【導(dǎo)讀】behalf.failure.sector?insurance,cars,hi-fiequipmentetc.managingdirector.havewithdoctors.

　　

【正文】 in dp rivp rivp riv qqq ,0,0,0 ?? . Following the reform we now have in dG M SG M SG M S qqq ,1,1,1 ?? and in dp rivp rivp riv qqq 1,1,1 ?? . The difference in total utilisation between GMS and private patients before the reform will be )( ,0,0 p rivGMS qq ? . The difference after the reform will be )( ,1,1 privGMS qq ? . The difference in the differences can be expressed as: )()( ,1,1,0,0 p r ivG MSp r ivG MS qqqq ??? i n dp r i vp r i vi n dG M SG M Si n dp r i vp r i vi n dG M SG M S qqqqqqqq ,1,1,1,1,0,0,0,0 ???????? ).()()()( ,0,1,1,0,1,0,1,0 i n dp r i vi n dp r i vi n dG M Si n dG M Sp r i vp r i vG M SG M S qqqqqqqq ???????? On the basis of the model outlined above, plus a number of reasonable assumptions, it should be possible to sign the above expression. Take the first two terms on the righthandside of the above equation, )( ,1,0 G MSG MS qq ? and )( ,1,0 privpriv qq ? . These terms represent the change in GP visits before and after the reform initiated by the patient for GMS patients and private patients. If we assume what is what is known as the “mon macroeconomic effect” then there should be no difference in the growth MB, MC Induced visits, GMS fee MC Private Private fee MC GMS *indpriv ? 0?indGMSQ rates for patient initiated visits for GMS and private patients. Thus overall these terms should be zero. Turning now to the latter two terms, )( ,1,0 indGMSindGMS qq ? and )( ,0,1 indprivindpriv qq ? , our model outlined above predicts that the first of these terms will be unambiguously positive. The change in reimbursement will drive induced visits for GMS patients to zero in period 1. The model is less clearcut regarding what will happen the second term. However, as discussed above, the strong likelihood is that it will be nonnegative . induced visits for private patients will not fall from period 0 to period 1 and may well rise. Summary statistics for Ireland show: GMS Private All 1987 1995 Thus differenceindifferences proportionately is ? = . Hence evidence shows that following change in reimbursement visits from GMS and private patients fell – thus no real evidence of SID. Small Area Variations Above analysis has assumed that doctors have “perfect” information and patients have inferior information – but doctors may not have perfect information either. Treatment rates for different conditions show wide geographical variation, even when controlling for population characteristics . diagram below 1000 100 Hysterectomy Tonsilectomy Diagram shows variation in rates per 100000 of population at risk for two procedures for three areas ( from left to right New England, Norway and West Midlands, England). Thus average rate of hysterectomy in New England is five times that in Norway. Why this degree of variation? Is it evidence of unnecessary or inappropriate care? Since very little of variation is explained by population characteristics there is a possibility that SID accounts for it. One possible reason is degree of uncertainty regarding certain conditions – thus wide range of practice styles are with accepted bounds. Suppose the “true” relationship between the degree of medical care devoted to a condition and the eventual health oute is given by S*. Doctors do not know this true relationship – instead we have two different “styles”, S1 and S2. S2 has strong belief in efficacy of medical intervention while S1 is more sceptical. If doctors of styles 1 and 2 are not distributed randomly across population then rates of medical intervention will vary. Possibility of local “bandwagon” effects especially if all doctors attended same medical school. Improvements in medical knowledge will eventually lead to S1 and S2 converging to S*. Does SAV have important welfare implications? Medical Care Health Status S2 S* S1 Suppose the true marginal benefit curve for a given treatment is MB* with optimal rate of intervention of R*. If there is underutilisation then perceived MB curve is MB1 giving a rate of utilisation of R1. This involves a loss of consumers’ surplus equivalent to triangle A. If perceived MB curve is MB2, then there is overutilisation at rate R2 and a loss of surplus of B. Suppose the “true” MB curve has slope of S and there are N munities across which intervention rates vary and let 2? represent the variance of use around the optimal use. Then it can be shown that the loss of consumer surplus is given by ?SN . MC MB1 MB* MB2 R1 R* R2 B A MB, MC Rate of intervention