Using the largest national readmission database in the United States, we show that the 30-day readmission rate after liver transplantation surgery was 30.6% based on 25,054 patients. Between 2010 and 2014, the number of liver transplantation surgeries has increased among patients older than 65 years of age and decreased among those between 40 and 64 years old. Over the same time period, both the inflation adjusted total index hospitalization and calendar year hospitalization costs have also increased significantly. The majority of 30-day readmissions were due to post transplant complications, with packed red blood cell transfusions being the most common intervention during readmission. Independent predictors of 30-day readmission were type of insurance, low and medium volume centers, hemodialysis, liver biopsy, infection, and prolonged LOS. Independent predictors of calendar year in-hospital mortality were 30-day readmission, age older than 64 years, non-alcoholic cirrhosis, and prolonged LOS.
30-day readmission is an important marker of quality of service provided during the index admission. Reduction of hospital readmissions is a declared goal in the United States economic and quality improvement agenda. Previous single center studies reported 30-day readmission rates after liver transplantation of 45%3,30. Wilson et al., in a study based on UNOS database, reported a 30-day readmission rate of 38% in this setting. Our analysis revealed a stable trend of annual 30-day readmission rate over the study period of 32–29%, which is lower than that reported in those earlier studies. This improving trend in 30-day readmission rate could be due to the adoption of preventative strategies over time. Russo et al., in a single center study, has reported a reduction in 30-day readmission rate after liver transplantation from 40 to 20% after the implementation of a multi-faceted readmission prevention strategy4. The strategy included (1) using observation status instead of inpatient admission when patients sought medical attention after LT and (2) performing same-day or in-office endoscopic retrograde cholangio-pancreatography (ERCP). The telemedicine based readmission prevention protocol proposed by the University of Pennsylvania has significantly decreased the post LT 30-day readmission rates from 32 to 16%17. While these and similar interventions can serve as a blue print for individual institutions to build their readmission prevention strategy after, their national reach and generalizability are yet to be determined.
We also report an increasing number of LT among older patients (65 years of age or older), along with an increase in index admission and calendar-year total hospitalization costs and requirement for hemodialysis. Although our study was not designed to identify the reason behind these findings, a possible hypothesis is that the transplant recipients being older and having more comorbidities as is reflected by the increasing requirement for hemodialysis derive the increasing hospitalization costs18. In addition, as opposed to the child Pugh Score, creatinine is a component of the MELD score. Therefore, after switching to the MELD score as basis of liver allograft allocation, a larger proportion of patients on the transplant waiting list had renal failure and needed hemodialysis19. This increase in the pool of patients with renal failure eligible for liver transplantation might also partially explain the increased rate of hemodialysis and costs post LT.
Our results show that patients in the age groups 18–39 and 40–64 experienced higher 30-day readmission rates compared with patients older than 65 years. On the other hand, Patel et al. found that younger age was associated with lower 90-day readmission rates3 after LT. The difference in the results obtained from the two studies can be explained by differences in study designs. Specifically, the study by Patel et al. was a single center study that included 325 patients who received LT between 2005 and 2015. Our study included 25,054 patients from around 2000 hospitals across the United States. While the exact reason for the decreased 30-day readmission rates among older patients is beyond the scope of the current analysis, several studies indicate that a less active immune system might actually benefit older patients receiving LT due to decreased risk of acute rejection20. Other possible causes might be poorer compliance with medical instructions and possibly poorer social support system among younger patients.
The association between prolonged LOS and early readmissions after LT we report is consistent with previous literature. Pereira et al. found that index LOS shorter than 9 days was associated with a lower 30-day readmission rate, LOS between 9 and 17 days was associated with higher 30-day readmission rate until the cut-off point of 17 days, after which 30-day readmission rates became lower again2. The authors concluded that patients with optimal health likely made up the bulk of the early discharges (LOS less than 9 days) and patients who needed medical optimization may account for the LOS longer than 17 days2. In our study more than half of the patients had a post-LT LOS of 11 days. Therefore, based on our and previous studies’ results, we propose the implementation of protocols focusing on reducing LOS during index admission as a measure to decrease 30-day readmission (4). However, the causal relationship between LOS and readmission rate can be complex, since prolonged LOS can lead to debility and thus increase readmission rate. It can also be a marker of more severe disease/higher comorbidity burden which in its turn is the true cause of higher readmission rate.
The median number of days from index discharge to first readmission was 8.1 days (IQR 3.6–15.6). Therefore, we speculate that some of the reasons for readmission could have been identified at the time of discharge or addressed as outpatient.
Acute renal failure, infection, need for a liver biopsy and discharge disposition are potentially modifiable predictors of early readmissions. Infection has been found to be a predictor of readmission in multiple studies7,21. Infections after LT is attributable to multiple factors including affection of immunogenic organ (liver), receiving an organ with variable cold ischemia time, being in the hospital, immunosuppressive medications, etc.22. All these factors put LTRs to risk of multiple drug resistant bacterial infections (particularly pneumonias, wound infections, cholangitis and bacteremia), as well as viral reactivation (HSV and CMV) and fungemia23,24,25. Liver biopsy is usually performed in the work up of abnormal liver function tests. Those can be due to acute cellular rejection, vascular thromboses, ischemic graft dysfunction or delayed graft function. Therefore, liver biopsy is a marker for these conditions, and implementing protocols to prevent these conditions rather than avoiding liver biopsy itself can potentially help in decreasing 30-day readmission. Approximately 15% of patients were discharged to various outpatient facilities such as skilled nursing facility (SNF), intermediate care facility (ICF) or another type of medical facility. We found that discharge disposition of transfer is an independent predictor of 30-day readmission. Contrary to our finding, Kothari et al. in a study based on 3072 LTRs reported that discharge to skilled nursing facility and inpatient rehabilitation are protective against 30-day readmissions when compared to discharge home or with home health26. However, Wilson et al. in a larger study found that disposition to these facilities was indeed a predictor of 30-day readmission6. In our study based on NRD, it is assumed that majority of patients return to the same hospital for admission in the immediate post-transplant period, even those living out of state.
Our study is the first to link the type of insurance provider with risk of 30-day readmission. We found Medicare and Medicaid insurance to be independent predictors of 30-day readmission after LT. This finding adds to the growing body of data that establish an association between insurance and disparity in treatment outcomes of multiple medical and surgical conditions. Medicaid insurance and no insurance have been shown to be associated with worse outcomes in the treatment of non-variceal upper gastrointestinal hemorrhage and LT27,28. In addition, Medicare/Medicaid primary payer status was also shown to be a 30-day readmission predictor in the post-coronary artery bypass graft population29. Along the same line, a national study by Nguyen et al., reviewing the NRD for 2013, revealed Medicaid and Medicare primary care payers as independent predictors of high annual burden and costs of hospitalization for patients with chronic gastrointestinal and liver diseases that are high utilizers of the healthcare system30.
There are several studies addressing the association of early readmission with annual hospital LT volume. The majority of those studies have shown better surgical outcomes in liver transplant centers with higher procedure volume13,14,31,32. In addition high procedure volume was associated with lower hospital resource utilization. The results of these studies are consistent with those of our national analysis. They suggest that limiting complex surgeries like liver transplantations to centers of excellence with medium to high yearly procedure volume may decrease costs and healthcare resource utilizations.
Our study has certain limitations based on nature of administrative database for research. First, this database relies on ICD-9-CM diagnostic coding and hence is predisposed to inaccurate entries or missing data33. However, the rate of missing data among the variables we used was less than 2.0%. In addition, ICD-9-CM codes have been shown to have a high specificity and sensitivity when used to study gastrointestinal diseases34. Second, factors such as medication use including immunosuppressants, objective laboratory values and radiology test results are not included in NRD. Donor related information is also not part of NRD. Therefore, polypharmacy, non-compliance with medications and specific immunosuppressive regimens could not be included in our analysis. Further studies with databases that include these factors are needed to clarify their specific contribution to 30-day readmission after LT. Third, for the same reason of limited data variables, we were not able to assess the severity of liver diseases by using the MELD score and the Child–Turcotte-Pugh (CTP) score, which are known predictor of mortality in patients with cirrhosis. However, we used instead well-validated and widely accepted cirrhosis severity classification criteria: the Baveno criteria. The Baveno criteria have been previously used for the stratification of liver cirrhosis severity using administrative databases35,36. Fourth, since NRD captures only in-hospital mortality, the 30-day and calendar-year mortality we report might be an underestimate of the true mortality rate. This is because patients who died at home, en-route to the hospital or in the emergency department were not included in our analysis. Lastly, inherent to the NRD’s dependence on state-based data (SID), readmissions occurring in a hospital located in another state could not be captured. However, it is assumed that majority of patients return to the same hospital for admission in the immediate post-transplant period, even those living out of state.
Despite these limitations, our study has several strengths. This study is the most recent, to our knowledge, reporting the 30-day all-cause readmission rate post liver transplantation, its predictors and its impact on treatment outcomes at the national level in the United States. The largest publicly available all-payer readmission database in the United States is used that minimized the likelihood of a beta error. Most importantly, the NRD is nationally representative and includes patients from hospitals that are small, medium, and large; rural & urban; privately or publicly owned; teaching & non-teaching; and for profit & not for profit, across 18–22 states. This makes the study results more readily generalizable to the United States. Furthermore, the unique variables in the database permitted us to explore factors such as hospitalization costs, household income estimates, insurance carrier and hospital factors, which are not commonly available in single-center and UNOS database studies.
In conclusion, this is the first study based on nationwide readmission database to determine the 30-day all cause readmission rate after liver transplantation, as well as its risk factors and impact on patient’s outcome. We found that the 30-day readmission rate is 30.6%. Between 2010 and 2014, the number of liver transplantation surgeries has increased among patients older than 65 years of age and decreased among those between 40 and 64 years old. Over the same time period, both the inflation adjusted total index hospitalization and calendar year hospitalization costs have also increased significantly. The majority of 30-day readmissions were due to post transplant complications. Independent predictors of 30-day readmission were type of insurance, low & medium volume centers, hemodialysis, liver biopsy, infection, and prolonged LOS. Independent predictors of calendar year in-hospital mortality were 30-day readmission, age older than 64 years, and prolonged LOS.
Early readmission not only increases economic burden on healthcare, but is also associated with increased calendar year mortality. The data we present on early readmission, its predictors, trends of liver transplantation and healthcare utilization can be helpful to the patients, clinicians, payers, and transplantation policymakers. Addressing potentially modifiable predictors of readmission, such as type of insurance, center volume, requirement of HD and discharge disposition may reduce the readmission rate. Although other predictors may not be modifiable, they can be used to identify patients at high risk of readmission, and who would thus benefit the most from interventions aimed at decreasing 30-day readmissions. Future interventions for all patients, and especially those at high risk for readmission, have the potential to decrease readmission rates and associated health care expenditures, as well as improve morbidity and mortality from liver transplantation.