Estimates of use and costs of behavioural health care: a comparison of standard and finite mixture models.

Estimates of health care demand are known to depend on the empirical specification used in the analysis. In this paper, an innovative specification, the finite mixture model (FMM), is employed to estimate the utilization of and expenditures on behavioural health care. Unlike standard specifications, the FMM has the ability to distinguish between distinct classes of users of behavioural health care (e.g. the 'worried well' and the severely mentally ill). This new model is tested against standard empirical specifications using data from the National Medical Expenditure Survey. Using common risk stratifiers, estimates of utilization and costs are generated with each specification. It is found that the FMM provides a much better fit of both expenditure and utilization data than standard specifications, particularly among high intensity users that standard models have been unable to represent adequately. Furthermore, the results provide preliminary evidence that there are (at least) two distinct groups of users of behavioural health care. The empirical advantages of the FMM translate into superior estimates of mean costs and utilization that have widespread application in rate-setting exercises. Copyright 2000 John Wiley & Sons, Ltd.