As the debate over healthcare costs intensifies and the issue of hospital readmission rates gets increased regulatory scrutiny, it is important to identify patients who may benefit from targeted interventions to decrease the risk of readmissions. Our study makes three important contributions to the field. First, to the best of our knowledge, our study is the first to show that in a cohort of patients (cirrhotics) at high risk of hospitalization, a relatively small group accounts for a disproportionate number of readmissions. Second, we demonstrate that racial disparities may be contributing to the risk of readmissions in patients with cirrhosis. We also corroborate previous studies that MELD score and diabetes are factors increasing risk of readmission in patients with cirrhosis. Third, we establish that cirrhosis patients with frequent readmission have significant barriers to transplantation that could potentially be targeted to increase access to transplantation and mitigate the high readmission rates.
Recent studies demonstrate that early re-hospitalizations among patients with decompensated cirrhosis are common, with 20–37% of patients being readmitted within 1 month of discharge 
. Early readmission was associated with MELD score, diabetes, and male gender, while the time to readmission in another study was predicted by MELD score, serum sodium level, and number of medications on discharge 
. In patients with hepatitis C-related cirrhosis, readmission rate at 1-year was 45% with a remarkably high admission rate of 73% among patients with hepatic encephalopathy 
. Taken together, these studies strongly suggest that the prevention of hospital readmissions in patients with end-stage liver disease represents a major unmet healthcare need deserving of the same level of attention from the healthcare community that has been accorded to patients with chronic obstructive pulmonary disease or congestive heart failure. Chronic disease management is now a widely accepted to address chronic disease outcomes in patients with heart disease and diabetes 
. However, while inconsistency of care of cirrhosis has been well-recognized and quality indicators have been put forth for cirrhosis care 
, chronic disease management in cirrhotics has not been well studied or established. The goal of our study was to identify cirrhosis patients at high risk of frequent hospital admissions and identify potential barriers to liver transplantation.
That patients with decompensated cirrhosis have high readmission rates would by itself not be surprising. However, a significant finding of our study was that a relatively small number of patients (not necessarily the medically sickest patients as assessed by conventional scoring criteria) accounted for a disproportionately large percentage of hospital admissions in this population. We found that MELD score was a strong predictor of frequent hospital admissions, a finding that is consistent with prior reports and with the MELD score as being a validated predictor of prognosis in patients with cirrhosis. In addition, race and diabetes were associated with frequent admissions. Prior studies have shown that African-Americans having the highest rates of preventable hospitalization for congestive heart failure, diabetes, and hypertension 
. Whether these findings are also true for patients with cirrhosis will require further research, including controlling for potential confounders of educational and socioeconomic status 
. Diabetes alone does not explain the racial disparity in readmission; given that race was still a factor despite controlling for diabetes. Given the increased risk of frequent admissions in patients with diabetes, our results raise the question whether better control of diabetes could be a potential intervention in this population to decrease hospitalization. Prior studies have shown that diabetes is also a risk factor for hepatic encephalopathy and may have contributed to frequent hospitalizations in these patients 
. Our study has implications for clinical practice for healthcare providers as well for hospital administrators, policy makers, and health insurance companies who make decisions regarding allocation of resources towards the care of these patients.
Since liver transplantation is now considered an important therapeutic option in patients with advanced liver disease, we were interested in determining potential barriers to transplantation in the group of patients at the highest risk of admissions. Not surprisingly, we found that this group had significant barriers to transplantation, including active alcohol use and addiction issues. Thus, our results raise the possibility that targeting this group for early and aggressive intervention directed at addiction and psychiatric comorbidities may be an option to decrease hospitalizations and increase access to transplantation. Interestingly, several patients in this group, despite their obviously advanced liver disease, had low MELD scores (<15). Since MELD score of great than 15 is required in the US for deceased donor transplantation, this subgroup represents unique challenges to manage their disease. Thus, it is intriguing to speculate whether interventions such as early transjugular intrahepatic portosystemic shunt placement in patients with cirrhosis, special considerations in the organ allocation system, or live donor liver transplant in patients with ascites would decrease the frequency of admissions in this group of patients and provide overall more cost-effective care 
. Our study identifies groups of patients who are ineligible for liver transplant and at high risk for admissions towards which specific, personalized interventions can be targeted to minimize risk of hospitalization. Indeed, insights gained from our study has led to changes in practice patterns at our institution, including early referral to drug and alcohol treatment programs, aggressive management of fluid status in patients identified on their index admission to be at risk of readmission, and more timely referrals to palliative care.
Some limitations of our study should be considered. We did not have data on education or socioeconomic status to control for potential confounders in terms of the differences we saw in terms of race. Since the goal our study was to discover risk factors for multiple readmissions, time-based analyses could not be used. Therefore, we controlled for non-uniform follow-up interval by including follow-up time in the model. In order to minimize the differences in follow-up time, patients with less than 90 days of follow-up were excluded from the study. Excluding these patients could potentially lead to bias, therefore, we also analyzed the data by including the group with shorter follow-up times and the predictive factors remained the same and just 3 additional patients were included in the hyper-admission group. The choice of 5 admissions for the cut-off was based on the both on the mean number of admissions per person year as well as the fact that this group accounted for half of the total admissions in the study group. These results are of particular interest because we show that targeting this small group of patients and optimizing their transplant status could potentially decrease the risk of half of all admissions.
In summary, a relatively small number of patients with end-stage liver disease account for a disproportionately large number of hospital admissions. Diabetes, race, and MELD score were predictive of high risk of admissions. In this group of patients at the highest risk of admissions significant barriers to transplant exist. Interventions designed to address these factors, including low MELD score despite having decompensated cirrhosis and addiction issues could potentially increase access to transplantation and decrease frequency of hospitalizations.