Abstract
Background
Cirrhosis is associated with significant healthcare utilization, yet data about in-hospital decompensations remain sparse. Additionally, the impact of liver transplant candidacy status on resuscitation and outcomes is largely unknown.
Aims
We aimed to evaluate the characteristics of resuscitation events for patients with cirrhosis with acute decompensation, analyzing liver transplant candidacy and intensive care unit (ICU) transfer parameters.
Methods
Retrospective single-center review of adult patients with liver cirrhosis who had a rapid response team (RRT) activation during hospitalization and no prior liver transplantation.
Results
Patients with cirrhosis who were liver transplant candidates were more likely to be younger (p = .003), have a higher serum total bilirubin (p = .015), higher INR (p < .001), and higher MELD 3.0 (p = .006). There was no significant difference in ICU transfer (p = .170) after RRT activation. Liver transplant candidates had a lower 30- and 60-day mortality (p = .008, p = .014) and were less likely to have a code status discussion after decompensation (p = .001). Lower serum albumin was associated with ICU transfer (p = .001). Patients who transferred to the ICU were more likely to have a code status discussion within 24 h after RRT (p = .011) without significant difference in 30- or 60-day mortality (p = .059, p = .277).
Conclusions
Liver transplant candidacy in patients with cirrhosis with acute decompensation is not clearly correlated with ICU transfer. Liver transplant candidates are more likely to be younger, have higher MELD 3.0 scores, less likely to have code status discussed after RRT, and have lower 30- and 60-day mortality rates. Patients who transfer to the ICU are more likely to have a code status discussion without any significant difference in 30- or 60-day mortality.
Keywords
Introduction
Cirrhosis is a significant cause of morbidity and mortality, with an estimated 1.47 million deaths globally in 2019, accounting for approximately 2.4% of worldwide deaths and representing a 10% increase from 2010.1,2 In the United States, cirrhosis prevalence and mortality have both increased since approximately 2009.3,4 Additionally, there has been an epidemiological shift to account for these increases, with general decline in cirrhosis due to viral hepatitis and increase in cirrhosis secondary to alcohol use and metabolic derangements.3,4 Hospitalizations for chronic liver disease have increased 25% from 2005 to 2015, and chronic liver disease tends to have the highest inpatient mortality out of all reasons for gastrointestinal hospital admissions with over 14 000 in-hospital deaths annually. 5 Chronic liver disease also accounts for over 325 000 Emergency Department visits annually and over 1 million clinic visits, representing a significant burden to the health care system. 5 Of note, terminology for liver disease has shifted over the previous decade, with both a transition from NAFLD to MASLD and the rise of the use of ACLF criteria, which continue to be heterogeneous.
In patients hospitalized with cirrhosis, acute decompensation events are estimated to occur in 10 to 30% and are the primary cause of hospitalization.6,7 Classically, decompensated cirrhosis is defined by the presence of clinical syndromes including hepatorenal syndrome, ascites, varices, hepatic encephalopathy, and/or jaundice. There are no clear evidence-based guidelines for which patients should have care escalated to the ICU during these decompensation events. However, when matched for illness severity, mortality rates of patients with liver disease admitted to the ICU were comparable to the mortality rates of patients admitted for other medical reasons, suggesting that criteria for escalation to the ICU should not differ based on etiology. 8
One study of cirrhotic patients admitted to the ICU found an in-ICU mortality of 42.7%, an in-hospital mortality of 54.1%, and a 6-month mortality of 75.1%. 9 Another study determined that patients who were admitted to a medicine floor prior to admission to the intensive care unit had higher mortality rates than those who were directly admitted to the ICU. 10 Additionally, the survival rates of patients with cirrhosis after ICU admission have increased in recent decades. 11 No differences in survival based on etiology of liver disease have been found. 12 Although data is limited on the most frequent causes of death in patients with cirrhosis, data suggest that upper GI bleed and sepsis are at the forefront, followed by hepatorenal syndrome and cerebrovascular and ischemic heart disease. 4
The high mortality rates of end-stage liver disease patients with decompensation events have raised questions about futility, particularly since admission to the ICU is most beneficial if the precipitating or underlying etiology of decompensation can be intervened upon, and many patients are not considered liver transplant candidates. 7 Of patients with cirrhosis admitted to the ICU, only about 3.8% undergo liver transplantation. 9 However, liver transplantation drastically improves survival, particularly in the most severely ill patients. 13
Despite the significant and increasing burden of chronic liver disease including prevalence, morbidity, mortality, and healthcare utilization, there are minimal data about in-hospital decompensations for this population, particularly in relation to escalation to intensive care units and liver transplant candidacy. In this study, we aimed to evaluate the characteristics of resuscitation events for patients with cirrhosis, looking specifically at liver transplant candidacy and ICU transfer parameters. We hypothesized that liver transplant candidates will have higher rates of ICU transfer and will have goals of care discussions more often following a Rapid Response Team (RRT) activation.
Materials and Methods
Study Design and Selection Criteria
This was a retrospective single-center review of adult hospitalized patients at two hospitals at a large academic institution from July 1, 2022 to July 31, 2023. Patients were included in the study if they were 18 years of age or older, had an International Classification of Diseases (ICD)-10 diagnosis of cirrhosis (K70.3, K74), and had a rapid response team (RRT) activation during admission. Patients were excluded if they did not meet the above criteria, if hospitalization was planned, if they had already undergone liver transplant prior to RRT, or if there was not a primary liver etiology for cirrhosis.
The first RRT called on each patient within the study time frame was assessed. Further RRTs called on the same patient were not included in this study, regardless of the time between the RRTs. Subsequent RRTs within the same hospitalization were rare, and these were excluded with the goal of avoiding the potential confounder of knowledge of previous RRT outcomes influencing clinician decision-making.
Data Collection
Demographics were extracted from the electronic medical record (EMR), including age at time of RRT, sex, and patient-reported race.
EMR data extraction was used to evaluate the etiology of cirrhosis (stratified as alcohol-related, autoimmune, chronic viral hepatitis, hemochromatosis, medications, MASLD/NAFLD, unknown, or other), severity of liver failure (as measured by MELD 3.0 using the most recent values prior to RRT; calculated using the OPTN calculator - https://optn.transplant.hrsa.gov/data/allocation-calculators/meld-calculator/), the individual laboratory components of the MELD score (serum total bilirubin, serum sodium, INR, serum albumin, and serum creatinine; most recent values prior to initiation of RRT were utilized) and liver transplant candidacy. Any patient who was actively undergoing liver transplant evaluation or was already on the liver transplant list at time of RRT was considered a liver transplant candidate. Any patient who was documented as not being a liver transplant candidate or who did not have transplant discussed at any point prior to RRT was considered to not be a liver transplant candidate.
Acute decompensation was defined as presence of an RRT event as indication of a significant change in clinical status, rather than by the more limited definition of decompensation often applied specifically to cirrhosis (eg jaundice, signs of portal hypertension, hepatorenal syndrome). The clinical measures of decompensation that prompted the RRT call were evaluated. This was documented per the standardized RRT indication list for the institution (acute bleeding, altered mental status, change in level of consciousness, chest pain, hypothermia, hypotension, increased FiO2, RN or MD concern, peer consultation, shortness of breath, stroke symptoms, tachycardia, tachypnea, other, no reason entered) and more than one indication could be selected in line with the process for RRT documentation. These were then grouped by similarity to facilitate statistical analysis, with overall RRT indication categories of cardiovascular (tachycardia, bradycardia, hypotension, chest pain), respiratory (increased FiO2, tachypnea, shortness of breath), neurologic (altered mental status, change in level of consciousness, stroke symptoms), and acute bleeding. The RRT categories of “RN or MD concern” and “peer consultation” were left out of final statistical analysis, as these were true of essentially all RRT events.
We evaluated the floor-level interventions performed during the RRT. Intervention was defined as administration of any fluid or medication, change in respiratory support, or further evaluation including blood tests, imaging, or EKG.
We also assessed if any ICU-related interventions were performed within 6 h of the RRT outcome decision if the patient was transferred to the ICU. This included initiation of renal replacement therapy (CRRT/HD), administration of vasopressors, initiation of invasive or NIPPV respiratory support, bronchoscopy, GI procedure (eg endoscopy, paracentesis), IR procedure, and cardioversion.
Outcomes
The primary outcome of interest was RRT disposition (ie transfer to ICU or not). This was abstracted from the EMR documentation of the RRT event, which clearly stated whether the patient was transferring to the ICU or not at the conclusion of the RRT.
Secondary outcomes included whether the patient underwent liver transplantation during the study timeframe, 30- and 60-day mortality after RRT event, and if a code status discussion was held with the patient and/or the patient's family within 24 h after initiation of the RRT event. Code status discussion was defined as change in code status, placement of an Advanced Care Planning (ACP) note, or explicit mention of “goals of care,” “care conference,” “code status,” or “prognosis” in clinician documentation. If code status was discussed, we analyzed if this discussion resulted in a change in code status.
Statistical Analysis
Descriptive analyses, demographics, and clinical characteristics were reported as median (quartile 1, quartile 3) and/or frequency (percentage). Continuous variables including age, laboratory values, MELD 3.0 score, and number of RRT interventions were analyzed using the Kruskal-Wallis test. Pearson's Chi-square test was used for comparison among groups of categorical variables including sex, race, etiology of cirrhosis, RRT indications, RRT interventions, transfer to the ICU, mortality, and code status discussion.
Univariate logistic regression was used with p-value of .05 or less considered significant. Two separate analyses were performed: variables compared based on transplant candidacy and variables compared based on RRT disposition.
Data extraction was performed using REDCap software (Version 13.7.31). All analysis was performed using BlueSky Statistics (Version 10.3.1-Pro, licensed to Mayo Clinic). This study was approved by our hospital institutional review board with appropriate consents obtained.
Results
Study Cohort Characteristics
88 patients with cirrhosis had RRT activations within the study period, of those, 66 patients met the inclusion criteria and were included in the data analysis. Of those, 37 were female (56.1%) and 29 were male (43.9%). Median age was 59.5 years (Q1 52.0, Q3 67.0). The most common etiology of cirrhosis was alcohol-associated liver disease (53.0%), followed by MASLD/NAFLD (27.3%) and autoimmune (10.6%). Median MELD score was 27.5 (Q1 19.0, Q3 34.8). Further demographics based on transplant candidacy and rapid response team outcome can be seen in Tables 1 and 4 respectively; there were no statistically significant differences between groups for sex, race, or cirrhosis etiology; there was a statistically significant difference in age between transplant candidacy groups (p = .003). 27 patients were transplant candidates (40.9%) and 39 were not (59.1%). Following RRT activation, 36 patients transferred to the intensive care unit and 30 remained general care status.
Baseline Characteristics Stratified by Liver Transplant Candidacy. Level of Significance p < .05 (Kruskal-Wallis Tests for Continuous Variables and Pearson's Chi-Square Test for Categorical Variables).
*Statistically significant.
Transplant Candidacy Status
Patients who were liver transplant candidates were more likely to be younger (p = .003). Sex, age, and race were not significantly different between groups. Several laboratory components of the MELD 3.0 score were different between groups, with higher bilirubin (p = .015), higher INR (p < .001), and higher albumin (p = .009) in liver transplant candidates; there was no significant differences in serum sodium and serum creatinine between groups. MELD 3.0 score was higher in liver transplant candidates, with a median of 31 in liver transplant candidates and a median of 22 in patients who were not transplant candidates (p = .006). Results are summarized in Table 1.
Liver transplant candidates were more likely to have RRT activation for cardiovascular causes (p = .039) and patients who were not transplant candidates were more likely to have RRT activation for acute bleeding (p = .050). There was no statistically significant difference in the number of floor-level or ICU-level interventions based on transplant candidacy. Additionally, there were no significant differences between implementation of specific RRT floor-level or ICU-level interventions based on transplant candidacy, except for EKG being more likely in patients who were liver transplant candidates (p = .008) and GI procedure (including upper endoscopy, paracentesis) being more likely in patients who were not transplant candidates. Results are summarized in Table 2.
Rapid Response Team Activation Indications and Interventions Stratified by Liver Transplant Candidacy. Level of Significance p < .05 (Kruskal-Wallis Tests for Continuous Variables and Pearson's Chi-Square Test for Categorical Variables).
*Statistically significant.
There was no difference in ICU transfer based on liver transplant candidacy (p = .170). However, patients who were liver transplant candidates were more likely to undergo transplant (p < .001) and more likely to be alive 30 and 60 days after the decompensation event (p = .008, p = .014). Patients who were not transplant candidates were more likely to have a code status discussion within 24 h after decompensation event than patients who were transplant candidates (p = .001). Table 3 provides results.
Outcomes Stratified by Liver Transplant Candidacy. Level of Significance p < .05 (Kruskal-Wallis Tests for Continuous Variables and Pearson's Chi-Square Test for Categorical Variables).
*Statistically significant.
Rapid Response Activation Disposition
There were no statistically significant differences in demographics between patients who were transferred to the ICU after RRT and patients who remained on the floor, including age, sex, race, and cirrhosis etiology. Additionally, there was no difference in transplant candidacy based on RRT disposition (p = .170). Of the laboratory values for MELD 3.0 calculation, only serum albumin was different between groups, with a lower serum albumin in patients transferred to the ICU (p = .001). Serum total bilirubin, serum sodium, INR, serum creatinine, and MELD 3.0 were not statistically different between groups. Results can be seen in Table 4.
Baseline Characteristics Stratified by Rapid Response Team Activation Outcome (ie Transfer to ICU or not). Level of Significance p < .05 (Kruskal-Wallis Tests for Continuous Variables and Pearson's Chi-Square Test for Categorical Variables).
*Statistically significant.
As demonstrated in Table 5, neurologic indication for RRT activation was negatively associated with ICU transfer (p = .007); cardiovascular, respiratory, and acute bleeding indications were not different between groups. Additionally, the number of floor-level interventions and the specific floor-level interventions were not statistically associated with RRT outcome, except for administering “other medications” (ie not diuretic, bronchodilator, antiarrhythmic, analgesic, or antibiotic) associated with patients remaining on the floor (p = .036).
Rapid Response Team Activation Indications and Interventions Stratified by Rapid Response Team Activation Outcome (ie Transfer to ICU or not). Level of Significance p < .05 (Kruskal-Wallis Tests for Continuous Variables and Pearson's Chi-Square Test for Categorical Variables).
*Statistically significant.
Transfer to the ICU was not associated with increased likelihood of liver transplantation (p = .198). Additionally, there was no statistically significant difference in mortality between transfer to the ICU and remaining on the floor at 30 or 60 days after RRT (p = .059, p = .277). Code status was more likely to be discussed after a decompensation event if the patient transferred to the ICU (p = .011). Table 6 shows these results.
Outcomes Stratified by Rapid Response Team Activation Outcome (ie Transfer to ICU or not). Level of Significance p < .05 (Kruskal-Wallis Tests for Continuous Variables and Pearson's Chi-Square Test for Categorical Variables).
*Statistically significant.
Discussion
In this study, we characterized differences in acute decompensation events in patients with cirrhosis between those who were potential or confirmed liver transplant candidates and those who were not transplant candidates. Compared to national data on liver transplant candidacy in the United States, our transplant candidates were more predominantly female (59.3% vs 38.7% nationally in 2021), more predominantly white (85.2% vs 68.9%), and more predominantly had cirrhosis secondary to alcohol abuse disorder (59.3% vs 36.3%). 14 These discrepancies are likely a result of the single-center nature of this study. As expected, we found that transplant candidates were more likely to be younger and tended to have more severe liver disease as measured by laboratory values including serum total bilirubin and INR and as measured by MELD 3.0 score. Median MELD 3.0 score for liver transplant candidates was 31, reflecting the severity of disease in our population as would be expected at a large academic referral center. However, there was no difference in transplant candidacy on the bases of sex, race, or cirrhosis etiology in our dataset. This was a reassuring finding, suggesting that progress has been made in recent years towards more equitable allocation. Approximately 81.8% of the dataset identified as white, which is comparable to the overall racial demographics of the state the study was conducted in. Further investigation, including a larger dataset and more longitudinal analysis, could help further support or refute these findings.
Our study found that liver transplant candidates were more likely to have a rapid response activation for cardiovascular concerns, of which hypotension was the most common reason for activation. Additionally, patients who were not liver transplant candidates were more likely to have an RRT activated for acute bleeding. These results are consistent with the significant differences in RRT interventions; EKG is more likely to be performed in liver transplant candidates and a GI procedure such as endoscopy is more likely to be performed in non-transplant candidates. There is no clear reason for these differences; further investigation would be needed.
We hypothesized that liver transplant candidates would be more likely to transfer to the ICU as the result of a decompensation event; we felt that knowing a patient's potential for receiving therapy to address the underlying process (ie transplant) would influence the RRT clinician's decision-making process and make them more likely to go to more extreme measures (ie ICU escalation) for a patient who had a greater chance of receiving life-saving therapies. However, we instead found that there was no difference in transfer to the ICU based on liver transplant candidacy. This suggests that clinician decision-making during decompensation events is more focused on the patient's current clinical status than on any perceived futility of care, which is particularly reassuring given recent studies showing that patients with cirrhosis have similar ICU mortality outcomes to patients admitted to the ICU for other medical reasons. 8
As expected, patients who were liver transplant candidates were more likely to undergo liver transplant and had an improved mortality. This is consistent with known significance of liver transplant on patient outcomes.
We also analyzed patient characteristics and RRT parameters between patients with cirrhosis who transferred to the ICU as a result of an RRT and those who did not, looking for any potential predictors of ICU transfer. Age, sex, race, cirrhosis etiology, and liver transplant candidacy were not significantly associated with ICU transfer, again suggesting that RRT clinicians focus more on a patient's clinical status than their demographics. From a clinical parameter perspective, serum albumin was the only laboratory parameter analyzed that was correlated with ICU transfer; patients with lower albumin were more likely to transfer to the ICU. However, there was no significant difference in MELD score between those who transfer and those who do not, suggesting that the features of the acute decompensation event may be more predictive than the severity of the patient's underlying disease. Further analysis of patient's clinical status at the time of RRT (eg vital signs, non-MELD labs) would be needed to confirm or refute this hypothesis.
An RRT called for neurologic concerns was more likely to result in the patient remaining on the floor. This is likely due to the prevalence of hepatic encephalopathy in patients with cirrhosis, and the ability to monitor and safely treat this on the floor in many situations.
Interestingly, there was no difference in liver transplantation rates or mortality between patients who transferred to the ICU and those who did not. It is possible that this reflects the success of the RRT system at triaging appropriately; patients who needed ICU level of care received it and had no mortality deficit and vice versa.
We also analyzed the occurrence of code status discussion in the 24 hours after RRT. Patients who were not transplant candidates were far more likely to have a code status discussion after a decompensation event. The relative lack of code status discussions for liver transplant candidates is likely due to the requirement that liver transplant candidates are full code during the transplant process in our facility; the clinical focus is on reaching the potentially lifesaving transplant. Code status discussions were also more likely if a patient transferred to the ICU due to a decompensation event. While this may be reflective of more severe illness prompting more urgent goals of care discussions, it also suggests that there is room for improvement in addressing a patient's illness severity and goals of care with them if they remain on the floor.
Limitations of this study include the single-center nature; this likely limits generalizability, particularly since this study was conducted at a large academic liver transplant referral center where patients may be more critically ill. However, no patients in this study were transferred from another institution for the admission analyzed. Additionally, this was conducted over two hospitals within the same health system (one with a liver transplant service and one with only a general gastroenterology service), which may add to generalizability. Approximately 81.8% of the cohort studied identified as white, which may also limit generalizability. In addition, since only the first RRT of each hospitalization was analyzed, it is possible that some subsequent ICU transfers were missed if a patient had multiple decompensation events within the same hospitalization.
ICU capacity may also have impacted triage decisions at the time of RRT and therefore impacted study results, particularly since decisions for escalation of care are multifaceted and may vary between clinicians and institutions. However, our study timeframe was carefully chosen to not include times of significant ICU strain secondary to COVID-19 to avoid ICU capacity restraints on decision of disposition after RRT activation.
Our clinical parameters during RRT only evaluated the general indication for RRT and MELD-associated laboratory values, which may also limit full understanding of predictors of ICU transfer. However, MELD-associated laboratory values are likely the most pertinent for patients with decompensating cirrhosis.
Conclusions
Contrary to our initial hypothesis, we demonstrate in this study that liver transplant candidacy in patients with cirrhosis and acute decompensation events is not clearly correlated with ICU transfer. Patients who are transplant candidates are more likely to be younger, have higher MELD 3.0 scores, less likely to have code status discussed after rapid response activation, and have lower 30- and 60-day mortality rates. For patients with cirrhosis and RRT activation, a lower serum albumin is the only clear predictor of ICU transfer. Patients who transfer to the ICU are more likely to have code status discussion but do not have any significant difference in 30- or 60-day mortality.
Footnotes
Abbreviations
Acknowledgements
None
Other Support
Grant support for REDCap data extraction software was utilized (UL1TR002377).
Author's Contributions
JF and AG contributed to study design; JF and MM contributed to data extraction; JF and AG contributed to data review and interpretation; JF contributed to manuscript preparation; all authors contributed to manuscript editing and finalizing.
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.
