In Pursuit of Meaningful Performance Measures for Postacute Care

The Improving Medicare Post-Acute Care Transformation Act (IMPACT Act) of 2014 established uniform reporting requirements across postacute care (PAC) settings for patient assessment data, quality measures, and resource use measures. The legislation aims to bring PAC settings into a value-based care paradigm by linking payments to care quality and advances efforts by the Centers for Medicare & Medicaid Services (CMS) to establish meaningful quality measures for priority areas. These changes in PAC settings follow several similarly aligned programs launched by CMS after passage of the Affordable Care Act, most of which have focused on adjusted payments to hospitals based on performance metrics that are thought to reflect care quality. These programs and the accompanying public reporting of facility performance are intended to incentivize quality-oriented policies, practices, and systems. These programs intrinsically rely on rigorous, valid, and practically useful quality measures, which should also be widely accepted, understandable, and able to motivate quality improvement of real-world care.

Measures linked to CMS action must always be held to a very high bar, particularly given their substantial cost and real-world ramifications. They must accurately identify high- and low-performing facilities and reflect meaningful care processes. To this end, the potential unintended consequences of implementation of such measures must be examined. Given the tight implementation schedule outlined in the IMPACT Act, the speed of adoption of new PAC setting measures appears to be outpacing opportunities to independently study candidate measure performance.

Potentially preventable readmission (PPR) measures, in general, have been an area of debate, with concerns ranging from construction of statistical measures, to exclusion of key factors, to unintended consequences in real-world application. A range of factors affect readmission risk, but they are not always considered in the PPR measures used by CMS. Emerging evidence highlights the potential for some performance measures, including the PPR measure, to mask or exacerbate underlying health disparities, thus highlighting the importance of empirical, independent investigation of measure discriminability and validity.¹

Malcolm and colleagues² investigate variation in newly implemented PPR measures across inpatient rehabilitation facilities (IRFs). The authors demonstrate that both the all-cause and PPR readmission measures are no different than the national mean for 98% to 99% of all IRFs. Adjustment factors included in the analysis were similar but not identical to the CMS approach. It is unclear whether the findings of Malcolm et al² reflect a ceiling effect for IRF performance or insufficient risk adjustment, or whether the existing readmission measures are insufficiently sensitive to underlying IRF care quality.

The study by Malcolm et al² is a valuable contribution to the literature that appropriately identifies the need for further research to understand reasons for the minimal variation in IRF readmission measures. Long-standing areas of debate in readmission measurement methods include outcome time frame selection (eg, why use 30 days?), elusive definitions of “preventable” and “avoidable,” adequacy of reflection of real-world care processes and quality, and inadequate risk adjustment through failure to account for differences in vulnerable populations. Prior work³ has demonstrated variation in IRF readmission rates by geography, region, and institutional profit status. Any or all of these factors may be contributing to the findings observed in this study.

The issue of robust social and economic risk adjustment itself is worthy of additional discussion. Across multiple studies,⁴ social determinants, including neighborhood-level socioeconomic disadvantage, individual socioeconomic status, living situation, and social support, have been associated with significantly increased risk of readmission, in some cases risk equal to or greater than that associated with chronic conditions. This remains an area of long-standing debate, with some researchers expressing concerns that adjusting for economic or social disadvantage would disincentive health systems from delivering high-quality care to disadvantaged populations, and others expressing concerns that financial penalties tied to readmission measures disproportionately affect safety-net health care institutions.^5-7 This leaves such institutions with fewer resources to engage in effective community-based partnerships to reliably reduce the influence of social forces on readmission risk. Whether different adjustment for social risk factors would significantly improve discriminability of the IRF readmission measure remains unclear, but it is an important area for future study.

The present findings also call into question assumptions about the transferability of some performance measure approaches among distinct settings in care. The IRF patient population is arguably distinct from other PAC settings, such as skilled nursing facilities. The IRF population often includes patients with higher acuity with functional deficits who are capable of tolerating the demands of intensive rehabilitation. Because the nature of care in IRFs is heavily focused on addressing function to support optimal and effective transition into the community, postdischarge continuity in IRFs may be improved by care processes more directly relevant to these functional needs. Among IRF patient populations, readmission predictive analyses inclusive of functional status have routinely demonstrated superior predictive ability than models based solely on medical comorbidities.⁸ Functional status appears to be associated with PPR among skilled nursing facility and home health patient populations as well,^9,10 suggesting that it is a worthwhile candidate for consideration in the PPR measure with potentially broad relevance across PAC settings.

Finding meaning in the IRF readmission measure for CMS, health care institutions, and patients alike has proven elusive. The lack of variation in IRF performance in the current study is certainly not an indication that further improvements in care continuity are unneeded. On the contrary, these findings remind us that developing practical, clinically sensitive, and meaningful performance measures always is (or should be) an iterative process. Before accepting the current measures for what they are, CMS should continue to monitor measure performance and consider adjustments that would improve discriminability and improve utility for all stakeholders. The IRF readmission measure and the clinical efforts that are motivated, or omitted, in light of performance on the measure are broadly relevant to all CMS readmission initiatives.

Published: December 13, 2019. doi:10.1001/jamanetworkopen.2019.17558

Open Access: This is an open access article distributed under the terms of the CC-BY License. © 2019 Gilmore-Bykovskyi A et al. JAMA Network Open.

Corresponding Author: Andrea Gilmore-Bykovskyi, PhD, RN, School of Nursing, University of Wisconsin, 3173 Cooper Hall, 701 Highland Ave, Madison, WI 53705 (algilmore@wisc.edu).

Conflict of Interest Disclosures: Dr Gilmore-Bykovskyi reported receiving grants from the National Institutes of Health (NIH) outside the submitted work. Dr Crnich reported receiving grants from the state of Wisconsin, Agency for Healthcare Research and Quality, and the US Department of Veterans Affairs (VA) outside the submitted work. Dr Kind reported receiving grants from the NIH and VA outside the submitted work.