Balanced Billing in the Belgian Health Care System: How to Predict the Financial Burden for Certain Groups of Patients?

Le résumé fourni par la source

Objectives: While policymakers already debated a lot on co-payments, there is an increasing concern about the huge increase of supplements as well. In this study we investigate the distribution of the burden of supplements. We want to describe the characteristics of the payer of supplements. Methodology: We use an administrative dataset of the different sickness funds with annual data for 2003 of a representative sample of about 300,000 individuals. We define supplements as the difference between the total amount paid by the patient and the official price (reimbursements from the health insurance co-payments); or as the total amount itself if there are no reimbursements. We estimate different specifications of the model. First, we start with a benchmark: OLS of the expenditures on several independent variables. Two-part models with logistic specification in the first part and different specifications in the second part (OLS, OLS on log expenditures and gamma with log specification) are estimated as well. Results: In general, the explanatory power of our models is rather low. Based on the OLS regression, we see a clear regional pattern in the total yearly amount of supplements paid. Older people and patients in their last year of life pay more supplements. People that are handicapped, disabled or chronically ill pay higher supplements as well. Patients that have preferential treatment (i.e. they have protection from the government for their co-payments) pay fewer supplements. In the two step model, we can see whether the reason for these patterns comes from the incidence of paying supplements or from the height of the supplements to be paid. We see that e.g. those with preferential treatment have an equal chance of paying supplements compared to the others, whereas the second part (in each specification) reveals that the absolute amount of supplements are lower. While the mean of OLS regression for supplements is - 60.77, the predicted mean of the two part model varies between 76.20 and 627.15 depending on the chosen specification. Conclusions: While OLS is a simple model giving a good overview for policymakers, the two-part model is more complicated and gives more details on the observed pattern of supplements. We can separate out effects on the probability to pay supplements versus the amount to be paid. It is clear that the patterns in the different specifications tell us more or less the same. More of concern is the fact that the estimated values of the mean of supplements vary a lot among the models. To be done: marginal effects will be calculated for each of the proposed models. Moreover, as the patients in the upper right tail of the distribution are the ones that are most likely to have financial problems, it is very important to see which model can best predict estimated expenditures in these regions.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.