Estimates of the demand for mental health services as a function of cost-sharing can be used to improve insurance design and mental health care resources planning. Nonexperimental studies of the effects of cost-sharing on mental health use are limited by unrepresentative samples and by possible self-selection into plans with more generous coverage by people who anticipate using mental health services. The Rand Health Insurance Experiment (HIE) avoided those problems by virtue of being a randomized trial that assigned health insurance those plans to participants sampled from a general population. Analysis of annual expenditures on mental health services in the HIE showed the effects of cost-sharing on each assigned plan, and of other participant characteristics, but lost information by adding together all episodes of care in each year. Also, the results by plan combine the effects of plan coinsurance, deductibles, and an upper limit on out-of-pocket spending. The proposed work extends that analysis to episodes of mental health care. It combines methods and data developed to study episoded of medical care with the extensive work done on the use of mental health services. The research will focus on care provided by mental health specialists, but a separately budgeted task describes episodes for other providers. Episodes are natural behavioral units that allow more precise estimates of the effects of insurance, and of individual characteristics, on behavior. They provide data on behavior within the year. The data can be used to separate the effects of coinsurance, deductibles, and the upper limit on spending; the resulting knowledge can be used to design better insurance and will resolve the controversy over the HIE's annual mental health results. Effects for mental health care will be compared with combination of a range of typical families and insurance policies. The descriptive analysis of mental health care episodes in a general population will yield distributions of frequency, type, drug use, and other characteristics. Multivariate regression models will be used to examine how prospectively measured characteristics affect use, such as the several components of mental health status, physical health status, health habits and attitudes, and demographic variables.
| Wells, K B; Keeler, E; Manning Jr, W G (1990) Patterns of outpatient mental health care over time: some implications for estimates of demand and for benefit design. Health Serv Res 24:773-89 |
| Keeler, E B; Manning, W G; Wells, K B (1988) The demand for episodes of mental health services. J Health Econ 7:369-92 |
