To Design Equitable Value-Based Payment Systems, We Must Adjust for Social Risk
This article originally appeared in Health Affairs on September 17, 2020, and was co-authored by Philip M. Alberti, senior director, health equity research and policy at the Association of American Medical College, and David R. Nerenz, PhD, director emeritus of the Center for Health Policy and Health Services Research at Henry Ford Health System.
In its recently released report to Congress, the Department of Health and Human Services Assistant Secretary for Planning and Evaluation (ASPE) concludes that many quality measures used in value-based payment programs—process and outcome measures specifically—should not be adjusted for the social risk of patient populations. Social risk factors exist at both individual and community levels and include challenges such as poverty, food insecurity, housing instability, homelessness, and lack of social support, all of which can affect health outcomes. The ASPE claims that social risk adjustment could mask unspecified inequities or excuse poor quality of care. But the report does not sufficiently articulate or support this claim. In this post, we clarify the distinction between health care quality and outcomes, show how differences in income and social risk across patient populations can affect providers’ quality ratings in value-based payment systems, and explain why risk adjustment for social needs helps advance quality and health equity.
Providers’ Performance on Quality Measures Can Be Affected by Differences in Patient Populations
The ASPE’s recommendation against adjusting outcome measures for social risk is based on the concern that adjustment would mask or excuse poor-quality care. We feel that this concern is misguided. First, it is crucial to understand that “outcome” and “quality” are not the same thing. They are related, but they may be only loosely related. Quality is inherent in the services provided or not provided to patients. Outcomes may reflect quality indirectly, particularly if the outcome assessment is separated in both time and place from the relevant care processes so that other factors emerge that have independent effects on the outcome. Inequities in outcomes are not necessarily inequities in quality of care. To obtain an accurate measure of quality, it is important to account for all patient factors that may affect outcomes through causal pathways other than quality of care. Variables such as age, sex, and chronic conditions have long been used in risk adjustment without question. There is now considerable evidence that social risk factors likewise can affect outcomes, apart from quality of care; thus, these factors should be included in risk adjustment.
A Tale of Two Hospitals
Imagine two hospitals—both have a 12 percent 30-day readmission rate following acute myocardial infarction for people with high incomes and a 13 percent readmission rate for people with low incomes. Imagine that the differences are primarily related to post-discharge events such as whether patients return to stable housing with access to healthy food, are able to understand and follow discharge instructions, and have social supports or community social services available.
Are the two hospitals in this example different in quality? No, their clinical care processes are equal for low-income and high-income populations, and both hospitals spend extra time and effort on patients with social needs. The different readmission rates relate to the social environments to which patients return after discharge.
Now, let’s imagine that the two hospitals differ only in their income case-mix. Ninety percent of Hospital A’s patients are wealthy, and 90 percent of Hospital B’s patients are low-income. The age, sex, and clinical characteristics of the patient populations are identical. Let’s calculate the overall readmission rate for each hospital based on the different case-mix: Hospital A’s rate is 12.1 percent; Hospital B’s rate is 12.9 percent.
Are the hospitals different in quality now? No—nothing changed from the initial example above; we just applied the effect of case-mix. To those paying attention to quality ratings, such as consumers and payers, however, the hospitals appear to be different now. That difference is not due to variation in quality but rather to variation in income case-mix. The difference in overall readmission rates is due to the added burden of social risk factors on the low-income population that result in worse outcomes for those patients, even when they receive the same quality of care. The impact of social risk on health outcomes has been well-documented by several major studies.
Risk Adjustment Helps Achieve Accuracy and Fairness in Value-Based Payment
There are several valid statistical methods to adjust for differences in case-mix. If the 0.8 percent difference in readmission rates between the two hospitals in our example would result in a bonus payment or financial penalties for one and not the other, is it fair and right to do the statistical adjustment to “remove” the effect of income case-mix differences?
We believe it is, and that adjustment is the only way to accurately reflect the equal levels of quality at each of the two hospitals. Without adjustment, a value-based payment program based on readmission rates would reward the hospital with more wealthy patients and penalize the one with more low-income patients, even if both hospitals achieved the same outcomes for both patient groups. It is possible that the hospital treating mostly low-income patients may provide lower-than-average quality of care to those patients, but that lower quality could be masked without risk adjustment. It is also crucial to note that validated risk-adjustment methods exist that do not “adjust away” real quality differences among entities being compared.
Social Risk Adjustment Does Not Remove the Incentive for Improvement
Hospital motivations for improvement, if linked to payment incentives, depend more on the design of the incentive payment system than on the quality measures used. For example, in a payment system linking a reward to deciles of performance in a linear fashion, a hospital in the ninth decile would have the same incentive to improve as one in the seventh decile. Social risk adjustment that might move a hospital from ninth to seventh decile would not change the incentive.
On the other hand, in a system with a single cut-off point—such as a financial penalty for hospitals below the twenty-fifth percentile on a measure but no penalty for those above the twenty-fifth percentile—adjustment for social risk (or other factors) may reduce the incentive to improve for hospitals right at the cut-off point that would receive a penalty without adjustment and avoid a penalty with adjustment. In such a system, it should be noted that all hospitals above the twenty-fifth percentile presumably would have no incentive for improvement, and hospitals far below the twenty-fifth percentile may view their situation as hopeless and therefore lack motivation to improve. Even with adjustment, these hospitals could be ranked as providing poor-quality care. Under such a payment system, adjustment for social risk would have a trivial effect on incentives for improvement.
Social Risk Adjustment Is Needed to Identify High-Performing Providers
Hospitals that serve a high volume of low-income patients but have good outcomes may provide outstanding quality care, but without adjustment for case-mix they may appear to be just average or even below average. With adjustment, they are much more likely to be identified as outstanding. Only if we can identify high-performing organizations across the full range of patients and communities served can we learn best practices and help others advance on the quality curve.
Social risk adjustment is critical to the design of effective value-based payment systems. Similar to common adjustments for age and comorbidities, adjustment for social risk is needed to avoid unfairly penalizing hospitals, health plans, physician practices, and other providers for patient and community characteristics beyond their control. Social risk adjustment will not mask poor quality of care nor will it disincentivize quality improvement. Health plans and hospitals serving large numbers of dual-eligible patients or other groups with one or more social risk factors will not be held to a lower quality standard in a system with social risk adjustment.
Value-based payment systems that do not adjust for social factors perpetuate health injustice and systematically increase inequities. Unfair penalties for providers caring for the sickest and most marginalized patients strip resources away from the very communities that need them most. The absence of social risk adjustment does nothing to address racial inequities in health and health care; if anything, it makes the disparities worse.
Given the growing body of evidence linking social risk factors with poor health outcomes, and the real-world consequences of value-based payment systems that do not account for these factors, we disagree with the ASPE report’s logic and its recommendation against social risk adjustment. We urge measure developers and other stakeholders to develop models that include fair, statistically valid, and logical social risk adjustment methodologies for value-based payment and performance ranking systems. These models will play a critical role in reducing disparities, advancing health equity, and supporting quality improvement.