Measurement of hospital performance through risk-adjusted mortality and readmission rates is a cornerstone of the US health system, factoring heavily into value-based purchasing and other strategies to incent quality improvement. A key principle underlying such performance measures is risk adjustment, that is, the ability to adjust in statistical models for differences in populations served by hospitals, such as case mix and disease severity, that confer differential risks of adverse outcomes and that may confound comparisons of hospital quality. Yet, identifying which variables to include in risk-adjustment models remains controversial, with different models yielding markedly different assessments of hospital quality.1
Using national Medicare data, the study by Pollock and colleagues2 showed that higher hospital prevalence of a do-not-resuscitate (DNR) status that is present on admission (POA) may be associated with higher risk-standardized mortality rates but lower risk-standardized readmission rates across several clinical conditions, leading to a lower likelihood of penalization under the Centers for Medicare & Medicaid Services Hospital Readmissions Reduction Program (HRRP). The authors recommended that hospital prevalence of POA DNR status should be included in pay-for-performance risk-adjustment calculations. The implicit assumption is that hospitals with low prevalence of POA DNR status unfairly receive penalties, whereas hospitals with high prevalence of POA DNR status inadvertently get rewarded by the HRRP. Pollock and colleagues2 should be applauded for their thoughtful, well-designed analysis and their desire to improve performance measures. However, it is important to critically appraise their assumption that including POA DNR prevalence in risk-adjustment models will improve the ability to accurately and fairly assess quality.
Conceptually, what are we adjusting for when we include DNR prevalence in risk-adjustment models? A DNR status refers to a patient’s expressed desire to forgo cardiopulmonary resuscitation (CPR) in the event of cardiac arrest, and POA DNR status requires that this preference be documented in the hospital record at the time of admission. Individuals with a DNR status often, but not always, have a limited prognosis, a serious illness, or advanced age. A DNR status is among the strongest predictors of 6-month mortality among survivors of a critical illness.3 If a DNR status were exclusively a proxy for illness severity, it might be appropriate to include it in risk-adjustment models to control for differences in case mix between hospitals.
Yet, a DNR status does not exclusively reflect illness severity; it also reflects the process that precedes the election of this status, namely, advance care planning (ACP) discussions between patients and health care practitioners about the type and intensity of care that patients wish to receive at the end of life. There is wide consensus that ACP discussions are essential to improving care quality,4 allowing patients to specify their values and goals for the end of life and clinicians to tailor care delivery accordingly. Patients who choose a DNR status after a thoughtful ACP discussion often value comfort-focused care and may forgo not only CPR but also other interventions that extend life, such as rehospitalization. Thus, hospitals that achieve high prevalence of POA DNR status through high-quality ACP discussions with patients who express conservative treatment preferences would be expected to have lower readmission rates. Moreover, on average, hospitals that consistently engage in ACP discussions with patients in outpatient settings or at the time of admission will have higher prevalence of POA DNR status than facilities that never do, given that the default approach in US hospitals is to assume patients want aggressive life-sustaining measures (full code) unless otherwise specified. Therefore, to the extent that hospital prevalence of POA DNR status reflects ACP rates, it would be inappropriate to adjust for a measure that corresponds to a hospital’s efforts to improve the quality and patient-centeredness of care.
However, hospital prevalence of POA DNR status does not equate with rates of ACP, either. Hospitals could engage in a high number of ACP discussions but still have a low prevalence of POA DNR status if, on aggregate, the patients they serve prefer aggressive care. For example, black and Latino patients more often prefer to receive full, life-sustaining measures compared with white patients.5 Accordingly, hospitals that serve racial/ethnic minorities may have a low prevalence of POA DNR status even if they have engaged in high-quality ACP process with all patients.
Hospital prevalence of POA DNR status does not necessarily reflect the quality of ACP. Identifying a patient’s goals for end-of-life care requires more than asking a blunt question about a patient’s desire to receive CPR in the event of cardiac arrest. It necessitates a sensitive discussion guided by what matters most to the patient as they envision their final days. With misperceptions about the likelihood of a favorable outcome of CPR and subsequent better quality of life, patients may make a spur-of-the-moment decision to accept full, life-sustaining measures at the time of hospitalization, particularly during a rushed conversation with a harried admitting resident who may have had little training in leading ACP discussions.6 Hospital prevalence of POA DNR status may depend largely on where, when, and how clinicians engage patients in ACP discussions, and therefore this rate cannot serve as a proxy for ACP quality.
What might be the implications, then, of including hospital prevalence of POA DNR status in risk adjustment for HRRP or other pay-for-performance programs, as Pollock and colleagues2 suggested? Hospitals with a high prevalence of POA DNR status, which tend to have higher mortality and lower readmission rates, would be at risk of incurring greater financial penalties under the HRRP. In addition to presenting conceptual problems, as outlined, this approach could undermine the quality of care.
Specifically, because the default in the US is to assume patients prefer to receive full code, unless they specify otherwise, incorporating hospital prevalence of POA DNR status in risk-adjustment models for HRRP creates a disincentive for hospitals to discuss and/or document DNR status at the time of admission. Moreover, incorporating POA DNR prevalence in risk-adjustment models creates a perverse incentive for those hospitals that promote ACP discussions at the time of admission to steer patients toward accepting aggressive life-sustaining measures (full code) even if palliation in anticipation of death may better align with patient preferences. Moreover, hospitals could game or misuse documentation by deferring the discussion and documentation of goals of care until 1 to 2 days after admission and thereby generate a lower prevalence of POA DNR status and less likelihood of incurring a penalty through the HRRP.
Adjusting for hospital prevalence of POA DNR status in HRRP risk models may inappropriately reward or penalize hospitals on the basis of a metric that may not reliably reflect relevant constructs, such as illness severity or patient preferences, and that may reflect higher-quality care processes. Doing so might have the unintended consequence of decreasing the frequency of ACP discussions at admission and reducing the overall quality and patient-centeredness of care. We believe a better system would be one that provides an incentive for a high-quality ACP process, rather than adjusting for the outcome of that process (POA DNR prevalence) in risk models. Rewarding ACP process measures creates an incentive to ascertain and document what the patient’s goals truly are for care at the end of life, regardless of what code status they elect. Furthermore, because ACP billing codes are time-based, this approach could reward longer ACP conversations (>15 minutes), which may increase the likelihood of robust, higher-quality discussions. This approach is consistent with the Centers for Medicare & Medicaid Services use of ACP billing codes for payment incentives in the advanced Bundled Payment Care initiative.7 Although opportunities exist for better risk-adjustment methods in the HRRP, including hospital prevalence of POA DNR status is not the right approach.
Published: July 14, 2020. doi:10.1001/jamanetworkopen.2020.10915
Open Access: This is an open access article distributed under the terms of the CC-BY License. © 2020 Wiener RS et al. JAMA Network Open.
Corresponding Author: Renda Soylemez Wiener, MD, MPH, Center for Healthcare Organization and Implementation Research, ENRM Veterans Hospital, 200 Springs Road, Building 70 (152), Bedford, MA 01730 (firstname.lastname@example.org).
Conflict of Interest Disclosures: None reported.
Funding/Support: This work was supported in part by resources from the Edith Nourse Rogers Memorial Veterans Hospital (Dr Wiener) and from the Susan J. and Richard M. Levy 1960 Distinguished Professorship in Health Care Delivery (Dr Barnato).
Role of the Funder/Sponsor: The funders had no role in the analysis and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Disclaimer: The views expressed herein are those of the authors and do not necessarily represent the views of the US Department of Veterans Affairs or the US government.
Identify all potential conflicts of interest that might be relevant to your comment.
Conflicts of interest comprise financial interests, activities, and relationships within the past 3 years including but not limited to employment, affiliation, grants or funding, consultancies, honoraria or payment, speaker's bureaus, stock ownership or options, expert testimony, royalties, donation of medical equipment, or patents planned, pending, or issued.
Err on the side of full disclosure.
If you have no conflicts of interest, check "No potential conflicts of interest" in the box below. The information will be posted with your response.
Not all submitted comments are published. Please see our commenting policy for details.
Wiener RS, Barnato AE. Implications of Including Hospital Do-Not-Resuscitate Rates in Risk Adjustment for Pay-for-Performance Programs. JAMA Netw Open. 2020;3(7):e2010915. doi:10.1001/jamanetworkopen.2020.10915
Customize your JAMA Network experience by selecting one or more topics from the list below.