EFFECTS OF COST SHARING ON PHYSICIAN UTILIZATION ame time. Given that the differences between the the peculiar interactions of demand and supply in low-income WOPI and the active population are health care. Belgiur bundant supply of much larger than the differences between the low- doctors, especially GPs(one of the highest densi- income WoPI and the WoPI, we feel safe ties in Europe) who earn, on average, relatively to reject the assumptions of the differences low incomes by Western European standards and model for the former group. Moreover, as the all through fee-for-service. Experience in other electio ountries suggests that such a combination would group from among the socially weakest, constitute ideal circumstances for a high degree are good reasons to believe that they are of supplier-induced demand. And indeed, physi- an adequate control group for the other w cian utilization rates are very high, with one eithe of the highest rates of doctor (again primar lly GP)consultation rates in Europe, and Some further test results of GP visits patients home in the world. That this combina Formal F tests of cross-equation equality restric. tion has, nevertheless, not led to unmatched levels tions on parameter estimates between user groups observers. care expenditure has puzzled many and between types of care confirm what is appar- One of the reasons which is sometimes offered ent from Tables I and 2. We mentioned already hat both models clearly show that the healthier for Belgiums relative success in terms of health care cost containment is its extent of patient cost active population is substantially more price- sharing. To a much higher degree than in most responsive than the wOPI population(hypothesis other European countries, Belgian patients have 2).Indeed, we feel that this result strongly sug- to pay a fairly large share out-of -pocket wher gests that the assumptions of the differences model are not acceptable. To test hypotheses 3(a) they want to see a doctor: except for some exempt and(b),we therefore concentrate on the levels about 60-70% of the total fee is reimbursed by The utilization rate is more price elastic for GP the sickness funds. These coinsurance rates are home visits than for gP office visits The latter much higher than those customarily employed in in turn more elastic-or rather, less inelastic other European countries, and comparable with the rates of coinsurance used in, for instance the In fact, the same pattern is also found for the Rand HIE. The interesting question that arises active population in the differences model. It is and which the hiE was not designed to answer- ot very meaningful to test the hypothesis for the is: to what extent is any price-sensitivity of health WOPI in the differences model because none of care users to these levels of coinsurance diluted by the estimated price elasticities is significantly dil- under pressure and fee-for-service remuneration? ferent from zero for that group The natural experiment that occurred in Bel- um in 1994 as a result of a drastic increase in CONCLUSION the(already fairly high) co-payment rates of the public health insurance system, offered an inter- esting opportunity to try and answer this ques Certain features of the Belgian health insurance tion. We used 10 years of data from the 31 system make it an interesting test case for some of regional offices of the largest Belgian sicknes Table 3. F-test results of cross-equation restrictions in levels model H F-test Findings 3a Vk: 2k 4.45 R Higher price-sensitivity for GP office visits than for specialist visits 3b Vk:1k 48.89 R Higher price-sensitivity for GP home visits than for GP office visits Critical value of 5% level is 2.6: R= null hypothesis rejected Copyright a 2001 John Wiley Sons, Ltd Health Econ.10:457-471(2001)EFFECTS OF COST SHARING ON PHYSICIAN UTILIZATION 469 same time. Given that the differences between the low-income WOPI and the active population are much larger than the differences between the low￾income WOPI and the WOPI, we feel safe to reject the assumptions of the differences model for the former group. Moreover, as the low-income WOPI are really a very selective group from among the socially weakest, there are good reasons to believe that they are not an adequate control group for the other WOPI either. Some further test results Formal F tests of cross-equation equality restric￾tions on parameter estimates between user groups and between types of care confirm what is appar￾ent from Tables 1 and 2. We mentioned already that both models clearly show that the healthier active population is substantially more price￾responsive than the WOPI population (hypothesis 2). Indeed, we feel that this result strongly sug￾gests that the assumptions of the differences model are not acceptable. To test hypotheses 3(a) and (b), we therefore concentrate on the levels model. The results are summarized in Table 3. The utilization rate is more price elastic for GP home visits than for GP office visits. The latter is in turn more elastic—or rather, less inelastic— than the utilization rate for specialist office visits. In fact, the same pattern is also found for the active population in the differences model. It is not very meaningful to test the hypothesis for the WOPI in the differences model because none of the estimated price elasticities is significantly dif￾ferent from zero for that group. CONCLUSION Certain features of the Belgian health insurance system make it an interesting test case for some of the peculiar interactions of demand and supply in health care. Belgium has an abundant supply of doctors, especially GPs (one of the highest densi￾ties in Europe) who earn, on average, relatively low incomes by Western European standards and all through fee-for-service. Experience in other countries suggests that such a combination would constitute ideal circumstances for a high degree of supplier-induced demand. And indeed, physi￾cian utilization rates are very high, with one of the highest rates of doctor (again primar￾ily GP) consultation rates in Europe, and probably the highest rate of GP visits at the patient’s home in the world. That this combina￾tion has, nevertheless, not led to unmatched levels of health care expenditure has puzzled many observers. One of the reasons which is sometimes offered for Belgium’s relative success in terms of health care cost containment is its extent of patient cost sharing. To a much higher degree than in most other European countries, Belgian patients have to pay a fairly large share out-of-pocket when they want to see a doctor: except for some exempt categories of low-income elderly and needy, only about 60–70% of the total fee is reimbursed by the sickness funds. These coinsurance rates are much higher than those customarily employed in other European countries, and comparable with the rates of coinsurance used in, for instance, the Rand HIE. The interesting question that arises— and which the HIE was not designed to answer— is: to what extent is any price-sensitivity of health care users to these levels of coinsurance diluted by supplier reactions in a situation of doctor incomes under pressure and fee-for-service remuneration? The natural experiment that occurred in Bel￾gium in 1994 as a result of a drastic increase in the (already fairly high) co-payment rates of the public health insurance system, offered an inter￾esting opportunity to try and answer this ques￾tion. We used 10 years of data from the 31 regional offices of the largest Belgian sickness Table 3. F-test results of cross-equation restrictions in levels model H Findings 0 F-test 3a 4.45 Higher price-sensitivity for GP office visits than for specialist visits
k:
2k=
3k R
k:
Higher price-sensitivity for GP home visits than for GP office visits 3b 48.89 1k=
2k R Critical value of F-test at 5% level is 2.6; R=null hypothesis rejected. Copyright © 2001 John Wiley & Sons, Ltd. Health Econ. 10: 457–471 (2001)