Evaluating a series of legislatively determined payment rate changes to MA plans, I found a negative relationship between MA payment rates and rates of ACS hospitalization rates among Medicare beneficiaries. Simulations based on these results suggest that reductions in MA payment rates may result in a small increase in ACS admission rates. Although there is considerable policy debate surrounding additional payments to MA plans relative to FFS spending, this study is the first to provide estimates of the potential effects of payment cuts on care provided to Medicare beneficiaries. Largely due to limited data availability, relatively little empirical work has been done in this area.
I did not find evidence that rates of elective procedures were related to MA payment rates, though restricting access to expensive procedures would be one way for plans to limit the cost of care. This finding is consistent with earlier work indicating that managed care plans achieve cost savings by paying lower prices rather than reducing services (Cutler, McClellan, and Newhouse 2000
). Rates of elective procedures will also remain stable if sicker patients remain in or disenroll to FFS Medicare to avoid HMO restrictions.
I focused on the effect of payment rates on hospitalizations for all beneficiaries rather than just MA enrollees because changes in payment rates will change the composition of the MA enrollee population as well as the benefits they receive. Spending on hospitalizations in FFS will be influenced by the resulting change in beneficiary demographics as well as any changes from practice patterns. This analysis is particularly relevant to the PPACA provisions, which are projected to achieve savings by reducing the number of beneficiaries enrolled in MA plans as well as the per-enrollee payments for those who remain. A drawback to this approach is that I cannot separately account for changes in hospitalizations driven by selective enrollment in response to payment cuts and those related to reductions in benefits for MA enrollees.
Comparisons of high- and low MA penetration counties offered a potential way to isolate the effect of MA payment changes on MA enrollees, though I generally failed to find a significant effect of payment changes on hospitalization rates in these counties. The limited influence of MA on practice patterns in low-penetration counties is consistent with the spillover and market share literature. Additionally, county-level hospitalization rates are weighted averages of the MA and FFS rates, so the modest changes in outcomes for the MA patients are likely dwarfed by the lack of change in FFS hospitalizations; for example, a 25 percent reduction in the rate of hospitalizations among MA enrollees would not significantly change the average rate of ACS admissions in the average low-penetration county. Separate regressions for MA and FFS enrollees provide descriptive evidence suggesting that the additional hospitalizations occur among MA enrollees. This likely reflects plans providing less generous coverage when payment rates are lower (Gold et al. 2004
Study findings should not be interpreted causally if other unmeasured factors drive both payment rates and hospitalization rates. Several analytic steps were taken to avoid this potential endogeneity. Legislatively determined MA payment rates create a quasiexperimental setting in which to consider the relationship between payments and quality. Use of county fixed effects controlled for persistent county-level differences in payment rates, beneficiary acuity, and propensity to use hospital care. Finally, I used standardized payment rates for an average enrollee rather than payments determined by the county's enrollee characteristics.
Like many other studies, interpretation of this research is limited by a lack of data on utilization by MA enrollees. Consequently, I focused on three states with superior data availability. While large numbers of Medicare beneficiaries, including many MA enrollees, reside in these states, results may not generalize to all states, particularly more rural areas.
Because my data ended before the introduction of Part D, MA enrollment was the only way for many seniors to access prescription drug coverage during the study period. Consequently, findings may overstate the relationship between payment rates and ACS hospitalizations if plans now use additional payments to provide benefits that are less effective than drug coverage at preventing hospitalizations. While MA enrollment was dominated by HMOs during the study period, the program currently includes many Private FFS and Preferred Provider Organizations. These types of plans typically do less care managing, which would reduce the potential for spillover effects to FFS, again suggesting that results in this paper represent an upper bound estimate of the relationship between MA payment changes and rates of ACS admissions. During the study period, large increases in payment rates were more common than large decreases. The simulation results assume that the response to a payment cut and payment increase are similar in magnitude but move in opposite directions.
It is important to note that quality and access are multidimensional constructs, and the rates of hospitalization included in this study only touch on some measures. This is unavoidable given current data availability. As PPACA changes are implemented, additional research will be needed to assess the implications of payment cuts for Medicare quality and access. Improved access to MA encounter data can aid these evaluations.