LATENT CLASS ANALYSIS OF DEMAND FOR MEDICAL CARE SERVICE

点击翻译按钮获取中文摘要

The objective of the research proposed here is to develop and evaluate alternative econometric models for the empirical analysis of the demand for medical care services. We propose a careful comparison and evaluation of a new modeling approach, based on the latent class (or finite mixture) model (LCM), with the currently leading paradigm for analyzing medical care services usage, the two-part model, (TPM). The TPM framework was proposed and empirically implemented by the RAND group within the context of the RAND Health Insurance Experiment (RHIE). The research objective will be implemented using the complete data from the RHIE, using both expenditure and count data measures of medical care utilization. The central focus in this research will be a comparative analysis of alternative econometric methodologies and model-building approaches, and the substantive implications for the responsiveness of medical care utilization to variations in price, income and health status variables. The LCM framework is more flexible but it is also conceptually and computationally more demanding. The advances of recent years make it feasible to reappraise the central and influential insights of the RAND study. Although the RHIE data are now old, they are nevertheless appropriate for the objective because they are generated within an experimental design in which household are assigned exogenously assigned insurance plans. There is scientific merit in a fresh analysis of the RHIE health utilization data using newer modeling tools, and reasons for expecting the proposed approach to yield new insights and to improve upon the well-established two-part modeling framework. We propose model selection criteria and cross-validation methodologies to compare the LCM and TPM frameworks with focus on the following issues: impact of insurance on demand for medical care at different levels of utilization: the classification of individuals into high- and low-use categories; prediction of average utilization of medical care services for specific patient types and hypothetical risk pools.