Health Status and Repeated Multiple Treatments in Long-Term Care: A Panel Structural Var Analysis

Abstract

This study analyzes the dynamic relationship between health status and expenditures on repeated multiple treatments, which are typical in long-term care. To facilitate causal inferences where complex dynamic interdependencies exist between many variables, we adopt a structural vector autoregression model for panel data of individuals. The model is estimated using a Bayesian shrinkage method which can simultaneously employ estimation and model selection for the lag length. Then, we employ a counterfactual analysis using impulse response functions. We analyze monthly claims data in the context of long-term care in Japan, where social insurance covers many formal services for elderly care at home. Our empirical analysis reveals several patterns of dependency between service utilization and their effects. In particular, we found that day care and outpatient rehabilitation share similar utilization patterns and also result in similar levels of improvement in health status, which implies that appropriate targeting can improve the effectiveness of service provision.

Keywords: Panel structural vector autoregression model; long-term care expenditure; long-term care insurance in Japan; high-dimensional data; Bayesian shrinkage estimation; causal inference

JEL Classification: I11; C33; J14; C11