An econometric model of health care utilization and health insurance in Switzerland

Abstract

This paper presents a preliminary econometric analysis on how different alternative plans affect the utilization of health care services in Switzerland. The data come from the 1992–1993 Swiss Health Survey (SHS). We briefly describe some institutional aspects of the Swiss health system which prevailed at the time of the SHS, with particular emphasis on health insurance plans and payment system of providers. We estimate a simultaneous two equation model containing latent variables in order to compare the probability of having at least one inpatient stay given that the insured has used some medical treatment for those who have purchased only a `basic insurance' plan and those who have purchased a supplemental insurance.

Introduction

As is well known, health care utilization is conditioned by the type of insurance plan of the user. There is a vast literature on this subject which examines the effect of the moral hazard problem which has been pointed out long-time ago by Arrow (1963). This is notably the case of the remarkable RAND health insurance experiment (RHIE) (see Manning et al., 1987; Newhouse and the Insurance Experiment Group, 1993). Other studies also point out that health care utilization may be conditioned by the economic incentive of providers. In particular, several studies pointed to the role of the consumer's information and the possibility of induced demand in fee-for-service insurance plans.

The purpose of this paper is to present a preliminary investigation of the effects of alternative insurance plans on the utilization of health care services in Switzerland.¹ The econometric analysis is based on the 1992–1993 Swiss Health Survey (SHS) carried out by the Swiss Federal Statistical office (SFSO).

Generally, the institutional aspects have important implications for modeling the relation between health insurance plans and health care utilization. Section 2describes the relevant aspects of the SHS data and the institutional details regarding insurance plans at that time in Switzerland. However, it is useful to indicate at this stage that these plans were mainly separated into: `basic' insurance concerning the so-called minimal `compulsory benefits'; `semi-private' and `private' insurance, the name being given because of the hospitalization in a two-bed room and a single-bed room, respectively. But these three type of insurance plans also varied along several other dimensions: coverage of supplemental inpatient and outpatient services, coverage of allied health personnel and payment system of care providers for each procedure or service.

In a recent study, Domenighetti et al. (1996) used the SHS data to analyze the variation in the consumption of five specific surgical procedures across subgroups of the Swiss population.² The analysis in Domenighetti et al. (1996) shows that surgery rates are much lower for those who have a `basic' insurance, and thus only entitled to hospitalization in public ward, and higher for those who have either a `private' or `semi-private' insurance.

The preliminary empirical investigation presented in this paper differs from Domenighetti et al. (1996) in two main aspects. First, we use the data on the utilization of health services by participants during the 12 preceding months instead of the data on the above-mentioned surgical procedures. Second, instead of a multiple logistic regression, we estimate a simultaneous equation model with latent variables to analyze these data.

As already observed by Cameron et al. (1988), among others, it is difficult to model the joint determination of health service utilization and health insurance choice by means of a tractable structural simultaneous equation model. In particular, econometric implementation of a structural model for insurance demand is quite complex. In this paper, we follow the same approach as in Cameron et al. (1988) and consider a simultaneous two equation model which simply relates use per person during the 12 preceding months – which may be, somehow, interpreted as annual use – to insurance plan.³ We have grouped together the two types of supplemental health insurance plans (`semi-private' and `private'), and the first equation is a reduced form equation of insurance choice. The second equation is a structural equation for the propensity that someone will have at least one inpatient stay given that he has used some medical treatment, conditional on the type of insurance plan he has selected. Thus, this equation contains endogenous latent variables as well as an endogenous dummy variable.⁴

The rest of the paper is organized as follows: Section 2contains a brief description of the 1992–1993 Swiss Health Survey and of the health insurance arrangements in Switzerland at that time. Section 3presents the simultaneous two-equation model with both latent variables and a dummy endogenous variable. It also presents estimates of this system. Our main conclusion is the following: if we control for other variables, the fact that the insured person has a supplemental health insurance has a positive effect on the probability that he will have at least one inpatient stay given that he has used some medical treatment. Thus, this result suggests that the conclusion reached in Domenighetti et al. (1996) on the basis of other data is confirmed.