Why It Matters
Because health and health care need improvement.
Processing ...

To Risk-Adjust or Not to Risk-Adjust? Yes

By John Gauthier | Monday, May 5, 2014

Amy Reid is a Senior Research Associate on the Results and Evaluation team at IHI and Don Goldmann, MD, is IHI’s Chief Scientific and Medical Officer. Reid and Goldmann are members of IHI’s Diversity and Inclusion Council and are passionate about health equity.

Reid Goflmann

In the United States, the National Quality Forum (NQF) is considering when and how to use risk adjustment for sociodemographic factors to assess quality of care, while balancing the need to make health inequities visible and hold the health care system accountable for addressing disparities. This is one of the most important debates in today’s pay-for-performance policy landscape. It is a fundamental issue that gets to the heart of fair benchmarking, public reporting, and policy in a country that has profoundly different outcomes depending on whether one is poor, comes from an underserved racial or ethnic minority, or is otherwise disadvantaged.

Risk adjustment is a statistical method that allows for comparison of outcomes when patient populations differ. For example, to compare readmissions between health facilities that serve populations with different risk profiles, risk adjustment could be used to control for socioeconomic status, insurance status, diagnosis, and other sociodemographic factors that might be associated with the outcome being measured. It’s like saying, “If these facilities served similar populations, do their readmissions rates still differ?”

Abundant data suggest that basing payment on unadjusted outcomes penalizes organizations that care for underserved populations including the poor, racial and ethnic minorities, uninsured and underinsured patients, and others who have poorer outcomes due, in part, to inequitable socioeconomic conditions. By tying payment to these poorer health outcomes, these organizations may not get the resources needed to continue to care for their high-risk populations. It seems fair, just, and reasonable to account for sociodemographic factors in comparing their performance to organizations serving populations with lower-risk profiles, as the “performance” is impacted by factors outside of the quality of care delivered. America’s Essential Hospitals and other organizations concerned about the future of safety net hospitals support adjusted outcomes.

LEARN MORE: Leading Population Health Transformation, Feb. 22-24, 2017, in San Diego, California

The concern is that inequity is hidden by such adjustments. It is critical that we continue to call out disparities, not bury them in complex statistical risk-adjustment models. This is why many equity advocates are concerned about risk adjustment, even though they know that organizations that care for the populations they are most passionate about start at a disadvantage. Historically, these concerns are reasonable and have merit. For example, experts we convened at IHI to discuss measuring mortality as a potential quality metric agreed that disparities should not be masked by presenting only risk-adjusted data. The success of our health system depends on identifying gaps experienced by disadvantaged populations, and tracking whether or not efforts to close these gaps are successful.
Despite the controversy, we believe that the path forward is clear: we must require both. If we use adjusted data for quality assessment, and merely suggest that organizations use unadjusted, stratified data to understand inequities, there is a danger that such measures will fall by the wayside. Some argue that requiring both adjusted and unadjusted data is too complicated. We disagree, and offer the following path forward. First, the population should be stratified by sociodemographic factors so disparities can be assessed. It is important to remember, for example, that African Americans may have worse outcomes than whites whether they are cared for in a safety net hospital or in a for-profit chain. If we cannot see disparities, we cannot address them. Second, when reporting data for pay-for-performance, public reporting, and benchmarking, outcomes generally should be risk adjusted. Both sets of data are critical, and both should be displayed publicly and prominently. This is not an either/or issue – it is a both solution.

first last

Average Content Rating
(0 user)
Please login to rate or comment on this content.
User Comments