Discontinuous provider response to financial incentives for quality improvement: An evaluation of performance thresholds in nursing home pay-for-performance
Typical P4P incentives give providers financial bonuses (e.g. add-on payments) for achieving pre-specified quality goals, including achieving a target threshold (e.g. performance above a pre-determined level), a relative rank (e.g. performance in the top 10% of all providers), or an improvement-based metric (e.g. improving performance over last year’s performance). While each of these quality goals has anticipated pros and cons, empirical evidence supporting their use is scarce. Understanding the tradeoffs of these design choices may enable design of more effective P4P programs.
Our objective is to investigate empirically the effect of using performance thresholds on provider response to financial incentives. While performance thresholds are easy for payers to implement, making them common in current P4P programs, their use may have several downsides. First, they give no incentive to improve beyond the targeted threshold. Thus, a ceiling of quality improvement may be observed at the threshold, blunting the average effect of P4P on quality improvement. Second, threshold-based payments give little incentive to low-performing providers with performance far from the threshold with little chance of achieving the targeted threshold.
We test the effect of using threshold-based financial incentives in the setting of nursing home P4P in three states with P4P programs, each of which define their thresholds for bonus payment differently: Colorado (where P4P was implemented in 2009 using a pre-determined threshold), Georgia (where P4P was implemented in 2007 using the state’s performance average as the threshold), and Oklahoma (where P4P was implemented in 2007 using the state’s median as the threshold).
In prior work we showed that on average, nursing homes in states with P4P improved performance. In this paper we test whether 1) these quality improvements are concentrated among nursing homes that were close to the performance threshold; 2) the thresholds resulted in a discontinuity in quality improvement around the threshold, with performance improving to just above the threshold but not beyond; and 3) facility response differs based on the uncertainty surrounding the set threshold (where there is more uncertainty about thresholds based on a state’s average or median performance compared to pre-determined level).
The primary data source is the nursing home Minimum Data Set, which contains detailed information on 100% of nursing home care admissions to Medicare and Medicaid nursing homes in the United States, which we use to create the facility-level quality measures included in each state’s P4P program. We compare changes in performance in over 1,000 P4P nursing homes to two control groups: nursing homes in neighboring states with similar performance levels and trends prior to each state’s P4P program but no P4P program; and a synthetic control group (selected from states that did not implement nursing home P4P).
Results from this analysis will provide important and new evidence on the use of performance thresholds in P4P programs.