Health plan enrollment and mortality in the Medicare program
使用Medicare数据，通过样本选择模型控制可观测和不可观测混杂因素，比较健康维护组织（HMO）与传统按服务收费（FFS）参保者的两年预测死亡率，发现调整不可观测因素后HMO死亡率更低，提示存在逆向选择。
Prior studies have found that Medicare health maintenance organization (HMO) enrollees have lower mortality (over a fixed observation period) than beneficiaries in traditional fee-for-service (FFS) Medicare. We use Medicare Current Beneficiary Survey (MCBS) data to compare 2-year predicted mortality for Medicare enrollees in the HMO and FFS sectors using a sample selection model to control for observed beneficiaries characteristics and unobserved confounders. The difference in raw, unadjusted mortality probabilities was 0.5% (HMO lower). Correcting for numerous observed confounders resulted in a difference of -0.6% (HMO higher). Further adjustment for unobserved confounders resulted in an estimated difference of 3.7 and 4.2% (HMO lower), depending on the specification of geographic-fixed effects. The latter result (4.2%) was statistically significant and consistent with prior studies that did not adjust for unobserved confounding. Our findings suggest there may be unobserved confounders associated with adverse selection in the HMO sector, which had a large effect on our mortality estimates among HMO enrollees. An important topic for further research is to identify such confounders and explore their relationship to mortality. The methods presented in this paper represent a promising approach to comparing outcomes between the HMO and FFS sectors, but further research is warranted.