Hospital Readmission as an Accountability Measure

In June 2009, the Centers for Medicare & Medicaid Services (CMS) began publicly reporting on its Hospital Compare Web site all hospitals' 30-day readmission rates for patients hospitalized and discharged with pneumonia, acute myocardial infarction (AMI), or heart failure (HF). This “shaming” of some hospitals and research showing that about 1 in 5 Medicare patients is readmitted within 30 days after hospital discharge highlighted the problem of hospital care transitions.¹ With passage of the Patient Protection and Affordable Care Act, CMS will begin holding hospitals accountable for their readmission rates and adjusting payments to hospitals in 2013 according to their rate of “excess” vs “expected” Medicare readmissions for pneumonia, AMI, and HF. Hospitals will now be subject to both the “shame” of public reporting and the likely more potent “stick” of decreased reimbursement. It is essential to consider these new policies in light of their likelihood of success and their potential for negative unintended effects as the health care industry attempts to address presumably elevated hospital readmission rates.

The Medicare Payment Advisory Commission identified pneumonia, HF, and AMI as common, costly causes of hospitalization, with a high proportion of potentially preventable readmissions, and recommended that Congress and CMS implement policies to address high rates. Proposed solutions included differential reimbursement for low- vs high-performing facilities and the bundling of inpatient and outpatient payments in an effort to better align performance incentives. These efforts led to inclusion of mandates in the Patient Protection and Affordable Care Act to reduce hospital readmissions by using risk-adjusted measures currently endorsed by the National Quality Forum and CMS. CMS aims to apply the rates from these 3 conditions to payments for all diagnosis related groups and expand the measurement to include chronic obstructive pulmonary disease, coronary artery bypass graft, percutaneous transluminal coronary angioplasty, and other vascular conditions in 2015.

Hospital readmissions and their preventability have been the subject of intense study, and their utility as a performance measure is often debated. One early study using chart review estimated that 1 of 7 readmissions for patients with diabetes, 1 of 5 readmissions for patients with HF, and 1 of 12 readmissions for patients with obstructive lung disease were attributable to substandard care.² According to a systematic review of 7 studies, low-quality inpatient care was associated with a 55% increase in risk for early hospital readmission.³ However, these analyses lack precision. Different reviewers have generated widely varying estimates of care quality, with estimates of readmission preventability ranging from 9% to 48%.⁴ Even different analysts examining the same data sets have reached differing conclusions. Such wide variability in estimates undermines the potential validity of using readmission rates in determining hospital reimbursement. Accountability measures should have a strong evidence base for their validity, should accurately measure whether high-quality care has been provided, and should have a low risk for unintended consequences.⁵ Some have observed that high readmission rates may be associated with reduced mortality and reflect severity of illness and actually high acuity of care.⁶

Several groups, including the developers of the CMS models, have used regression techniques with complex analyses of administrative data to estimate the proportion of potentially avoidable hospital readmissions. Using 2005 Medicare data, the Medicare Payment Advisory Commission estimated that 13.3% of 30-day hospital readmissions were potentially preventable and cost an additional $12 billion.⁷ However, the proprietary nature of this particular methodology prevents others from assessing its validity. The nonproprietary CMS algorithms involve hierarchical logistic regression modeling to calculate hospital risk-standardized 30-day all-cause readmission rates.⁸ These models include 2 demographic variables (age, sex) and use more than 20 International Classification of Diseases, Ninth Revision codes to adjust for baseline differences in severity of illness at hospital presentation that might influence readmission risk. Although the CMS model has been validated against chart-review methods and compares favorably, both sets of methods fail to account for a large degree of the variance observed in hospital readmissions. For example, the C statistic for the HF model is 0.6, only modestly better than chance.⁸ This makes sense, given the number of conceptual factors influencing readmission risk that are outside of the hospital purview. Thus, although hospital readmissions have become the gold standard for assessing the effectiveness of hospital discharge processes, this outcome measure remains a surrogate for the real outcomes of interest—health, quality of life, and value.

Other options include use of process of care measures, and CMS has accumulated a wealth of experience with indicators for clinical management of pneumonia, AMI, and HF. In similar fashion, CMS should develop care transitions process measures that document adherence to evidence-based practices such as high-quality medication reconciliation, telephone follow-up, or use of nurse-directed case management. In addition to this information, CMS might require hospitals to collect more direct measures of health and quality of life from patients after discharge, beyond patient satisfaction as captured by the Hospital Consumer Assessment of Healthcare Providers and Systems. Such standardized process measures and patient information could supplement hospital readmission rates in determining performance and could be factored into decisions about reimbursement.

Implementing accountability measures also carries the possibility of negative unintended effects. In particular, a policy that links payment to performance without adjusting for important cofactors runs the risk of unfairly reducing payments to hospitals caring for a high proportion of minority or economically disadvantaged patients. The CMS model does not adjust for socioeconomic status or race/ethnicity. In its review of 30-day readmission rates for appropriateness as a quality measure, the National Quality Forum justified this omission according to a desire to avoid masking economic or racial disparities in care quality related to hospital readmissions.⁹ This assumes that racial disparities in readmission rates exist, but reports examining the association between race and hospital readmission have not found race to be a significant readmission predictor.⁴

Economically disadvantaged patients have poorer overall health status, fewer community resources, and poorer access to primary care, all of which potentially confound assessment of readmissions risk. Critical access (ie, “safety net”) hospitals already face decreases in CMS Disproportionate Care payments that presumably will be offset by increased provision of health insurance to the patients served by these hospitals. Even though several programs to reduce hospital readmissions have proven efficacious and even cost-effective in randomized controlled trials,¹⁰ the resources needed to implement them may not be available at many cash-strapped hospitals. Additionally, many hospitals lack sufficient quality improvement expertise to fully overhaul their discharge process, and surrounding communities may not have sufficient primary care resources.

Moreover, altering Medicare payments solely according to readmission rates may actually penalize hospitals with low mortality rates. Patients dying during or after an index hospitalization are excluded from readmission calculations, but no adjustments are subsequently made during hospital risk standardization for hospital-specific mortality rates, a separate indicator already collected by CMS. In a study comparing risk-adjusted hospital readmission rates and mortality rates in HF, using available online CMS data, facilities with the lowest mortality had the highest readmission rates.⁶ In other words, hospitals may have reduced mortality rates by offering the treatment benefits of hospitalization to the sickest patients at risk for hospital readmission. Thus, a policy that fails to adjust for mortality rates carries the potential to unfairly curtail reimbursement to facilities offering high-quality, evidence-based care.

To avoid such negative unintended consequences resulting from the proposed reimbursement policies, CMS should consider adjustment for socioeconomic status and mortality in its models for 30-day all-cause readmission risk or including these case-mix considerations as a factor in reimbursement decisions. This method could avoid unnecessarily penalizing critical access hospitals caring for a high proportion of economically disadvantaged patients and guard against penalizing hospitals saving lives through higher-quality care.

Hospital readmission rates represent an important, if imperfect, proxy measure for poor-quality inpatient and outpatient care and poor care transitions. Linking hospital readmission rates to reimbursement is a complex issue that may have unintended negative consequences. Health care policy makers and the health care industry should give careful attention to developing innovative care transitions measures and refining readmissions analyses used for hospital reimbursement.