划，但在第三，他们可以选择他们的计划。这 允许研究小组检查计划的影响和不利的影响， 在利用率和成本方面进行选择。之间的唯一区别 受试者干预前后是衰老和衰老的影响之一 年另一个好处是，来自管理式护理计划的遭遇数据是 可用于从基于索赔的系统中增加数据。相同的方法 估算费用-使用赔偿计划单位价格-将用于所有 布局这种分析可以对配偶和家属以及 为员工。 该研究将采用变化分数作为患者特征的函数 和用于切换的指示器。分层方法可用于嵌套 家庭中的个体，可能使用两部分模型。前后对比 转换者和保留者的收益率估计差异 计划+逆向选择效应和计划直接效应。分析 按服务（MH与其他）将提供有关差异选择的信息 和响应能力。 在第二个子项目中，他们将模拟分配结果 使用福特和的替代风险调整方法（PIP-DCG和HCC） 哈佛员工。数据使用模拟和其他信息 补偿，他们可以评估任何再分配的幅度，以及 谁会受到影响底层建模方法将是 与Keeler、Carter和纽豪斯（JHE 1998）的研究相同。

Description: This project takes advantage of a natural experiment in choice of plan for employees of the Ford Motor Company. In the first two years they have no choice of plan, but in the third, they can chose their plan. This allows the research team to examine the effects of plan and of adverse selection in terms of utilization and costs. The only difference between the subjects before and after is the intervention, and the effects of aging one year. A further bonus is that encounter data from managed care plans is available to augment data from claims based systems. The same method for imputing costs - using indemnity plan unit prices - will be used for all plans. This analysis can be conducted for spouses and dependents as well as for employees. The study will employ change scores as a function of patient characteristics and an indicator for switching. Hierarchical methods can be used to nest individuals within families, possibly using two-part models. The before-and-after difference for those who switched and those who stayed yield estimates of the plan + adverse selection effect and the plan effect directly. Analysis by service (MH vs. other) will provide information on differential selection and responsiveness. In the second sub-project, they will simulate the distributional consequences of alternative risk adjustment methods (PIP-DCGs and HCCs) using Ford and Harvard employee. data. Using the simulations and information on other compensation, they can assess the magnitude of any redistribution, as well as saying who would be affected. The underlying modeling methods will be the same as that in Keeler, Carter and Newhouse (JHE 1998).