3

Methods to Account for Social Risk Factors

This chapter returns to the methods of accounting for social risk factors and discusses in more detail how the various methods, individually and together, might be leveraged to help the Centers for Medicare & Medicaid Services (CMS) attain its goals for value-based payment (VBP) relative to the status quo, which generally does not account for social risk factors. The committee’s approach takes as a point of departure the aim to achieve four policy goals:

1. Reducing disparities in access, quality, and outcomes;
2. Quality improvement and efficient care delivery for all patients;
3. Fair and accurate reporting; and
4. Compensating providers fairly.

Arguments in support of accounting for social risk factors in VBP are frequently framed in terms of fairness to providers in performance measurement and payment. Although these are direct policy goals of the 10 methods to account for social risk factors the committee identified in its third report, they are but intermediary goals and means of achieving what the committee views as the indirect, but principal, goals of any approach to accounting for social risk factors—reducing disparities in access, quality, and outcomes and improving quality and efficient care delivery for all patients. In other words, this series of reports aims to inform health care payers and administrators. However, the committee recognized throughout their deliberations that accounting for social risk factors is a technical exercise that can influence the lives of real Medicare beneficiaries, especially those with social risk factors.

Any approach to accounting for social risk factors (including the status quo) will achieve important policy goals but could also have unintended consequences. The committee identified five categories of potential unintended consequences:

1. Avoiding patients with social risk factors
2. Reducing incentives to improve the quality of care for patients or enrollees with social risk factors
3. Underpayment to providers disproportionately serving socially at-risk populations
4. Negative symbolic value: Perceptions of different standards for different populations
5. Obscuring disparities

This chapter begins with a review of the methods CMS could use to account for social risk factors. The committee reviewed literature on a range of methods to account for social risk factors in quality measurement and payment, with the aim to be more inclusive. In its third report (NASEM, 2016a), the committee identified 10 methods in four categories that could be used individually or in combination to account for social risk factors. These categories are

1. Stratified public reporting;
2. Adjustment of performance measure scores;
3. Direct adjustment of payment; and
4. Restructuring payment incentive design.

Three categories of methods—stratified public reporting, adjustment of performance scores, and direct adjustments of payment—build on the existing payment system. Only restructuring incentive design presents entirely new approaches.

In this report, the committee provides specific examples and methods for accounting for social risk factors. In doing so, the committee underscores the differences in the goals of each approach that guide their applications. For example, the goal of stratified public reporting is to allow a decision maker (e.g., patient) to observe and act on differences in performance for different types of patients. In this context, CMS will have to consider both the reliability of stratified estimates along with what strata are meaningful to patients in order to support informed decision making. Adjustment of performance measure scores, like stratified reporting, affects what patients observe about the performance of a provider or health plan and CMS. On their own (that is, without stratified data), adjusted scores by definition send a single performance signal that accounts for differences

in the mix of patients served but does not make disparities apparent. In contrast, adjusting payment algorithms (through either the third or fourth category of methods) is intended to alter the incentives for the plan or provider directly. The reliability of those measures will affect the balance of incentives and risk inherent in the payment formula: noisy measures impose risk and diminish the returns to improvement efforts.

In its third report (NASEM, 2016a), the committee identified certain advantages and disadvantages for each of the 10 methods (see Table C4-1 in Appendix C). In this report, the committee aligns the advantages with its four policy goals and the disadvantages with the five unintended consequences. In addition, the committee describes how the four categories of methods could be used to achieve the four policy goals, as well as the potential unintended consequences that could result from any method. The trade-offs between policy goals attained and potential unintended consequences should be carefully considered, and, as the committee concluded, any approach should seek to minimize potential unintended consequences, especially those to patients with social risk factors (NASEM, 2016a).¹ Thus, the chapter also discusses how CMS could mitigate potential unintended consequences. Finally, any approach to accounting for social risk factors will interact with the underlying incentive design to achieve certain policy goals or produce certain adverse consequences. As the committee concluded, strategies to account for social risk factors for measures of cost and efficiency may differ from strategies for quality measurement, because observed lower resource use may reflect unmet need rather than the absence of waste, and thus lower cost is not always better, while higher quality is always better (NASEM, 2016a).² Thus, the committee believes that the policy goals achieved and negative consequences that any approach to accounting for social risk factors produces need to be considered for each specific VBP program. To illustrate how CMS might select and combine methods and the benefits and harms that may result from those combinations, the committee presents an example, the Hospital Readmissions Reduction Program (HRRP). The categories of methods, the committee’s policy goals, and the potential unintended consequences are described comprehensively in Appendix C.

A. STRATIFIED PUBLIC REPORTING

Public reporting aims to make quality of care and outcomes visible to consumers, providers, payers, and regulators (IOM, 2007). Provision of quality information to these stakeholders can lead to quality improve-

___________________

¹ See Conclusion 4 in the committee’s third report (NASEM, 2016a).

² See Conclusion 7 in the committee’s third report (NASEM, 2016a).

ment for all beneficiaries through reputational incentives and by increasing market share (i.e., influencing beneficiaries’ choice of provider or plan) for reporting units (e.g., the provider, plan, physician group) with higher performance, especially when linked to behavioral nudges (IOM, 2007). Stratified public reporting provides this information for specific subgroups. To account for social risk factors in Medicare payment, provider and plan performance could be stratified by reporting unit characteristics (e.g., safety-net hospitals, minority-serving institutions, providers with a similar share of patients with social risk factors) or by patient characteristics within reporting units (i.e., by social risk factor) (Casalino et al., 2007; Martino et al., 2013; MedPAC, 2013; NQF, 2014; Price et al., 2015). Such stratified public reporting could lead not only to quality improvement (through similar mechanisms as for publicly reported overall performance), but also to disparities reduction. Just as overall performance can lead to quality improvement for all beneficiaries, publicly reported performance scores stratified by social risk factors could influence beneficiaries’ choice of provider or plan by allowing patients with social risk factors to see which providers or plans provide the best care for patients like them. Similarly, such stratified reporting gives providers and plans reputational incentives to reduce disparities within their organizations. Stratified public reporting by patient characteristics within reporting units could also increase fairness and accuracy in public reporting by allowing “apples to apples” comparison—for example, by comparing performance for low-income patients across different hospitals. Similarly, stratification by reporting unit characteristics allows comparisons to be made within peer groups (e.g., among safety-net providers), who may have differing abilities to achieve targets or improve performance owing to differences in the social composition of their patients and to resource constraints.

Comparisons based on stratified data could also help lower-performing providers and plans to identify top-performing peers and learn from their practices to reduce disparities and improve quality and efficiency. Because public reporting with stratification by patient characteristics is the only method that presents information on subpopulations and can therefore highlight any disparities that may exist, it is also the only category of methods that would allow CMS to monitor disparities. Thus, if monitoring disparities is an important policy goal, any approach to account for social risk factors must include public reporting stratified by patient characteristics within reporting units.

Although stratified reporting can help achieve the goals of reducing disparities for patients with social risk factors, quality improvement and efficient care delivery for all patients, and fair and accurate public reporting, it does not influence provider compensation. Additionally, a potential unintended consequence of the method is that stratification could create

the perception that different patients are entitled to different standards of care. This risk is expected to be minimal when stratifying by patient characteristics within reporting units, because comparisons typically compare performance across different strata (e.g., performance for black patients compared to whites), which tends to highlight disparities. The risk may be slightly greater when stratifying by reporting unit, because comparisons typically compare performance between different reporting units within the same strata (e.g., safety-net hospital to safety-net hospital), which could lead to the perception that certain types of providers are held to lower standards of care. Moreover, stratification by reporting unit characteristics does not reveal differences within units (such as those attributable to patient characteristics or quality of care). Notably, public reporting of unadjusted performance data stratified on reporting unit characteristics but not on patient characteristics does eliminate the incentive for providers to avoid high-risk patients in order to increase their apparent performance.

Finally, practical limitations of stratification may include small sample sizes, which may limit precision. These are limitations of all the methods the committee identified, but are more severe for stratification. However, methods exist to address these limitations—for example, aggregating across multiple years and suppressing estimates with insufficient sample size or reliability. Because subgroups are concentrated in a small subset of providers and because some providers have low volumes of patients, it may only be possible to report the quality of care for certain subgroups for a subset of all Medicare plans/providers, but those providers/plans are likely to contain nearly all of Medicare beneficiaries from that subpopulation (CMS, 2016a).

B. ADJUSTING PERFORMANCE MEASURE SCORES

Adjusting performance measure scores aims to estimate the true quality of reporting units (i.e., the quality a reporting unit would have if all units had the population average patient). In other words, such adjustment aims to statistically minimize the effect of factors that may independently influence performance indicators used in VBP, such as social risk factors that make it difficult for providers disproportionately serving socially at-risk populations to improve or achieve performance benchmarks under the status quo (which does not take these factors into account). Adjustment of performance measure scores can be done for the average disparity within a provider or plan, the average disparity between providers and plans, or both. (Note, the provider or plan refers to the reporting unit so, for example, if the reporting unit is at the hospital level, differences between doctors within a hospital remain within a provider.)

Differences within providers or plans are stronger evidence of the effect of social risk factors on performance indicators than differences between

providers or plans, because such between-provider or between-plan differences could reflect the unmeasured influence of the provider or plan (i.e., true differences in quality) (Elliott et al., 2001; Jha and Zaslavsky, 2014; Zaslavsky et al., 2001). Thus, adjustment for within-provider and within-plan differences avoids adjusting for quality differences. Such within-provider and within-plan adjustment would result in more accurate performance measurement, which could better focus behavioral nudges into better providers or plans to reduce disparities and improve quality. Additionally, because social risk factors should no longer have substantial influence on performance measure scores, accurate adjustment would reduce incentives to avoid patients with social risk factors.

Under the status quo, which does not generally account for social risk factors, it is more difficult on average for providers and plans disproportionately serving socially at-risk patients to achieve performance targets owing to the influence of social risk factors. Although incentives to improve care under the status quo may be diminished when performance measure scores are adjusted for within-provider differences, this incentive reflects a disadvantage of these providers under the status quo (the greater average difficulty and greater resources needed to achieve benchmarks) rather than a benefit of not including social risk factors in the existing VBP.

Incorrect adjustment of performance measure scores could also produce several potential unintended consequences. For example, whereas adjusting for within-provider differences accounts for differences between subpopulations within a provider (e.g., subgroups with high and low levels of social risk factors) and is more likely to capture differences arising from patient characteristics, adjusting for between-provider or between-plan differences may result in incorrectly measuring provider performance. This mismeasurement may occur because adjusting for between-provider differences effectively assumes that patients with social risk factors receive care from providers of the same quality as do patients without social risk factors. However, evidence suggests that providers disproportionately serving socially at-risk populations may provide lower-quality care compared to hospitals serving the general population (NASEM, 2016b). Thus, observed differences in quality and outcomes by social risk factors measured at the provider or plan level may capture both differences in patient characteristics as well as an unmeasured influence of a provider or plan characteristics linked to overall quality (such as true differences in quality). By conflating differences owing to social risk factors and true differences in quality, adjusting for between-provider differences would remove incentives to improve care, especially for patients with social risk factors. The risk is greater for threshold-based incentives; incentive schemes with continuous reward functions (e.g., paying for improvement) are less subject to this concern. This is a serious disadvantage that CMS would want to consider care-

fully. Nevertheless, addressing between-provider differences in quality of care is an important part of improving health equity. Improving quality measurement and restructuring incentives may help reduce such disparities, but other policy initiatives may be required in addition.

Importantly, any effects of adjusting performance measure scores on payment are indirect—meaning, they do not affect provider compensation unless payment is adjusted based on performance scores. Additionally, effects on payment require consideration of the particular payment strategy. If the incentive payment is based on adjusted performance scores, this could increase resources to providers serving socially at-risk populations, which these providers could invest in strategies to improve quality of care for all patients and to reduce disparities. Finally, although adjusting performance measure scores for within-provider differences creates an estimate of the average disparity within providers, neither the magnitude of this disparity nor subgroup performance are apparent in the publicly reported performance scores. Even if both adjusted and unadjusted performance scores are publicly reported, only whether a provider is doing better or worse relative to the average disparity is visible. Thus, adjusting performance measure scores does not make disparities visible unless performance scores stratified on patient characteristics within reporting units are also publicly reported.

C. DIRECT ADJUSTMENT OF PAYMENT

Direct adjustment of payment refers to any adjustments in payment that by themselves do not affect performance measure scores. This could be done by adjusting the payment formula for social risk factors directly (without adjusting performance measures) (CMS, 2015b) or by setting different benchmarks for payment for different strata of social risk factors (Damberg et al., 2015). By accounting for the increased resources (i.e., estimated costs) needed to care for socially at-risk populations, directly adjusting payments avoids unintentionally redistributing resources away from (i.e., underpaying) providers who serve patients with social risk factors and reduce incentives to avoid these patients. More favorable allocation of resources to these provides would increase their resources (Damberg et al., 2015), which they could invest in reducing disparities and improving quality and efficiency. However, if the payment formula is adjusted directly, providers could be awarded despite poor performance or poor outcomes, which would reduce incentives to improve care. Finally, because directly adjusting payments does not affect publicly reported measures, this method does not improve the accuracy of performance scores. Therefore, adjustments of payments do not make disparities visible unless this method is coupled with public reporting stratified by patient characteristics within reporting units. Relatedly, if payment is directly adjusted, but performance

is still reported without adjustment, then there could be incentives to avoid patients with social risk factors.

D. RESTRUCTURING PAYMENT INCENTIVE DESIGN

Restructured payment incentive designs do not explicitly incorporate measures of social risk factors, but they do implicitly account for them. In so doing, like directly adjusting for payment, this implicit adjustment accounts for the increased resources needed to care for socially at-risk populations and therefore avoids unintentionally underpaying providers who serve these populations and reduces incentives to avoid patients with social risk factors. Payment incentives can be restructured in several ways. For example, in addition to other rewards and penalties, providers and plans could receive a bonus for having low disparities (Blustein et al., 2011; Casalino et al., 2007). This has the obvious advantage of directly incentivizing disparities reduction. Similarly, providers and plans could receive a bonus for improving quality and efficiency relative to their own benchmark (i.e., paying for improvement) (Casalino et al., 2007; Rosenthal et al., 2004). This would directly incentivize quality improvement and efficiency, but may also reward providers at lower levels of absolute performance. Like directly adjusting payments, restructuring payment incentive design does not affect publicly reported measures and therefore does not improve the accuracy of performance scores. Restructuring incentive design does not make disparities visible unless it is combined with public reporting stratified by patient characteristics within reporting units.

Comprehensive descriptions of the 10 methods, as well as their advantages and disadvantages can be found in Table C4-1 in Appendix C. Table 3-1 summarizes how different categories of methods to account for social risk factors might achieve the committee’s four policy goals. Table 3-2 summarizes how different categories of methods might result in unintended consequences. These tables capture the primary effects of the different methods on policy goals and adverse consequences. Just as any approach will interact with the underlying incentive design, any method to account for social risk factors may interact with other methods when used in combination. In many cases, combinations of methods may help mitigate the risks of any method alone.

GENERAL CONSIDERATIONS FOR MITIGATING UNINTENDED CONSEQUENCES

Approaches that promote equity will tend to reduce disparities in outcomes related to access and quality of care for patient populations between socially disadvantaged versus advantaged populations. As the

committee concluded, any specific approach to accounting for social risk factors requires continuous monitoring with respect to its four policy goals (NASEM, 2016a).³ VBP programs may also risk creating perverse incentives, such as incentives to underdeliver care for patients with social risk factors or to otherwise reduce the provision of beneficial care. These adverse consequences undercut the fundamental objective of equitable health care. Therefore, as the committee concluded, it is also important to minimize potential unintended adverse consequences to patients with social risk factors and to monitor the effect of any specific approach to accounting for social risk factors to ensure the absence of any unanticipated adverse effects on health disparities (NASEM, 2016a).⁴

The committee takes this opportunity to expand on two potential unintended consequences about which some opponents of accounting for social risk factors have raised particular concerns: reducing incentives to improve care for patients with social risk factors and for patients overall, and obscuring disparities. In particular, the committee suggests how these unintended consequences might be mitigated.

Some opponents of accounting for social risk factors worry that by making it easier for providers and plans to reach performance targets or rewarding them at lower levels of absolute performance, accounting for social risk factors may remove incentives to improve quality and efficiency (in particular, to exceed benchmarks) for patients with social risk factors (Bernheim, 2014; Jha and Zaslavsky, 2014; Kertesz, 2014; Krumholz and Bernheim, 2014; O’Kane, 2015). This concern arises from the fact that there is substantial variation in performance among providers and plans disproportionately serving socially at-risk populations, and yet there is evidence that it is possible to provide optimal care and achieve quality benchmarks for socially at-risk populations (NASEM, 2016b). The committee emphasizes that the fact that some providers do well with socially at-risk populations does not mean that it is not more difficult to do so on average. This greater average difficulty arises from the fact that provider actions may be insufficient to overcome the influence of social risk factors on performance indicators used in VBP. Additionally, improving quality of care and achieving good outcomes (i.e., quality benchmarks) for patients with social risk factors may require greater costs, time, and effort compared to doing so for more advantaged populations. Thus, the standard for taking social risk factors into account should not be that it is impossible to provide optimal care, but that it is more difficult and more costly on average. Accounting for social risk factors in quality measurement and payment can be seen as accounting for this greater average difficulty and greater average

___________________

³ See Conclusion 9 in the committee’s third report (NASEM, 2016a).

⁴ See Conclusion 4 in the committee’s third report (NASEM, 2016a).

TABLE 3-1 Policy Goals of Methods to Account for Social Risk Factors

NOTE: This table pertains to alternative methods to account for social risk factors in Medicare payment. It does not capture the policy goals achieved under the status quo (which generally does not account for social risk factors). Disadvantages of the status quo are described in detail in Chapter 1. This table assumes the base payments to providers and plans are adequate.

TABLE 3-2 Potential Unintended Consequences of Methods to Account for Social Risk Factors

NOTE: This table pertains to alternative methods to account for social risk factors in Medicare payment. It does not capture the potential unintended consequences of the status quo (which generally does not account for social risk factors). Disadvantages of the status quo are described in detail in Chapter 1. This table assumes the base payments to providers and plans are adequate.

cost to improve quality and achieve performance benchmarks for patients with social risk factors. The committee recognizes that certain methods may diminish (but not entirely remove) incentives to improve quality and reduce disparities, but the committee also suggests that any approach should be sure to include sufficient incentive for quality improvement overall, as well as for socially at-risk populations. As described in prior sections, achieving this might require a combination of reporting and payment methods.⁵ Critics of adjustment are also concerned that accounting for social risk factors would obscure differences that arise from poor quality care (Krumholz and Bernheim, 2014). It is true that a risk of some methods to account for social risk factors is the perception of different standards for different populations, which could have negative symbolic value, as described above. As intimated earlier, this may be due to the fact that variation in performance arises through multiple mechanisms, including not only the influence of social risk factors, but potentially genuine differences in the quality of care provided. Approaches that adjust for and report by patient characteristics and within-provider differences are less subject to this concern than approaches that adjust for and report by provider characteristics. At the same time, the committee emphasizes that its approach to accounting

___________________

⁵ See Recommendation 6 in the committee’s third report (NASEM, 2016a).

for social risk factors is not intended to obscure disparities that do exist and in fact seeks to reveal disparities by social risk factor. For this reason, the committee reiterates that, if CMS’s goals for VBP include monitoring and reducing disparities, because only public reporting stratified by patient characteristics within reporting units makes disparities visible by providing quality information for different subgroups, stratified public reporting must be part of any approach to improve on the status quo.

The committee notes that some restructuring of payment formulas may still be needed to ensure that there are sufficient incentives for health plans and providers to improve access, quality, and outcomes for groups that are

disadvantaged by high levels of social risk factors. Payment formulas that incentivize improving care for socially at-risk individuals and communities may include paying for performance or change in performance for subgroups with high levels of social risk factors. Furthermore, improving health equity may require both accounting for social risk factors in payment and quality improvement interventions.

HOSPITAL READMISSIONS REDUCTION PROGRAM EXAMPLES

The HRRP requires CMS to reduce a share of the base operating payments to acute care hospitals paid under the Inpatient Prospective Payment System that have the highest readmission rates (CMS, 2016b). CMS implemented the HRRP beginning in fiscal year (FY) 2013 (October 1, 2012). Currently, the program calculates excess readmissions for six conditions: acute myocardial infarction, heart failure, pneumonia, chronic obstructive pulmonary disease, total hip arthroplasty or total knee arthroplasty, and coronary artery bypass graft surgery. To calculate the payment reduction, CMS first calculates a hospital’s excess readmissions. The algorithm used to calculate excess readmissions captures an individual hospital’s performance compared to that of hospitals nationally over a 3-year performance period. The excess readmission measure is then risk adjusted using a methodology endorsed by the National Quality Forum (NQF) to account for differences in patient characteristics; factors currently included in the adjustment include certain demographic characteristics, clinical comorbidities, and patient frailty (NQF, 2014). CMS also accounts for planned readmissions. CMS then uses the adjusted excess readmissions measure to calculate the payment adjustment. For FY 2017, the maximum reduction is 3 percent. According to a Kaiser Family Foundation analysis of CMS data, in FY 2017, based on performance for the period of July 2012 through June 2015, more than half of hospitals nationwide will be penalized under the HRRP (Rau, 2016). The average hospital penalty among penalized hospitals is estimated to be –0.73 percent, totaling approximately $528 million (Rau, 2016).

As described in prior sections, CMS has several options to account for social risk factors (in addition to the existing risk adjustment): (A) stratified public reporting; (B) adjustment of performance measure scores; (C) direct adjustment of payment; and (D) restructuring payment incentive design. One overarching caveat for the following examples is that the committee assumes the social risk factor indicators adopted by any method are measured reliably and with sufficient precision to allow for meaningful applications.

A. Stratified Public Reporting in the HRRP

For the HRRP, stratified public reporting could involve showing readmission rates separately for patients who identify as white, black, Hispanic, or other racial and ethnic groups. In principle, such stratification could be done with any discrete categories of social risk factors including those created from continuous measures such as income (e.g., patients with incomes above and below the median of the observed distribution of all relevant patients). There would be a logical limit to how many strata could be reported and sample-size considerations would also apply.

Stratified reporting would show the public (who now see only a single clinically adjusted rate for a hospital’s overall performance) two important pieces of information: the different composition of patients across hospitals and the readmission rates for the groups shown. It would be possible to see how big the disparities are on average, whether they always exist, and whether a particular institution does better with some groups than others and by how much.

Stratified data will help all patient groups select hospitals with the best readmission rates for someone in their group. They can also help hospitals find peers who perform better for patients with high social risk factors from which the hospitals could learn better practices. Last but not least, stratified public reporting of hospital readmissions would allow CMS to track whether the HRRP is exacerbating disparities as some have worried. As described above, of all the methods of accounting for social risk factors only stratified public reporting can generate these “information” benefits for patients, hospitals, and CMS. If CMS views these mechanisms as important to achieve the policy goals of reducing disparities in access, quality, and outcomes; quality improvement and efficient care delivery for all patients; fair and accurate reporting; and compensating providers fairly, then stratified reporting must be part of any approach to improve on the status quo, where public reporting obscures differences in performance for high- and low-social risk factor groups. For example, CMS could do for the HRRP what it has done for Medicare Advantage and Part D plans for their Healthcare Effectiveness and Data Information Set (HEDIS) and Consumer Assessment of Healthcare Providers and Systems (CAHPS) scores—stratifying performance scores by four racial and ethnic groups (CMS, 2016a).

B. Adjustment of Performance Measure Scores in the HRRP

Adjusting performance scores for social risk factors in the context of the HRRP could be accomplished by adding social risk factor indicators to the adjustment formula presently in use. Imagine for example that race, ethnicity, and income are included in this formula. This would mean that

some hospitals that now receive a penalty but have more patients than average from low-income, racial, and ethnic minority backgrounds would have adjusted performance that is sufficiently good to lower or even eliminate their penalty. Conversely, a hospital that almost incurs a penalty currently with a very affluent patient population might incur a penalty if such adjustments were added.

If the adjustment takes account of the average disparity between high- and low-social risk factor groups (i.e., the marketwide disparity) and leaves the effect of any disparity within a hospital (i.e., the hospital-specific disparity), this adjustment may improve measurement and payment equity. Safety-net hospitals would likely benefit from lower penalties, which might help them better serve their patients. But it would also be true that the incentives to improve care for patients with high social risk factors might be curtailed at lower levels of absolute performance than the incentives to improve care for patients with higher social risk factors—a hospital with patients with high levels of social risk factors gets to the “no-penalty” zone with poorer absolute readmission rates than a hospital with patients with low levels of social risk factors. With accurate adjustment, hospitals serving low-social risk factor patients might see increased penalties and then be appropriately motivated to work on reducing readmissions for all patients. Accurate adjustment would also reduce the awards for avoiding patients with high social risk factors—they should no longer have as big an effect on a hospital’s penalty though they may still be more expensive to treat and manage readmissions effectively. It should be noted that the potential incentive disadvantages of high social risk factor hospitals reaching no-penalty zones as easily as low social risk factor hospitals reflect a trade-off with equity and accuracy of measurement—incentives might be more evenly and equitably distributed. Increasing incentives for high social risk factor hospitals might incorporate additional methods such as those under the categories of direct adjustment of payment and restructuring payment incentive design in combination with those under the categories of stratified public reporting and adjustment of performance measure scores. Notably, this problem is a property of threshold-based incentives; incentive schemes with continuous reward functions are less subject to this concern.

If adjustment accounts for all (within and between) differences in performance that are associated with social risk factors, then it removes any incentive to reduce the disparity in performance for patients with high social risk factors. Such “overadjustment” may also eliminate the incentive to improve readmissions altogether by pushing hospitals into the no-penalty zone although they could still improve their readmission rates. For this reason, adjustment for only within-facility differences has been recommended by NQF (2014).

C. Direct Adjustment of Payment in the HRRP

Adjusting payment only for social risk factors in the context of the HRRP could be accomplished by adjusting the benchmarks for payment rather than the readmission rate itself. For the HRRP, which uses a comparison of actual to predicted readmissions to determine the penalty, this would mean having two different adjusted measures—one for public reporting (without adjustment for social risk factors) and one for payment (with social risk factors). Adjustment of the payment algorithm could also be done not by allowing the benchmarks to reflect social risk factor differences but instead to account for the estimated additional costs of providing care at the same level of performance for patients with high-social risk factors. The policy goals achieved and potential unintended consequences of directly adjusting payment for the HRRP are similar to adjusting performance measure scores except that adjusting payment alone would leave visible in the single reported readmission rate the reduction in average performance associated with patients with social risk factors. Without stratification this adjusted rate may or may not be better than the status quo (a single rate adjusted for clinical risk factors but not for social risk factors) in terms of the information value for consumers choosing a hospital. Likewise a combination of stratified public reporting with adjustment of performance measure scores is not obviously better or worse (but different) than a combination of stratified public reporting and direct adjustment of payment. Because of the informational benefits of stratified public reporting identified above, any type of adjustment (including none) is better when paired with stratified public reporting than not.

D. Restructuring Payment Incentive Design in the HRRP

Finally, the HRRP could be entirely reformulated in a number of ways that account for social risk factors. One simple suggestion would be to reward improvement rather than the level of performance, with the caveat that some kind of “maintenance of high performance” award would be needed. This approach could be combined with any of the other approaches for increased transparency (stratified public reporting) or accuracy/fairness of measurement and payment (either adjustment of performance measure scores or direct adjustment of payment under the conditions noted above). Some novel approaches to the redesign of incentives may stand alone in terms of serving CMS’s goals—they would effectively incorporate stratification and adjustment in some way. Performance on readmission rates (adjusted for clinical factors as now) could be awarded points on a scale from some minimum performance to “best in class” performance as with the Hospital Value-Based Payment program (CMS, 2015a). There could be

two scales, one for patients with high social risk factors and one for patients with low social risk factors, and the points could be added together with or without weighting (where the weights reflect the differential costs of improving care for high-risk populations) or additional points for improvement on one or the other performance scales or for disparity reduction.

REFERENCES

MedPAC (Medicare Payment Advisory Commission). 2013. Report to the Congress. Medicare and the health care delivery system. Washington, DC: MedPAC.

NASEM (National Academies of Sciences, Engineering, and Medicine). 2016a. Accounting for social risk factors in Medicare payment: Criteria, factors, and methods. Washington, DC: The National Academies Press.

NASEM. 2016b. Systems practices for the care of socially at-risk populations. Washington, DC: The National Academies Press.

NQF (National Quality Forum). 2014. Risk adjustment for socioeconomic status or other sociodemographic factors. Washington, DC: National Quality Forum.

O’Kane, M. 2015. Comment on the advance notice of methodological changes for calendar year 2016 for Medicare Advantage call letter. http://www.ncqa.org/public-policy/comment-letters/medicare-advantage-03-2015 (accessed December 20, 2016).

Price, R. A., A. M. Haviland, K. Hambarsoomian, J. W. Dembosky, S. Gaillot, R. Weech-Maldonado, M. V. Williams, and M. N. Elliott. 2015. Do experiences with Medicare managed care vary according to the proportion of same-race/ethnicity/language individuals enrolled in one’s contract? Health Services Research 50(5):1649-1687.

Rau, J. 2016. Medicare’s readmission penalties hit new high. http://khn.org/news/more-than-half-of-hospitals-to-be-penalized-for-excess-readmissions (accessed October 31, 2016).

Rosenthal, M. B., R. Fernandopulle, H. R. Song, and B. Landon. 2004. Paying for quality: Providers’ incentives for quality improvement. Health Affairs (Millwood) 23(2):127-141.

Zaslavsky, A. M., L. B. Zaborski, L. Ding, J. A. Shaul, M. J. Cioffi, and P. D. Cleary. 2001. Adjusting performance measures to ensure equitable plan comparisons. Health Care Financing Review 22(3):109-126.

This page intentionally left blank.