Readmissions Metrics in Hemodialysis: Do the Specifics Matter?

• hemodialysis
• ESRD
• hospitalization
• payment
• policy
• Medicare

The 2010 Affordable Care Act required establishment of the Hospital Readmissions Reduction Program by 2012 in the United States. Under this mandate, which aimed to improve quality by reducing hospital readmissions, the Centers for Medicare and Medicaid Services (CMS; the largest single payer in the United States) developed payment adjustment factors to reduce hospital reimbursement for unplanned readmissions that occurred within 30 days of an index discharge for a set of diagnoses. Endorsed by the National Quality Forum, a quality metric clearinghouse, these measures subsequently were adopted by most other payers in the United States. However, although clearly undesirable for patient experience and quality of life, whether 30-day readmission rates truly reflect the quality of care delivered by a hospital remains controversial.¹

Readmissions among Medicare beneficiaries are costly and are particularly common among patients on hemodialysis in the United States, with 35% of index discharges followed by an unplanned readmission within 30 days.² These readmissions are often for causes not related to the initial admission (Figure 1). This readmission rate exceeds readmission rates for other chronic conditions, such as congestive heart failure or chronic obstructive pulmonary disease.² Multiple factors are associated with an increased likelihood of readmission, only some of which may be modifiable by hospitals and dialysis facilities. These risk factors include extremes of age, the presence and severity of both medical and psychologic comorbid conditions, longer index hospitalization and need for mechanical ventilation during the index hospitalization, and more admissions over the preceding 6 months.^3,4 Limited data suggest that early evaluation and intervention in the dialysis facility may reduce the overall risk of readmission.^5,6

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Causes of rehospitalization within 30 days as a function of the cause of the index hospitalization, demonstrating that index hospitalizations and readmissions are often for unrelated causes. Derived from data contained in the US Renal Data System 2017 Annual Data Report.² CVD, cardiovascular disease; Rehosp, rehospitalization; VA, vascular access.

Unplanned hospital readmission rates were incorporated into the ESRD Quality Incentive Program in the United States in 2015, such that worse performance on the standardized readmission ratio (SRR) metric could place a dialysis facility at increased risk of financial penalties of up to 2% of Medicare income in subsequent years. The SRR reflects the number of unplanned readmissions per facility during a calendar year that occurs between 4 and 30 days after an index discharge, divided by the expected number of unplanned 30-day readmissions.⁷ The latter number is derived from a model designed to account for patient characteristics, the dialysis facility, and the discharging hospital. The SRR excludes certain unplanned readmissions, including those primarily for cancer, mental health, or rehabilitation.⁷ Importantly, this measure focuses on the number of readmissions rather than the number of patients who have a readmission. As a result, a small minority of patients within a facility who have high readmission rates could have a marked effect on a facility’s SRR. The SRR has been criticized for several shortcomings, including that some readmissions are unavoidable, potentially reflecting patient morbidity and illnesses that are not specifically related to the index discharge.^8,9

In this issue of the Journal of the American Society of Nephrology, Lin et al.¹⁰ evaluate the relationship between prior hospitalizations and the probability of related and unrelated 30-day readmissions among prevalent patients on hemodialysis in the United States. To assess relatedness of the index hospitalization and readmission, principal diagnoses for each hospitalization were categorized by organ system. Readmissions with a principal diagnosis within the same organ system as the index hospitalization were classified as related. Among 651,386 index hospitalizations in the US Renal Data System database in 2013–2014, roughly 45% occurred in individuals with up to one hospitalization in the year before the index hospitalization, 35% had two to four hospitalizations, and 20% had five or more hospitalizations. In adjusted analyses, having more hospitalizations in the preceding year was associated with a significantly higher probability of both related and unrelated 30-day readmissions; however, for all index discharges, readmission was more often for an unrelated than for a related cause. Related 30-day readmissions were more common after hospitalizations categorized within cardiovascular, pulmonary, digestive, and injury or poisoning organ systems. On the basis of these results, the authors asserted that CMS should consider refining 30-day readmission metrics to account for clinical relatedness and prior hospitalization burden.

Although there is room for improvement in the SRR, accounting for relatedness may not optimize this metric. Lin et al. worked hard to create an algorithm for relatedness that could be scalable to the policy level, but further assessment of validity is required to know if related versus unrelated readmissions are being classified correctly and in such a way that appropriately identifies quality care. Consider the following two patients: patient 1 was admitted with new onset atrial fibrillation during dialysis (organ system: cardiovascular) and readmitted within 30 days for exertional chest pain to rule out a myocardial infarction while in sinus rhythm (organ system: cardiovascular); patient 2 was admitted with influenza and required a prolonged intensive care unit stay (organ system: infection), during which she lost 6 kg of weight, and after discharge the dialysis target weight was not updated, and she was readmitted 5 days later with a diagnosis of heart failure (organ system: cardiovascular). Although, clinically, the readmission for patient 1 likely is unrelated to the index hospitalization, the algorithm classifies this as related. In contrast, although the readmission for patient 2 is both related to the index hospitalization and is modifiable at the dialysis facility, the current algorithm classifies this as unrelated. These examples highlight how difficult it is to transform administrative billing data into practical assessments of quality. Ultimately, in designing a metric, the trade-off is whether to be more restrictive, as proposed by Lin et al., increasing the risk of misclassification and gaming, or broader, as operationalized in the current metric, increasing the risk that quality becomes defined by nonmodifiable events. In designing the SRR, CMS elected for the latter. Regardless of the approach, both require rigorous adjustment for patient factors that are associated with an increased risk of hospitalization.

Despite there being few proven strategies to prevent readmissions in this population, we put forth the following recommendations on the basis of our own experience: (1) have a regular discussion of hospitalized patients on dialysis between the inpatient and outpatient teams, including a conversation at the time of discharge; (2) ensure that discharge paperwork is sent to the patient’s dialysis facility (and not just to the primary care physician); (3) scrutinize admission and discharge medication lists (especially in the era of electronic medical records); (4) reassess the patient’s target weight after discharge; (5) reconcile medications (ideally with review of pill bottles) as soon as possible after discharge; (6) have the nephrologist see the patient during the initial treatment (or as soon as possible) after discharge; and (7) coordinate visits to other providers/specialists as soon as possible after discharge. We suspect that, regardless of the measure specifications incorporated into a rehospitalization metric, these interventions will result in better performance on any metric that exists, and, far more importantly, will make a meaningful difference in the quality of care.

• Copyright © 2019 by the American Society of Nephrology