HMO COST PERFORMANCE: A SIMULTANEOUS EQUATIONS APPROACH

过去的研究表明，健康维护组织（hmo）已经

Past research suggests that Health Maintenance Organizations (HMOs) have reduced per-capita medical expenditures in a variety of settings. Efforts to uncover the source and extent of these savings have focussed on the role of bias selection in explaining any differences between the HMOs and their fee-for-service (FFS) counterparts. Biased selection occurs when individuals with systematically lower expenditures enroll in HMOs. Its impact on medical expenditures depends in large part on the factors that influence a consumer's health plan decisions, but little is known about the manner in which these decisions are made. Beginning in 1988, military health care beneficiaries in two states were given the option of enrolling in an HMO or staying with a standard FFS plan. Data were collected on utilization, expenditures, and socioeconomic factors for a random sample of the eligible beneficiaries, as well as a control group. This demonstration project provides a social experiment ideally suited for investigating the role of bias selection. This proposal outlines an econometric model of consumer behavior in h the enrollment decision depends in part on the mean and variance of expected cost differences. Conditional on this choice, total expenditures are observed. Using maximum likelihood techniques, a consistent estimate of the cost savings can be obtained. The resulting estimates identify the determinants of individual health plan choice, and the extent of any cost savings adjusting for bias selection. In particular, they yield estimates of the responsiveness of HMO enrollment to expected cost savings and present health status. The following policy simulations are suggested: 1) what impact does a subsidy of X dollars have on HMO enrollment and per-capita expenditures? 2) how does HMO enrollment vary if the per-visit copayment is doubled? 3) what cost savings can be expected if the entire sample population were enrolled in the HMO? The results of these simulations should have broad implications for health care policy.