Background
End-stage renal disease (ESRD) patients on hemodialysis experience a high burden of hospital admission and readmission. In 2015, ESRD patients were admitted to the hospital 1.7 times per year on average, and about 35% of hospital discharges among these patients were followed by a readmission within 30 days of discharge [
1] – almost double the readmission rate of the general Medicare population. This contributes to the overall economic burden of ESRD, as approximately one-third of Medicare expenditures for ESRD patients are for inpatient costs [
1]. The Centers for Medicare & Medicaid Services (CMS) have implemented quality metrics designed to incentivize reducing readmissions, including the standardized readmission ratio (SRR) for dialysis facilities. Several studies have focused on identifying predictors of readmission among ESRD patients, including demographics [
2], comorbidities [
3,
4], clinical characteristics of the hospital stay [
5], psychosocial factors [
6], and nephrology visits and other processes of care after initial hospitalization [
7,
8]. However, less is known about health outcomes after readmission.
Readmission within 30 days is associated with 1-year mortality in the community-dwelling Medicare population [
9], and we have previously demonstrated an association between readmission and 1-year mortality among prevalent ESRD patients [
10]. The relationship between early readmission among incident ESRD patients and longer-term outcomes remains unexplored. Patients who have had an early readmission may represent a high-risk subgroup that have a different disease course and mortality risk than the already elevated risk experienced by ESRD patients [
11]. Characterizing long-term outcomes among those who have had an early readmission may inform clinical decision-making and interventions to improve care for this substantial group. We used the United States Renal Data System (USRDS) to estimate the association between early readmission in the first year of dialysis and mortality, hospitalization, and transplantation in the second year among a cohort of incident hemodialysis patients who survived at least 1 year with primary Medicare coverage in the United States.
Discussion
In this analysis of the U.S. Medicare population initiating hemodialysis in 2010–2013 who survived at least 1 year on dialysis, nearly 1 in 5 experienced a hospital admission with 30-day readmission in the first year of hemodialysis. We found that 30-day readmissions were associated with increased risk of subsequent long-term outcomes in the second year of hemodialysis, including mortality, hospitalization, ICU utilization, and a lower likelihood of kidney transplantation. Patients with readmissions were at substantially higher risk of poor outcomes than either patients who had no admissions in the first year or patients who had a hospital admission with no readmission. Our findings imply that the detrimental effects of readmission are long-lasting and affect a large proportion of incident dialysis patients. To our knowledge, this is the first study to identify patients with hospital readmissions in their first year of dialysis as a group at high risk for poor long-term outcomes.
Our observed associations likely have complex mechanisms. One possible explanation for our observed association between hospital readmission and subsequent poor outcomes is that hospital readmission is a marker for patients who have a poor long-term prognosis. We found patients with readmissions in the first year of dialysis had higher rates of comorbidities than patients who were either not admitted or admitted but not readmitted. Chan, et al. [
3], also found that patients who were readmitted had higher rates of comorbidities, as well as more severe disease. While we adjusted for the presence of comorbidities in multivariable analyses, and performed additional sensitivity analyses using comorbidities from claims as well as the 2728 form, we did not adjust for all possible diagnoses, or for comorbidity severity. Therefore, there is likely residual confounding of our results by underlying patient health status.
Another potential explanation for our findings is that physiologic changes in ESRD patients during hospitalization contribute to poor long-term outcomes. Chan et al. found decreased levels of hemoglobin, albumin, phosphorus, calcium, parathyroid hormone and weight after hospitalization among dialysis patients; these lab values were inversely correlated with length of stay [
14]. The authors suggested that hospitalization could be described as an “acute inflammation-malnutrition syndrome” which impacts laboratory markers associated with mortality in ESRD. Readmissions result in increased time spent in the hospital, potentially magnifying the effect of malnutrition. Increased time spent in the hospital may also increase the risk of hospital-acquired infections, which are a major cause of mortality among ESRD patients [
1]. It is unclear from our study - or other previous studies - whether hospitalization or readmission events have a causal effect on long-term outcomes, or whether they are markers of poor prognosis, or both. The most effective approach to improving outcomes for these patients likely involves both attempts to prevent readmission and interventions to reduce mortality risk for patients who have been readmitted.
Patients who were Admit+ / Readmit+ in our study were less likely than those who were Admit-, or Admit+ / Readmit-, to receive a kidney transplant. Kidney transplantation is the preferred treatment for ESRD, conferring improved survival and quality of life to patients as well as decreased healthcare costs [
15]. However, not all patients benefit equally from transplant. A study by Lynch et al., found that patients who were frequently admitted to the hospital while on the waitlist for kidney transplant had increased waitlist mortality, increased healthcare utilization, and inferior graft and overall survival [
16]. Simply increasing access to transplantation among those with readmissions may not result in substantially improved outcomes without a focus on improving their underlying health status.
The readmissions rate in our study is lower than the USRDS estimate of 35% among prevalent patients; our analysis was at the patient-level, as opposed to the admissions-level, and focused only on incident hemodialysis patients who survived at least 1 year. Our observed rate of 19% was slightly lower than both the rates we reported in a previous study of the Medicare population (23%) [
10] and a study by Chan, et al., that included the non-Medicare population (22%) [
3]; this difference may be due to the fact that patients must have survived at least 1 year to be included in our cohort. While no other studies have directly examined the association between readmissions and long-term mortality, our estimated effect of hospital admissions alone on mortality (HR: 1.84, 95% CI: 1.78–1.90) was comparable to that observed in a study of long-term mortality among ESRD patients after ICU admission (HR: 2.32, 95% CI: 1.84–2.92) [
17].
Our results should be interpreted in the context of several limitations not mentioned above. First, there is the potential for misclassification of variables in administrative databases such as USRDS. Second, we did not have access to laboratory data during follow up, which may have provided insight into preliminary markers of poor outcomes. Third, our assessment of hospital admissions and readmissions began at 90 days after dialysis initiation; patients may have had hospital admissions prior to this 90-day window that were not captured in our assessment. However, our primary purpose was to assess the association of outcomes subsequent to readmissions after a period of stabilization following dialysis initiation, and sensitivity analyses examining readmissions from day 1 showed similar results to our primary analyses. Additionally, under CMS' Hospital Readmissions Reduction Program hospitals are incentivized to avoid readmission by keeping patients in the ED or in observation stays, and we did not include these encounters in our primary analysis. In fact, results from our sensitivity analyses including ED visits/observation stays suggest that the sicker patients with the worst outcomes are less likely to be kept in the ED and instead admitted. However, it is important to note that our results reflect the actual outcomes subsequent to readmissions, as they are occurring under current policies. Finally, although we controlled for many relevant covariates, there is the possibility of residual confounding by factors not well-captured in our data, such as socioeconomic status, laboratory data, and severity of comorbid conditions. One major strength of our analysis is the use of USRDS, a national database of incident dialysis patients on Medicare. As the vast majority of ESRD patients in the United States are eligible for Medicare, our results are likely to be generalizable to the population of U.S. hemodialysis patients.
While potential interventions have been identified to prevent readmissions after hospitalization among ESRD patients – for example, additional physician visits [
7] or hemoglobin monitoring [
14] – very little is known about preventing poor outcomes among patients once they have had a readmission. Further studies are needed to identify modifiable factors associated with poor long-term outcomes for ESRD patients with hospital readmissions in their first year on dialysis, who represent about 20% of patients starting hemodialysis. Identification of these modifiable factors is the first step to clarifying potential interventions and directing appropriate clinical resources to this high-risk group.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.