In its June 2013 report to Congress, MedPAC offered ways to refine the CMS Hospital Readmissions Reduction Program (HRRP). These included issues of stability (and efficiency) due to dealing with multiple condition-specific measures, the calculation of the existing CMS payment penalty, the inverse relationship between readmission and mortality for heart failure rates, and the topic of socioeconomic status (SES) and risk adjustment.
The bias against high Disproportionate Share Hospitals (DSH) apparent in the CMS payment policy is particularly concerning. This comes at a time when Medicare DSH payments are being directly adjusted as part of ongoing reforms and Medicaid DSH payments are being cut nationally in accordance with the Affordable Care Act. This heightened sensitivity brings urgency to the discussion of how to correct for the perceived SES bias, specifically whether measures reflecting SES should be included in the current risk-adjustment formula. If additional measures reflecting SES are to be considered, it will be important to separate the effects that may be attributed to generally lower performance in low income areas from those attributable to the complexities of treating a challenging population. In other words, can the risk adjustment method help us distinguish whether hospitals that care for poorer patients perform worse because they don’t do a good job, or because their patients are more difficult to care for?
Is it advisable to assign providers to risk groups for the purposes of risk adjustment? This blog hopes to answer that question.
MedPAC argues that the use of peer groups, based upon the share of Medicare patients qualifying for Social Security benefits, is a preferred solution that can adjust for SES while maintaining transparency. MedPAC’s proposed solution envisages providing readmission reports that are unadjusted for SES alongside peer group targets that are used for computing financial incentives. Thus the solution will provide two measures: a public report that will provide an outside entity insight into hospital performance, and a financial calculation that adjusts for factors beyond hospital control, in theory associated only with the effects of SES (albeit at an average rate for the peer group).
Leaving aside the issue of whether a single factor, such as SSI, provides an adequate representation of the effects of SES, this approach is problematic for a number of reasons. First, in terms of transparency, to whom is it transparent? The public report will not account for SES factors acknowledged to be beyond the control of the provider and worthy of consideration when evaluating performance. If there is a thought that this information should guide the consumer, then how should they select the provider for treating their condition? Some providers will appear better on a risk-adjusted performance basis across all providers, but worse than their peer group average and vice versa. More importantly than external perception, how should the provider view itself internally? Should quality teams pushing change within the hospital be concerned with their relative position or be satisfied? Will a financial incentive telling one story counteract a public report telling another? Having two intentionally discordant results does not promote transparency no matter which way you slice it.
Second, the construction of peer groups needs to be treated with caution. Since the peer group average would determine the magnitude of a penalty, providers will argue vociferously with regard to both the applicability of cut-off points for inclusion and the fairness of assigning providers to their group. This issue is compounded at the state level where there are fewer providers from which to draw peer groups and a decile will therefore include an uncomfortably large range of SSI requiring modification to obtain any face validity. Modifications may in turn result with peer groups of 1 or 2 hospitals. Moreover, using a single variable such as SSI to create a peer group may not be tenable because it assumes that there is no separate interaction between poverty and the challenges of providing care that should fairly be accounted for: providers may be placed in peer groups comprised of urban and rural locations that face different challenges in providing care for the poor or provide services, such as inpatient psychiatric services, that may attract more complex patients whose poverty complicates treatment of the underlying condition. Providers may also need to be reassigned to different peer groups should their SSI percentage change over time. While in theory splitting hospitals into evenly divided groups based upon a single variable is manageable, in practice the process is far more complicated.
If the clinical model is insufficient (i.e. persistent disadvantage in rankings for providers caring for low income individuals), additional factors such as literacy or English language proficiency have also been suggested, as well as data elements such as income that are present at a community level. Exogenous adjustment for such factors is preferable to peer groupings. Calculating a single, explicit, exogenous factor to be used in conjunction with the risk-adjustment model retains separation of performance measure relative to the individual patient from the community/provider level SES adjustment. Much like the exogenous IME or DSH adjustments in PPS that relate more to the community/provider than the individual patient, an exogenous SES adjustment would provide an explicit and quantified adjustment for communities and providers with a disproportionate level of patients in lower SES statuses. This type of SES adjustment is explicit and empirically derived, rather than derived as part of a black box peer grouping process, and can be assessed as to whether its magnitude makes sense or if it varies across measures or locations. No single provider could claim to be treated unfairly and the method upon which the adjustment is based would correctly be subject to public scrutiny and comment. Put simply it would be transparent. Further, it would also focus efforts to reduce the need for such an adjustment on the community services needed to address the issue.
In summary, as discussed in previous blogs, risk adjustment for provider rates that account for SES should include additional clinical and demographic factors that can be shown to improve predictive performance. In addition, those additional factors should be incorporated into a model that is based on continuous variables rather than a categorical model based on peer groups.