Publicly Reported Readmission Measures and the Hospital Readmissions Reduction Program: A False Equivalence?

Hospital readmissions have emerged as a central focus of quality improvement in the United States. For years, studies reported that patients discharged from the hospital had a high risk for being readmitted (1). In 2009, the Centers for Medicare & Medicaid Services began publicly reporting risk-standardized readmission rates for acute myocardial infarction, heart failure, and pneumonia (2–4), highlighting the performance of hospitals for Medicare beneficiaries hospitalized with these conditions nationwide. The conditions were selected because they represent high-prevalence, high-risk, and high-cost conditions known to have high readmission rates. As part of the Patient Protection and Affordable Care Act, the Hospital Readmissions Reduction Program (HRRP) introduced financial penalties for hospitals with higher-than-expected readmission rates, which were already being publicly reported for these conditions. Subsequently, readmission measures were developed for other conditions (5). Moreover, because readmission rates for the selected conditions did not reflect the overall readmission quality at a given institution, a hospital-wide risk-standardized readmission measure was developed by our group more than 6 years ago (6). The measure underwent public scrutiny and was approved by the National Quality Forum before being publicly reported, starting in 2014.

In their current study, Butala and colleagues (7) used a nationally representative all-payer database—the Nationwide Readmissions Database—to assess whether a hospital’s performance on the 3 HRRP conditions reflects readmission rates for other conditions or insurance groups. They found that relative performance varied not only among conditions but also for the same condition among different payer groups.

A key question is whether this finding represents an important gap in knowledge. The 3 readmission measures in the HRRP were not intended to define the totality of a hospital’s performance. Hospitals are complex organizations, with different teams and processes governing care and outcomes of patients with different diagnoses. Further, readmission rates for various conditions are known to lack perfect concordance. For example, readmission rates for the 3 HRRP conditions decreased more than those for other conditions soon after the program was introduced (8).

The premise and conclusions of Butala and colleagues’ study may represent a misunderstanding of what is publicly reported. Although only select, condition-specific readmission measures are included in the HRRP, the Centers for Medicare & Medicaid Services inpatient quality reporting program includes many more readmission measures that are publicly reported (5), including the hospital-wide readmission measure. Butala and colleagues conclude that “Current publicly reported measures may not be good surrogates for overall hospital quality related to 30-day readmissions.” However, the publicly reported hospital-wide measure, which they did not study, is designed to be just that. What the authors did find is that hospital excess readmission ratios for the HRRP conditions, which are used to determine penalties and convey hospital performance, differ from ratios for other conditions for some hospitals. This finding, however, does not support the conclusion that publicly reported measures do not reflect hospital quality—only that these 3 do not.

Nevertheless, it is true that the use of the hospital-wide measure for HRRP might enable the program to create a broader incentive for improvement. This is particularly germane given that readmission rates for the 3 initial target conditions under the HRRP have improved more than those for nontarget conditions (8). The initial implementation of the HRRP was focused on 3 conditions because corresponding publicly reported measures already existed, evidence for the feasibility of reducing readmissions was still limited, and it enabled hospitals to focus on a subpopulation of patients to develop protocols and interventions to reduce readmissions. However, after several years of sustained improvements in readmission rates for target conditions and plateauing of readmission rates for nontarget conditions (8), inclusion of the hospital-wide readmission measure in the HRRP may be an appropriate next step. Its focus on the entirety of readmission performance would allow hospitals to design locally relevant strategies to reduce readmissions in the context of their highest-risk patient populations. Moreover, if the HRRP’s success with the target conditions could be replicated for nontarget conditions, a larger improvement in the readmission outcomes for patients, as well as the overall costs for the health system, probably would result. The Medicare Payment Advisory Commission therefore has recommended a transition to the hospital-wide readmission measure in the HRRP (9). However, a penalty structure based on the hospital-wide measures would need to be refined, because transposing the current structure might increase the penalty burden for hospitals (10), affect disparities through disproportionate effects on safety net hospitals (9), and inadvertently dilute the focus on high-priority conditions.

An interesting aspect of Butala and colleagues’ study (7) is its assessment of hospitals’ performance on readmissions of non-Medicare patients, a group for which there is no system for reporting or comparing hospital readmission outcomes. The limitations of the Nationwide Readmissions Database, including the inability to track patients or hospitals across calendar years or to identify specific insurance-based patient groups (dual-insured and fee-for-service vs. managed care), create challenges to reliably assessing differences in readmission performance across age and insurance groups. Nevertheless, the differences observed by Butala and colleagues suggest that investigations are needed to elucidate the factors that drive the readmission risk in other patient groups, as well as strategies for risk mitigation. Moreover, a sustained improvement in readmission performance in these patients would require a valid system for hospital profiling as well as incentives to lower readmissions.

In conclusion, the need to evaluate health policy interventions continues. Current readmission measures and policies certainly can be improved. Science is needed to guide approaches to ensure the safety, fairness, validity, and effect of health policies, such as the HRRP. We already have a publicly reported measure that captures the overall hospital quality signal; building on this measure might be the best way to acknowledge the idea that financial incentives under a program to reduce readmissions should be based on more than a set of selected conditions.