Incorporating patient-related sociodemographic factors into performance measurement systems is essential for providing an accurate assessment of the quality of care delivered.
Our healthcare payment system is in the midst of a great transformation from one primarily based on volume of services provided to one based on the value of those services to the individuals and communities who receive them. Naturally, this transformation is not just about payment or making care more affordable; it is also about making care safer, ensuring that all persons and families are engaged as partners, promoting effective communication and coordination of care, promoting the most effective prevention and treatment practices, and enabling healthy living for communities (ie, focusing on achieving the priorities of the National Quality Strategy).1 Ambitious goals have been set for this transformation: HHS Secretary Sylvia Burwell announced in January 2015 that 85% of all Medicare fee-for-service payments will be tied to quality or value by 2016, increasing to 90% by 2018; and 30% of Medicare payments will be tied to quality or value through alternative payment models (eg, accountable care organizations) by the end of 2016, increasing to 50% of payments by the end of 2018.2
As we work to build a healthcare system that not only delivers better care but is smarter about how dollars are spent and makes people healthier, we are identifying metrics to manage and track our progress. These metrics, which are used for public reporting and payment, are known as performance measures—as opposed to quality improvement indicators used internally by organizations to improve the quality of care they provide. Performance measures are now being used in over 2 dozen federal programs, including those that determine payments to hospitals, long-term care facilities, health plans and physicians, and pay-for-performance (or value-based payment as it is currently called), and they are becoming a standard part of commercial contracting as well.
However, health outcomes can be influenced by many factors other than the healthcare services received: patient-related factors such as existing clinical conditions and sociodemographic status (SDS) also play an important role. Avoiding incorrect conclusions or inferences about the quality of care delivered is important to consumers and patients in making informed decisions about where to obtain care; for payers, health plans, and providers regarding rewards and penalties; and for providers and plans in terms of reputation and the ability to improve care for the various subpopulations they serve. In particular, as healthcare moves toward increasing the use of financial rewards for better quality and financial penalties for substandard quality, a substantial risk of penalizing healthcare organizations and providers who serve more disadvantaged populations is present. Failing to account for the greater difficulty in achieving good outcomes in socially and economically disadvantaged populations could set up a series of adverse feedback loops that result in a downward spiral of access and quality for those populations. Therefore, it becomes increasingly important that our metrics accurately measure those aspects of quality of care under the control of the providers with as little input from sources out of their control as possible.
What other sources might confound the measurement of quality? There is a large body of evidence that various sociodemographic factors influence healthcare outcomes, and thus influence results on outcome performance measures.3-5 SDS refers to a variety of socioeconomic (eg, income, education, occupation) and demographic factors (eg, age, race, ethnicity, primary language). Also, there are disparities in health and healthcare related to some sociodemographic factors.6 Given the evidence, the overarching question is what, if anything, should be done about sociodemographic factors in relation to performance measurement, and thus performance-based payments? More specifically, the question being asked is: “How would the performance of various providers compare if, hypothetically, they all had the same mix of patients?”
Given that healthcare outcomes are a function of patient attributes (including SDS) as well as the care received, and because in the real world people are not randomly assigned to providers for healthcare services so that all providers have the same mix of patients, some type of approach is desirable for ensuring an apples-to-apples comparison when examining outcome performance in real-world settings. One approach is through risk adjustment.
What Is Risk Adjustment?
Risk adjustment is a statistical approach used to level the playing field when reporting performance measures by adjusting for differences in risk among patients. Risk adjustment takes into account patient-level characteristics that may be risk factors for the measure outcome, but are unrelated to the quality of care. Adjusting for these characteristics improves the ability to fairly compare quality measures between healthcare providers.
Risk-adjusted performance measure rates are calculated by comparing a provider’s actual (or observed) rate to their expected rate. The expected rate is determined by the case mix of the provider’s patients, using statistical modeling to take into account patient-level clinical and sociodemographic characteristics. For example, using hospital readmission rates as the outcome, a provider that serves a larger proportion of elderly, chronically ill patients will have an expected readmission rate higher than that of another provider which serves a younger, healthier population. The ratio of the provider-observed rate to the provider-expected rate is then multiplied by the overall rate from a nationally representative sample, which allows for standardization in order to compare rates across providers.
Concerns and Unintended Consequences
Why has the performance measurement world been reluctant to embrace SDS risk adjustment? The first and most important concern about adjustment for sociodemographic factors is that disadvantaged patient groups, on average, might receive worse quality of care. In other words, differences in observed performance might reflect actual differences in the processes of care for disadvantaged versus other patients that would be “adjusted away.” These and other ideas regarding the desirability of SDS risk adjustment are summarized in the
In 2014, CMS contracted the National Quality Forum (NQF) to address these issues. The NQF convened an expert panel composed of multiple stakeholders with a variety of experiences related to outcome measurement and disparities to consider if, when, and how outcome performance measures should be adjusted for SDS. The panel’s recommendations included the following8:
The panel went on to recommend a 2-year trial period, during which all measures newly introduced for NQF endorsement, as well as any measures already NQF-endorsed, could be considered for SDS risk adjustment.
What Is PQA Doing to Address SDS Risk?
In light of the NQF trial period, which began April 2015, Pharmacy Quality Alliance (PQA) convened an SDS Risk Adjustment Advisory Group, which is comprised of PQA member organizations with experience in the topic. The goal of the group is 2-fold: 1) identify which PQA measures may be appropriate for SDS risk adjustment; and 2) recommend a valid risk adjustment methodology for those measures, which includes determining which SDS variables will be used for adjustment and how to report the measure rates by plan. The group is currently focusing on the 3 medication adherence measures that are used in the CMS Star Ratings: proportion of days covered for diabetes medications, cholesterol lowering medications (statins), and blood pressure medications (renin-angiotensin system antagonists). Risk adjustment models for these 3 measures will be developed and reviewed by the group, with the goal of making a final recommendation by fall 2015.
Moving forward, measure developers are encouraged to assess newly developed and existing measures to determine if sociodemographic status risk adjustment may be appropriate. PQA will incorporate SDS risk adjustment consideration into its rigorous measure development process, and bring forward any risk-adjustment measures for NQF endorsement consideration. Multiple stakeholders will be monitoring the NQF 2-year trial period to gain an understanding of how measure developers are using SDS risk adjustment and what variables are the most important. As payers and purchasers begin to incorporate SDS risk-adjusted performance measures into public reporting or pay for performance programs, there will be a need to evaluate the impact of risk adjustment to determine if there are unintended consequences.