Renal Trajectory Patterns Are Associated With Post-Discharge Mortality in Patients with Cirrhosis and Acute Kidney Injury

Abstract

Background & Aims:

Little is known about long-term outcomes of acute kidney injury (AKI) in patients with cirrhosis. Outcomes can vary with stage of AKI, chronic kidney disease (CKD), and trajectory of renal function.

Methods:

We collected data from the Department of Veterans Affairs and identified 6917 patients with cirrhosis who developed AKI during hospitalization at any of its 127 hospitals, from 2004 through 2014. We used latent class analysis of serial creatinine measurements during index hospitalization to determine trajectories of renal function.

Results:

Overall, 32% of patients died within 90 days of discharge from the hospital and 48% of patients died within 1 year. We identified 5 distinct in-hospital renal trajectories: mild AKI with full improvement (24.8% of patients died within 90 days), severe AKI with rapid improvement (24.7% of patients died within 90 days), moderate AKI with partial improvement (33.7% of patients died within 90 days), moderate to severe AKI with partial improvement (42.0% of patients died within 90 days), and severe AKI with minimal improvement (48.0% of patients died within 90 days). Trajectories were significantly associated with mortality within 90 days and 1 year of mortality. Patients with severe AKI with minimal improvement had the highest risk of death within 90 days (adjusted odds ratio, 3.08; 95% CI, 2.54-3.72) and within 1 year (adjusted odds ratio, 2.71; 95% CI, 2.25-3.27) compared with patients with mild AKI with full improvement. The highest 90-day post-discharge mortality (65.2%) was observed in patients with normal or near-normal prehospitalization renal function who developed severe AKI with minimal improvement during hospitalization.

Conclusions:

In an analysis of almost 7000 veterans with cirrhosis who were hospitalized for AKI, we found the pattern of renal trajectory to associate with mortality after discharge. Renal trajectory patterns can be used to identify subgroups of patients with cirrhosis and AKI who should receive intensive post-discharge management.

Meeting Abstract:

Mindikoglu AL, Hernaez R, Liu Y, Kramer JR, Taylor T, Kanwal F. Long-Term Outcomes in Patients with Acute Kidney Injury: A Retrospective Cohort Study. Gastroenterology 2019, Vol. 156, Issue 6, S-1348. Abstract has been presented on May 21, 2019 at Digestive Disease Week (DDW) 2019, San Diego, CA. Control ID: 3148648.

INTRODUCTION

Acute kidney injury (AKI) is one of the most common life-threatening complications of cirrhosis.¹ Although several studies have examined the clinical course and outcomes of patients with cirrhosis and AKI, most of them reported short-term in-hospital mortality with limited data on outcomes of patients who survived hospitalization.^1–3 Identifying the longer-term outcomes can help inform the timing of heightened surveillance after hospital discharge. These data can also guide healthcare professionals and hospitals to more efficiently align their interventions designed to reduce adverse outcomes in AKI patients with the highest risk of death.

Although AKI is associated with a high risk of mortality, this risk is not similar across all patients. For example, several studies showed significantly higher short- and long-term mortality in patients with AKI and chronic kidney disease (CKD) compared with patients with AKI alone.^4–7 With the aging population and rising prevalence of cardiovascular risk factors and non-alcoholic fatty liver disease, the burden of CKD is likely to increase in patients with cirrhosis. Furthermore, the severity of AKI at the time of hospitalization is a strong determinant of short-term outcomes. Changes in serum creatinine (Cr) during hospitalization also predict outcomes in patients with AKI. However, there are limited data on the patterns of renal function in patients with cirrhosis and AKI, their frequency, and how they associate with longer-term outcomes independent of other predictors such as demographics, liver disease severity, and comorbidities.

In this study, we aimed to determine the long-term survival of patients with cirrhosis discharged following hospitalization for AKI and identify the effect of CKD, renal trajectory patterns, and other risk factors to predict 90-day and 1-year mortality.

METHODS

Study Data and Patient Population

We acquired data from Veterans Affairs (VA) Corporate Data Warehouse (CDW) using VA Informatics and Computing Infrastructure (VINCI) workspace.⁸ We used VA CDW outpatient, inpatient, vital status (includes the date of death)⁹ and purchased care (includes services paid but provided outside of VA) files. VA Vital Status file combines data from Medicare, VA, Social Security, and VA Compensation and Pension Benefits to determine the date of death (sensitivity 98.3% and specificity 99.8% relative to National Death Index).⁹

We first identified patients hospitalized with cirrhosis defined by more than one ICD-9 code for cirrhosis (571.2, 571.5, and 571.6) or its complications (varices [456.0, 456.1, 456.2, 456.20, 456.21], hepatic encephalopathy [572.2, 572.20], portal hypertension [572.3], hepatorenal syndrome [572.4], other sequela of chronic liver disease [572.8], ascites [789.5], or spontaneous bacterial peritonitis [567.23]) between 01/01/2004 and 12/31/2014.¹⁰ For patients with multiple cirrhosis-related hospitalizations in the study time frame, we sampled patients’ first hospitalization for this study.

We excluded patients whose length of stay was either less than 24 hours or longer than 30 days; the former likely represented short-term, elective hospitalization, whereas the latter might have been related to rehabilitation or longterm care. We also excluded patients who had liver or kidney transplant or were on dialysis before or during the index admission as immunosuppressive medications, and dialysis can affect serum Cr levels.

We identified patients with AKI if they had at least 0.3 mg/dl increase in serum Cr from the baseline serum Cr based on the revised consensus recommendations of the International Club of Ascites.¹¹ We used serum Cr value obtained within three months before index hospitalization and used the one closest to the admission date as baseline serum Cr, as previously described.³ For patients who did not have any serum Cr within three months before hospitalization, we used the value obtained within 24 hours of admission as baseline serum Cr. For each patient, we selected the second serum Cr measured within 48 hours of the baseline serum Cr. We excluded patients who did not have a second Cr value to determine the change from baseline. For patients who had multiple values within these 48 hours, we used the highest one to define AKI. From the cohort with AKI, we excluded patients who did not have at least two serum Cr values following the diagnosis of AKI from this study. Lastly, because our goal was to identify patients who might need heightened surveillance after hospital discharge, we limited the study to patients who were discharged alive.

We used Kidney Disease Improving Global Outcomes (KDIGO) criteria¹² to define CKD. We classified patients as having CKD if they had at least two estimated glomerular filtration rate (eGFR) values < 60 ml/min/1.73m² at least 90 days apart within one year before index hospitalization or had > one ICD-9 code for CKD within one year before index hospitalization. We calculated GFR by using the 4-variable Modification of Diet in Renal Disease (MDRD) Study equation.¹³

Further details related to Methods are included in Supplementary Materials.

Outcome Measures

The primary outcome was mortality within 90-days and 1-year following hospital discharge. We defined the beginning of the 90-day and 1-year period from the date of discharge. Because the first 90 days after discharge may correspond to the highest risk period following hospitalization and interventions, we focused on 90-day mortality in the primary analyses. We chose 1-year in addition to 90-day as our study timeframe because it represents the standard timeframe used in studies of other chronic medical conditions such as cardiovascular diseases.¹⁴

Statistical Analysis

Descriptive statistics were presented as median (interquartile ranges) or mean (SD) and frequency (percentage) for continuous and categorical variables, respectively.

We used a group-based trajectory modeling (GBTM)/semi-parametric mixture modeling approach to determine whether there are distinctive in-hospital serum Cr trajectory profiles.¹⁵ This trajectory analysis identifies groups with similar trajectories based on maximum likelihood estimates, with each patient assigned a probability of membership in each trajectory. We used the time of serum Cr measurements as unique data points. To assess how many latent trajectories best fit the data, we used a two-stage model selection process¹⁶ by evaluating both 1) the Bayesian Information Criteria (BIC) for the empirical fit of each model, as well as 2) clinical judgment.¹⁷ To evaluate the adequacy of the final model, average posterior probability membership (AvgPP) and odds of correct classification were calculated for each trajectory group. The minimum AvgPP was greater than 0.7, with odds of correct classification (OCC) of 5.0 or more for all groups.¹⁶

We conducted bivariate and multivariate logistic regression models to evaluate the effect of demographic, clinical, and healthcare utilization factors as well as in-hospital Cr trajectories on 90-day and 1-year mortality, respectively. All analyses were conducted with SAS Version 9.4 (SAS, Cary, NC)¹⁸, and trajectory analyses were performed using PROC TRAJ macro.^{19, 20} All other visualizations were created using the ggplot2 package in R software 64x,3.6.1.²¹

To gain insight into the effect of pre-existing renal function on AKI outcomes, we also conducted GBTM to identify Cr trajectory profiles within one year before index hospitalization. For this GBTM, patients had to have at least two serum Cr tests within one year before hospitalization to be included in the analysis. We used the analytical approach described in the section above for these trajectories. We examined the joint effects of pre-hospitalization (pre-AKI) and in-hospital (AKI) renal function trajectories on 90-day and 1-year mortality.

Development of trajectory models and evaluation for the fit of both trajectory models are included in the Supplementary Materials and Table S1.

RESULTS

Patient Characteristics

A total of 121,197 patients with cirrhosis were hospitalized at VA hospitals between 1/1/2004 to 12/31/2014. After excluding patients whose length of stay was less than 24 hours (n=2643), who had liver or kidney transplant, had end-stage renal disease (ESRD) and dialysis (n=5427) prior to index hospitalization and those who did not have an available baseline and 2^nd serum Cr to diagnose AKI (n=39,214), a total of 73,913 patients with cirrhosis were identified. Among 73,913 patients, 15% (n=11,082) developed AKI. After further excluding patients with AKI who died during the admission (n=2396), whose length of stay ≥ 30 days (n=458), had dialysis/ESRD or kidney transplantation during the index hospitalization (n=371) and did not have at least 2 data points (n=940), a total of 6917 patients with AKI met the inclusion criteria and were discharged alive following AKI-related hospitalization. Figure S1 displays the steps included in the cohort identification.

Table 1 shows the characteristics of patients with cirrhosis and AKI. Further details related to patient characteristics are included in Supplementary Materials, Tables S2, and S3.

Table 1.

Baseline Characteristics 6917 Inpatients with Cirrhosis and AKI

Characteristics
Age (years), Mean (SD)	64.25 (10.40)
Gender, n (%)	N (%)
Male	6763 (97.77)
Female	153 (2.21)
Age groups (years), n (%)
Age < 50	424 (6.13)
Age 50-65	3757 (54.32)
Age > 65	2736 (39.55)
Race, n (%)
White	4774 (69.02)
African-American	1407 (20.34)
Others	180 (2.60)
Unknown/missing	556 (8.04)
CKD, n (%)
Not CKD	3896 (56.32)
CKD	2817 (40.73)
Unknown/missing	204 (2.95)
Etiology of cirrhosis, n (%)
Hepatitis C	854 (12.35)
Alcohol	1908 (27.58)
Hepatitis C and Alcohol	1033 (14.93)
Other etiologies	3122 (45.14)
Ascites, n (%)
No	2896 (41.87)
Yes	4021 (58.13)
Hepatic encephalopathy, n (%)
No	5313 (76.81)
Yes	1604 (23.19)
Hepatocellular carcinoma, n (%)
No	6151 (88.93)
Yes	766 (11.07)
Gastrointestinal hemorrhage, n (%)
No	6421 (92.83)
Yes	496 (7.17)
Baseline MELD-Na
<10	34 (0.49)
10 -<20	1001 (14.47)
≥ 20	1885 (27.25)
Unknown/Missing	3997 (57.79)
Discharge MELD-Na
<10	1455 (21.04)
10-<20	3227 (46.65)
≥ 20	1169 (16.90)
Unknown/Missing	1066 (15.41)
Clinical visit within 15 days after discharge, n (%)
None	4220 (61.01)
Specialty only	170 (2.46)
PCP only	2366 (34.21)
Both	161 (2.33)

AKI, acute kidney injury; CKD, chronic kidney disease; MELD-Na, model for end-stage liver disease-sodium; PCP, primary care physician.

In-Hospital Renal Trajectory Patterns

On average, patients had 6 (SD=1.5) separate Cr measurements during hospitalization. We chose a five-trajectory model for AKI patients (Figure 1). The in-hospital renal trajectory patterns included Group 1) mild AKI with full improvement (43.1%), Group 2) severe AKI with rapid improvement (7%), Group 3) moderate AKI with partial improvement (26.6%), Group 4) moderate to severe AKI with partial improvement (12.1%), and Group 5) severe AKI with minimal improvement (11.2%).

Figure 1.

In-hospital renal trajectory patterns in patients with cirrhosis and AKI during the index hospitalization.

Associations between Renal Trajectory Patterns and Post-Discharge Mortality

Renal trajectory patterns predicted 90-day and 1-year mortality in patients with AKI. Ninety-day mortality ranged from 24.7% in patients with severe AKI with rapid improvement to 48.0% in those with severe AKI with minimal improvement. We found a similar trend in 1-year mortality. Compared to patients with mild AKI with full improvement, there was a stepwise yet modest increase in the risk of 90-day mortality among patients who had moderate AKI with partial improvement (adjusted odds ratio [aOR]=1.46, 95% confidence interval [CI]=1.27-1.69) (Table 2). Mortality risk was substantially higher in patients who had moderate to severe AKI with partial improvement (aOR=2.31, 95% CI=1.92-2.78) and severe AKI with minimal improvement (aOR=3.08, 95% CI=2.54-3.72). Older age, ascites, hepatocellular carcinoma, model for end-stage liver disease-sodium (MELD-Na) score between 10 and <20 and MELD-Na ≥ 20 had higher post-discharge 90-day mortality risk following AKI related hospitalization with aOR of 1.05 (95% CI: 1.04-1.05), 1.52 (95% CI: 1.35-1.71), 2.33 (95% CI: 1.95-2.78), 2.16 (95% CI: 1.84-2.54) and 8.19 (95% CI: 6.72-9.98), respectively (Table 2). Compared to patients who did not have a clinic visit within 15 days after discharge, the mortality risk was lower for patients who were seen by their primary care physician only (aOR=0.66, 95% CI=0.58-0.74), for patients who were seen by specialists only (aOR=0.55, 95% CI=0.37-0.80), or for those who were seen by both primary care physicians and specialists within 15 days after discharge (aOR=0.40, CI=0.26-0.60) (Table 2). These associations persisted in the analysis that examined the risk of 1-year mortality.

Table 2.

Logistic Regression Analysis Predicting Post-Discharge Mortality in Patients with Cirrhosis and AKI Using Inpatient Renal Trajectories and Characteristics

	90-Day OR (95%CI)		1-Year OR (95%CI)
	Unadjusted	Adjusted	Unadjusted	Adjusted
A. Renal (serum Cr) trajectories (ref=1.Mild AKI with full improvement)
2. Severe AKI with rapid improvement	1.01 (0.80-1.26)	1.11 (0.86-1.42)	1.16 (0.95-1.41)	1.22 (0.98-1.52)
3.Moderate AKI with partial improvement	1.56 (1.37-1.77)	1.46 (1.27-1.69)	1.85 (1.64-2.08)	1.66 (1.45-1.90)
4. Moderate to severe AKI with partial improvement	2.30 (1.96-2.70)	2.31 (1.92-2.78)	2.41 (2.06-2.82)	2.16 (1.81-2.59)
5. Severe AKI with minimal improvement	2.87 (2.44-3.38)	3.08 (2.54-3.72)	2.83 (2.41-3.33)	2.71 (2.25-3.27)
Male	1.45 (1.00-2.09)	1.17 (0.78-1.76)	1.65 (1.18-2.30)	1.31 (0.91-1.89)
Age (year)	1.03 (1.03-1.04)	1.05 (1.04-1.05)	1.04 (1.04-1.05)	1.05 (1.04-1.06)
Race (reference=white)
African-American	0.94 (0.83-1.07)	0.98 (0.84-1.13)	0.94 (0.84-1.06)	0.99 (0.86-1.13)
Others	0.77 (0.55-1.08)	0.89 (0.61-1.29)	0.67 (0.49-0.91)	0.77 (0.55-1.08)
Unknown	1.73 (1.45-2.07)	1.50 (1.19-1.88)	1.52 (1.27-1.82)	1.47 (1.17-1.84)
Etiology (reference=hepatitis C)
Alcohol	1.26 (1.05-1.50)	1.22 (0.99-1.49)	1.05 (0.89-1.24)	1.00 (0.84-1.21)
Hepatitis C and Alcohol	0.86 (0.70-1.06)	0.94 (0.75-1.18)	0.79 (0.66-0.95)	0.88 (0.72-1.08)
None	1.55 (1.31-1.83)	1.19 (0.98-1.45)	1.44 (1.24-1.68)	1.05 (0.88-1.25)
Ascites (reference=no)	1.65 (1.48-1.83)	1.52 (1.35-1.71)	1.89 (1.71-2.08)	1.77 (1.58-1.97)
Hepatic encephalopathy (reference=no)	0.92 (0.81-1.03)	0.80 (0.70-0.92)	1.05 (0.94-1.17)	1.00 (0.88-1.14)
Hepatocellular carcinoma (reference=no)	1.84 (1.58-2.14)	2.33 (1.95-2.78)	1.86 (1.59-2.17)	2.34 (1.96-2.78)
Gastrointestinal hemorrhage (reference=no)	0.86 (0.71-1.05)	0.90 (0.72-1.13)	0.83 (0.69-1.00)	0.88 (0.72-1.08)
CKD (reference=none)
CKD	1.07 (0.97-1.19)	0.74 (0.65-0.85)	1.30 (1.18-1.44)	0.89 (0.79-1.01)
Unknown	1.87 (1.41-2.49)	1.29 (0.90-1.86)	1.67 (1.25-2.22)	1.13 (0.79-1.62)
MELD-Na discharge (reference=<10)
10- <20	2.02 (1.74-2.35)	2.16 (1.84-2.54)	1.91 (1.68-2.16)	2.01 (1.75-2.31)
>=20	6.16 (5.17-7.35)	8.19 (6.72-9.98)	5.14 (4.35-6.08)	6.64 (5.50-8.02)
Unknown	1.33 (1.10-1.62)	1.36 (1.11-1.67)	1.40 (1.19-1.65)	1.49 (1.25-1.78)
Clinic visit within 15 days after discharge (reference= none)
Specialty only	0.60 (0.42-0.85)	0.55 (0.37-0.80)	0.93 (0.69-1.27)	0.87 (0.62-1.22)
PCP only	0.63 (0.56-0.70)	0.66 (0.58-0.74)	0.75 (0.68-0.83)	0.79 (0.70-0.88)
Both	0.46 (0.31-0.67)	0.40 (0.26-0.60)	0.65 (0.47-0.89)	0.56 (0.40-0.80)

abbreviations: AKI, acute kidney injury; CKD, chronic kidney disease; Cr, serum creatinine; MELD-Na, model for end-stage liver disease-sodium; OR, odds ratio; PCP, primary care physician.

Secondary and subgroup analyses:

The key findings did not change when we limited our analysis to patients who had Cr test within three months before hospitalization (Table S4). The renal trajectories showed similar patterns between patients without ascites and MELD-Na < 20 and between those with ascites and MELD-Na ≥ 20 (Table 3, Figure 2).

Table 3.

Logistic Regression Analysis Predicting Post-Discharge 90-Day Mortality in Patients with Cirrhosis and AKI Using Inpatient Renal Trajectories and Characteristics Stratified Based on Ascites and MELD-Na Score

	Ascites		Discharge MELD-Na Score
	Without ascites	With ascites	MELD-Na <20	MELD-Na ≥ 20
Renal trajectories (reference= 1. Mild AKI with full improvement)
2. Severe AKI with rapid improvement	1.15 (0.77-1.72)	1.73 (1.34-2.24)	1.10 (0.81-1.48)	1.35 (0.83-2.18)
3.Moderate AKI with partial improvement	1.64 (1.28-2.10)	1.25 (1.04-1.49)	1.44 (1.20-1.72)	1.13 (0.83-1.55)
4. Moderate to severe AKI with partial improvement or worsening*	2.69 (1.93-3.75)	2.22 (1.74-2.82)	2.50 (2.00-3.13)	3.23 (2.02-5.16)
5. Severe AKI with minimal improvement	3.79 (2.71-5.30)	2.80 (2.20-3.56)	2.67 (2.10-3.39)	3.64 (2.34-5.68)
Male	1.53 (0.75-3.14)	1.03 (0.62-1.70)	1.21 (0.73-2.00)	1.04 (0.43-2.52)
Age (year)	1.04 (1.03-1.05)	1.05 (1.04-1.05)	1.05 (1.04-1.06)	1.03 (1.01-1.04)
Race (reference=white)
African-American	0.95 (0.75-1.22)	0.98 (0.82-1.18)	0.99 (0.83-1.18)	0.93 (0.67-1.30)
Others	1.25 (0.67-2.31)	0.77 (0.48-1.22)	0.73 (0.46-1.17)	1.29 (0.53-3.12)
Unknown	1.18 (0.79-1.74)	1.75 (1.31-2.35)	1.30 (0.96-1.75)	2.12 (1.20-3.74)
Etiology (reference=hepatitis C)
Alcohol	1.15 (0.83-1.60)	1.24 (0.95-1.62)	1.38 (1.06-1.79)	0.98 (0.64-1.49)
Hepatitis C and Alcohol	0.87 (0.60-1.25)	1.02 (0.76-1.38)	0.94 (0.70-1.26)	0.99 (0.62-1.57)
None	1.07 (0.78-1.46)	1.23 (0.95-1.60)	1.18 (0.92-1.52)	1.09 (0.71-1.68)
Ascites (reference=no)			1.86 (1.60-2.15)	0.74 (0.56-0.97)
Hepatic encephalopathy (reference=no)	1.18 (0.94-1.50)	0.65 (0.55-0.78)	0.80 (0.67-0.97)	0.89 (0.69-1.16)
Hepatocellular carcinoma (reference=no)	3.05 (2.37-3.92)	1.94 (1.49-2.51)	2.37 (1.91-2.93)	2.65 (1.71-4.12)
Gastrointestinal hemorrhage (reference=no)	0.97 (0.66-1.43)	0.95 (0.72-1.26)	0.97 (0.73-1.29)	0.92 (0.61-1.38)
CKD (reference=none)
CKD	0.83 (0.65-1.04)	0.68 (0.58-0.81)	0.74 (0.62-0.87)	0.77 (0.55-1.08)
Unknown	1.62 (0.89-2.94)	1.12 (0.70-1.79)	1.24 (0.74-2.07)	0.81 (0.31-2.09)
Clinic visit within 15 days after discharge (reference= none)
Specialty only	0.45 (0.21-1.00)	0.61 (0.39-0.95)	0.54 (0.33-0.89)	0.51 (0.26-1.00)
PCP only	0.56 (0.45-0.69)	0.71 (0.61-0.83)	0.69 (0.59-0.80)	0.57 (0.43-0.76)
Both	0.30 (0.13-0.68)	0.47 (0.29-0.77)	0.33 (0.19-0.58)	0.55 (0.28-1.10)
MELD-Na score discharge (reference= <10)
10- <20	2.66 (2.02-3.49)	1.86 (1.52-2.28)
>=20	15.12 (10.77-21.22)	5.68 (4.44-7.28)
Unknown	1.56 (1.11-2.18)	1.25 (0.96-1.62)

AKI, acute kidney injury; CKD, chronic kidney disease; MELD-Na, model for end-stage liver disease-sodium; PCP, primary care physician.

^*Partial improvement (without ascites and MELD-Na < 20) or worsening (with ascites and MELD-Na ≥ 20)

Figure 2.

In-hospital renal trajectory patterns in patients with cirrhosis and AKI stratified based on ascites and MELD-Na score: A) without ascites. B) with ascites. C) MELD-Na<20. D) MELD-Na≥20. abbreviations: Cr, serum creatinine; MELD-Na, model for end-stage liver disease-sodium.

We examined pre-hospitalization renal trajectory patterns to gain insight into the observed protective effect of pre-existing renal function, including CKD, on AKI outcomes. After following the criteria described in Methods, we chose a four-trajectory model for pre-hospitalization trajectory patterns (Figure S2). The four distinct groups were patients with 1) normal/near-normal renal function (52.6%), 2) mild (33.0%), 3) moderate (12.1%), and 4) severe (2.3%) pre-existing renal dysfunction.

Of the 6917 patients with AKI, 2461 (43.1%) patients had a mild kidney injury (initial Cr values were approximately in the range of 1.2-1.4 mg/dl followed by a quick recovery in their renal function after AKI (Group 1) (Figure 1). Most of these patients (88.2%) had a normal/near-normal renal function with Cr values approximately in the range of 0.8 to 1.2 mg/dl within one year before hospitalization (Figures S2). The 90-day mortality in Group 1 was 24.8% (Figure 3).

Figure 3.

Combined effects of pre-hospitalization within one year before index hospitalization) and in-hospital renal trajectory patterns on the risk of 90-day post-discharge mortality in patients with cirrhosis and AKI. abbreviation: G; group

A total of 388 (7.0%) patients had severe AKI with rapid improvement (Group 2). Initial Cr values were between 3.0 to 3.5 mg/dl with improvement in renal function during the hospital course; the Cr values plateaued approximately in the range of 1.2 and 1.5 mg/dl for this group (Figure 1). More than one-fourth of these patients (26.3%) had a normal/near-normal renal function before AKI; 90-day mortality was 31.4% in these patients. Most of the remaining patients had a mild chronic renal disease (pre-hospitalization Cr values approximately in the range of 1.3-2.2 mg/dl) and moderate chronic renal disease (pre-hospitalization Cr values approximately in the range of 1.8 to 2.8 mg/dl) (Figure S2); for these subgroups, the AKI hospital trajectory represented an acute yet moderate worsening in their chronic trajectory which subsequently returned to baseline renal function with 90-day mortality ranging from 18.8% to 24.5% (Figure 3).

A total of 1523 (26.6%) patients had moderate kidney injury followed by rapid – yet incomplete – improvement with Cr values plateauing around 1.5-1.6 mg/dl range (Group 3 in Figure 1). More than 50% of these patients had pre-existing mild to moderate chronic kidney disease (Figure 3), with the acute episode representing severe acute-on-chronic injury with a rapid return to baseline function. The mortality in these subgroups ranged from 20.3% to 28.4% and was lower than the mortality in patients who had normal/near-normal renal function before developing AKI (41.0%) (Figure 3).

A total of 721 (12.1%) patients had a trajectory that represented moderate renal dysfunction with Cr values in approximately 2.1-2.8 mg/dl range throughout the hospital course (Group 4 in Figure 1). Similar to Group 3, this group was mixed in regards to pre-hospitalization renal function. Specifically, approximately 18.6% of these patients had normal/near-normal renal function before developing AKI; this group had the highest post-discharge mortality, with a 90-day mortality of 55.2% patients. The pre-hospitalization trajectory for most of the remaining patients represented mild to moderate chronic kidney disease and was not much different from the trajectory during their hospital stay – the 90-day mortality for this group ranged from 28.8% to 46.6% (Figure 3).

In all, 677 (11.2%) patients had severe renal dysfunction with minimal improvement during hospitalization (Group 5 in Figure 1). More than 40% of these patients had moderate chronic renal dysfunction before index hospitalization; in this group, the 90-day mortality was 45.3% and lower than the mortality in the 13.6% and 27.5% of patients who had normal/near-normal renal function or only mild renal dysfunction, respectively prior to their index hospitalization, with 90-day mortality approaching and exceeding 60% (Figure 3).

DISCUSSION

Our results show that in-hospital trajectory patterns (Figure 1) had an effect on 90-day and 1-year mortality (Table 2). We also found that pre-existing renal function (Figure S2) was an important factor because, within each trajectory-based group, the outcomes differed based on pre-existing renal patterns (Figure 3). A subgroup of patients whose in-hospital trajectories returned to baseline renal pattern had better outcomes. However, others, who presented with partial or minimal improvement in their in-hospital renal trajectory had the worst outcome with the highest short-term mortality. Specifically, patients with a normal/near-normal baseline renal function who presented with severe AKI with minimal improvement had the highest the 90-day mortality. In general, patients with prior chronic renal dysfunction had better outcomes than those with normal/near-normal renal function before developing AKI.

Our findings are also important for the clinical management of AKI patients at the time of their discharge from the hospital. These results suggest that a tailored approach, matching surveillance, and follow-up intensity to individual risk, could improve patient outcomes. For example, aligning interventions for patients in the high-risk renal trajectory subgroups might be one way to reduce the otherwise high risk of death in AKI. Our data also show that outpatient follow-up within 15 days of discharge resulted in lower overall mortality, with the lowest mortality in patients managed by both primary and specialty care providers. We also found that less than 40% of hospitalized patients with cirrhosis who survived AKI were seen within 15 days in outpatient clinics following discharge, findings similar to those observed in our previous work.²² Based on these data, an outpatient follow-up with primary care physicians and specialists within 15 days after discharge might be warranted for all patients with AKI, particularly for those at the highest risk of mortality based on trajectory patterns. An evaluation for simultaneous kidney transplantation should be considered for patients with cirrhosis and AKI whose inpatient renal function worsens or improves minimally. Lastly, given the high risk of mortality within 90 days of discharge, the renal trajectory patterns can inform decisions about the futility of care, especially among patients who are not eligible for liver or kidney transplantation.

Our study lacked information on the cause of AKI. Patients with cirrhosis and AKI whose serum Cr trajectories showed marked improvement might have had AKI secondary to pre-renal azotemia in the presence of gastrointestinal bleeding or sepsis. In contrast, patients whose renal function improved minimally might have had a hepatorenal syndrome that progressed to acute tubular necrosis. Additionally, some patients with AKI might have been treated with vasoconstrictors; we were unable to ascertain the type and length of treatment. Notwithstanding the underlying mechanisms of AKI, our data provide support to the importance of renal trajectory patterns in predicting mortality post-hospitalization. These patterns can be easily recognized in electronic medical records, and thus broadly used to risk-stratify patients for close follow-up, transplant evaluation, or palliative care referral. As our current analysis focused on post-discharge mortality as the only endpoint, we did not evaluate the improvement or worsening of renal function after discharge. Our renal trajectory pattern analysis was based on serum Cr values. Given the fact that the increase in serum Cr is expected to lag for 2 to 3 days before it reaches out a plateau after an acute decline in actual GFR²³, renal trajectories may have different sensitivities for a decline in actual GFR. Actual GFR in AKI setting can only be measured by real-time GFR measurement technologies²⁴, which are not widely available in clinical practice.

Several reasons explain why our study population represented a small subset of the larger cohort of patients with cirrhosis. First, the sampling frame for our overall cohort included all patients hospitalized with cirrhosis defined by more than one ICD-9 code for cirrhosis or its complications between 01/01/2004 and 12/31/2014. Of note, for patients with multiple cirrhosis-related hospitalizations in the study time frame, we sampled patients’ first hospitalization for this study. Prior studies on AKI prevalence used a cross-sectional approach that sampled any hospitalization or follow-up that occurred during a specified timeframe. For instance, in one study, the proportion of AKI episodes in all follow-ups was 13%²⁵, this was even lower than the proportion of subjects in our cohort who had AKI at first index hospitalization. Further, because we sampled patients’ first hospitalization for this study, our study cohort included cirrhosis patients who were likely at an earlier (less severe) phase of the cirrhosis severity spectrum. This may explain why only a small subset of our patients met the criteria for AKI.

Our results might lack generalizability because a higher percentage of the study population was male and most likely had higher serum Cr levels compared with female patients with cirrhosis.^{26, 27} Therefore, future studies are needed to assess renal function trajectories and associated mortality risk in women with cirrhosis.

In summary, we found that in patients with cirrhosis discharged following AKI related hospitalization, renal function trajectories were significantly associated with 90-day and 1-year post-discharge mortality risk. Renal function trajectories-predicted post-discharge mortality and may serve as the basis of future risk stratification tools in AKI. Patients with moderate to severe AKI with partial or minimal improvement in renal function following AKI, especially those with normal renal function prior to AKI, represented the highest risk group. Close monitoring and early referral to hepatologists and nephrologists may reduce the otherwise high risk of mortality in this subgroup. For patients who are ineligible for transplantation, the renal trajectory patterns can guide timely referral to palliative care to improve the quality of end-of-life care for patients with cirrhosis.

Supplementary Material

Supplementary Figure 1

Supplementary Figure 1. Steps included in the identification of the final cohort with cirrhosis and acute kidney injury (AKI). Cr, serum creatinine; ESRD, end-stage renal disease; KT, kidney transplantation; LT, liver transplantation.

Supplementary Figure 2.

Supplementary Figure 2. Prehospitalization renal trajectory patterns in patients with cirrhosis and acute kidney injury before index hospitalization. Cr, serum creatinine.

Need to Know.

Background:

Outcomes of patients with cirrhosis and acute kidney injury (AKI) vary with stage of AKI, chronic kidney disease (CKD), and trajectory of renal function.

Findings:

In an analysis of 6917 patients with cirrhosis who were hospitalized for AKI in the Veterans’ Administration, we found the pattern of renal trajectory to associate with mortality after discharge.

Implications for patient care:

Renal trajectory patterns can be used to identify patients with cirrhosis and AKI who should receive intensive post-discharge management.