Do Medicare HMOs still reduce health services use after controlling for selection bias?

This study models the relationship between Medicare beneficiary decisions to join Medicare HMOs and subsequent health services utilization. The relationship between health plan choice and utilization is thought to be endogenous because of favorable selection into HMOs. Previous studies found significantly lower inpatient utilization among Medicare HMO enrollees than among nonenrollees, but lacked strong controls for selection bias. Thus, a firm conclusion could not be drawn as to whether the observed differences were attributable to the HMO practice setting or to baseline differences in the illness profiles of the two groups studied. The present study uses simultaneous equations methods, including discrete factor estimation, to test the effect of Medicare HMOs on utilization when strong controls for selection bias are imposed. The model was run on a panel of 1993-1996 data from the Medicare Current Beneficiary Survey, supplemented with linked data on Medicare HMO characteristics and area supply characteristics. The study found that even when favorable selection is controlled for, Medicare HMOs significantly reduce both the probability of hospitalization and the number of inpatient days used by those who are hospitalized. Medicare HMOs do not, however, appear to reduce the use of physician services.