The effects of quality bonuses on quality of care in Medicare Advantage

Pay-for-performance (P4P), a policy reform in which payment to providers or health plans is connected to observable quality indicators, is quickly proliferating throughout the health care system. In 2012, Medicare implemented the Quality Bonus Program (QBP) in Medicare Advantage (MA), the market for private Medicare plans. The MA QBP is the first P4P program to incentivize health insurance plans, rather than health care providers. All MA plans are eligible to receive bonus payments based on an overall plan quality rating, which is calculated from a series of performance measures for clinical quality and patient satisfaction. Ratings are based on a “star” rating system, with ratings ranging between 1 (lowest quality) and 5 (highest quality) stars. Incentives in the program are very large: payouts to plans were approximately $3 billion in 2012 – around 20 times larger than the incentives in the first year of Medicare’s hospital P4P program.

In this paper we evaluate the effects of the MA QBP on plan quality. We identify the effect of the Demonstration on quality using a differences-in-differences design, comparing quality among individuals enrolled in MA plans to quality among individuals enrolled in commercial plans (CM) before and after the implementation of the Demonstration. Data from this study come from the Health Care Cost Institute database, a unique multi-payer commercial claims database. The unit of analysis will be the patient-year, and we include all individuals age 55-75 enrolled for the entire year. This results in an MA (CM) cohort that includes between 430 thousand and 4.24 million enrollees (between 1.54 million and 1.92 million enrollees), depending on the year.

We measure quality for each cohort-year using a subset of 6 measures that were used for determining bonuses in Medicare Advantage that are part of the HEDIS measure set. These measures can be calculated for beneficiaries in both MA and CM plans using the HCCI claims data. We calculate these measures at the cohort-by-year level, allowing us to observe differential changes in quality in MA vs. CM.

In addition to estimating the effect of the Demonstration on overall quality in MA, we also extend the analysis in two ways. First, we explore the possibility that the quality-boosting effects of the Demonstration spilled over from MA into CM. To determine whether spillover effects exist, we perform a robustness check where we calculate CM quality measures using only claims data from counties with less than 10% MA penetration. We will also estimate alternative specifications of our regressions, interacting our treatment variable with MA penetration. If the treatment effect is decreasing in MA penetration, we will conclude there is evidence of spillovers.

Second, we explore the mechanisms through which MA insurers may have improved quality. Specifically, we will test whether they boost quality by sending patients to higher quality providers or by convincing providers to give higher quality care. This supplementary analysis will provide rare insight into the relationship between insurers and providers and will provide policymakers and researchers with important information about how insurers influence provider behavior.