Attendance at a Transitional Liver Clinic May Be Associated with Reduced Readmissions for Patients with Liver Disease

Abstract

Introduction:

Patients with liver disease have high rates of early hospital readmission, but there are no studies of effective, scalable interventions to reduce this risk. In this study, we examined the impact of a Physician Assistant (PA)-led post-discharge Transitional Liver Clinic (TLC) on hospital readmissions.

Methods:

We performed a cohort study of all adults seen by a hepatologist during admission to a tertiary care center in 2019 (excluding transplant patients). We compared those who attended the TLC with those who did not with respect to 30-day readmission and mortality. Propensity score-adjusted modeling was used to control for confounding.

Results:

Of 498 patients, 98 were seen in the TLC. 35% had alcoholic liver disease, and 58% had cirrhosis. Attendees were similar to non-attendees with respect to demographics, liver disease characteristics and severity, comorbidities, and discharge disposition. 30-day cumulative incidence of readmissions was 12% in TLC attendees, compared to 22% in non-attendees (p=0.02), while 30-day mortality was similar (2.0% vs. 4.3%; p=0.29). In a model using propensity score adjustment, TLC attendance remained associated with reduced readmissions (SHR 0.52; 95% CI, 0.27–0.997; p=0.049). The effect of TLC was greater in women compared to men (p=0.07) and in those without chronic kidney disease (p=0.02), but there were no differences across other subgroups.

Conclusions:

Patients with liver disease seen in a PA-led TLC may have a significant reduction in the 30-day readmission rate. Randomized trials are needed to establish the efficacy of PA-led post-discharge transitional care for this population.

INTRODUCTION

Patients with liver disease suffer from high rates of hospitalization and readmission. Cirrhosis is associated with 30-day readmissions exceeding 25%, mostly caused by ascites and hepatic encephalopathy.^1,2 These complications of advanced liver disease are also common in alcoholic hepatitis, where 30-day readmission rates are 30%.^3–5 Readmissions in these populations are preventable in 22–37%, suggesting opportunities to improve outcomes.^6,7

Despite the growing recognition of the importance of liver disease readmissions, few studies have explored interventions to reduce this burden. Of the studied interventions, several have not shown efficacy due to small samples of severely ill patients,⁸ limitations in study designs, and reliance on isolated interventions.⁹ Two programs reduced readmissions using same-day radiology, endoscopy, and a “day-hospital” for outpatient treatment.^10,11 These resource-intensive programs may not be scalable to other settings. All of the above studies were performed exclusively in patients with cirrhosis, and only one was randomized.⁸

In response to high readmission rates, we established the Transitional Liver Clinic (TLC) in 2019. This model provides a dedicated service for patients to see a hepatology Physician Assistant (PA), where they can receive scalable, sustainable, coordinated, patient-centered care. In this study, we examined all patients with liver disease hospitalized in 2019 to assess the impact of the TLC on readmissions. We hypothesized that TLC attendance would be associated with reduced readmissions.

METHODS

The Transitional Liver Clinic

In April 2019 we initiated the TLC, which allows patients to see a hepatology PA for timely post-discharge transitional care. In this model, for eligible patients (non-hospice, non-transplant), the inpatient hepatology team sends an electronic health record (EHR) message to schedule the TLC appointment. Inpatient providers are encouraged to refer all patients, but the provider makes the final referral decision. The PA then reviews the EHR for complexity, and a face-to-face visit is scheduled within 14 days (for moderate complexity) or 7 days (for high complexity). Complexity is estimated based on CMS medical decision-making standards.¹² Due to high no-show rates, we refined the process in July 2019 to include a nurse telephone call within 2 business days of discharge. During this call, the patient is queried as to overall health, problems following discharge, medication reconciliation, confirmation of upcoming testing/procedures, and confirmation of the TLC visit. The nurse also provides education and answers questions. If the patient does not answer, a HIPAA-compliant voicemail is left when possible. The PA and a hepatologist are also available.

At the TLC visit, the patient undergoes a history and physical exam, with attention to liver-related complications and contributing factors. In addition to direct care (ordering/coordinating tests/procedures, medication adjustment), referrals can be made for additional services. These can include (1) social work for financial assistance and linkage to alcohol/substance use disorder treatment and community resources, (2) nutrition for dietary counseling, and (3) pharmacy for medication review and prescription assistance. The PA also provides education, ensures follow-up with other providers, and assists in scheduling follow-up with community providers and services.

Study Population

The study cohort consisted of all patients potentially eligible for the TLC, including adults (age ≥18) admitted to the Indiana University Health Academic Health Center between January and December 2019 who were seen in the hospital by hepatology for any liver-related issue. We included patients who were admitted to either the primary hepatology service or to another service with hepatology consultation. Patients with advanced liver disease are generally admitted to the primary service with overflow to the hospitalist service with hepatology consultation when the primary service census is full. The discharge planning process is the same regardless of service. We excluded individuals who died during the admission and those discharged with hospice. We also excluded patients undergoing liver transplant (LT) evaluation and post-LT patients because these patients receive outpatient care by a separate LT team. Patients were followed from the day of discharge for 30 days to assess for readmission and up to 90 days to assess for mortality. The Indiana University Institutional Review Board approved this study.

Outcomes

The primary outcome was all-cause 30-day readmission. Readmissions were captured from the EHR and the Indiana Network for Patient Care (INPC), a health information exchange connecting over 110 hospitals, 18,000 practices, and 50,000 providers across Indiana.¹³ The secondary outcome was mortality, also abstracted from the EHR and INPC. For patients lost to follow-up without mortality data, an online obituary search was performed.¹⁴ We did not collect emergency visits that did not result in readmission.

Variables

The primary exposure was TLC attendance. We recorded whether a nurse phone call was made and whether the patient was successfully contacted.

We collected additional variables that could influence TLC attendance and readmission. Demographic information included age, sex, race/ethnicity, and health insurance (private, Medicare, Medicaid, none). The presence of a support person was determined based on social work documentation. Distance to the medical center was calculated as the distance between centroids of the patient’s home and the medical center zip codes. The primary reason for admission was categorized as acute kidney injury (AKI)/electrolyte disturbance, altered mental status/hepatic encephalopathy, acute viral hepatitis, alcoholic hepatitis/alcohol related, ascites/anasarca/hydrothorax, elevated liver enzymes (without any of the aforementioned diagnoses), gastrointestinal/variceal bleeding, infection/sepsis, and other. We recorded whether patients were admitted to the primary hepatology service or were seen by the consult service. Underlying liver disease was categorized as alcohol, viral hepatitis, nonalcoholic fatty liver disease, cholestatic/autoimmune, and other. The presence of cirrhosis (based on characteristic clinical, laboratory, radiologic, and endoscopic features) was noted, as were ascites and hepatic encephalopathy. Laboratory results nearest to discharge included serum albumin, sodium, creatinine, bilirubin, and international normalized ratio to calculate the Model for End-Stage Liver Disease-Sodium (MELD-Na).¹⁵ We collected comorbidities: AKI during hospitalization, chronic kidney disease (CKD), congestive heart failure, chronic obstructive pulmonary disease, mental health disorder (including depression, anxiety, and others), and substance use disorder (including alcohol and illicit substances). Discharge location was noted (home vs. facility). We did not include non-TLC follow-up due to a lack of reliable data outside of our healthcare system.

Statistical Analysis

The distributions of variables by TLC status, clinical service, and implementation period were assessed using two-sample tests for continuous and categorical variables. Exact and non-parametric methods were utilized where sample size or the normality assumption appeared tenuous. Unadjusted tests of association between TLC attendance and 30-day readmission and mortality were conducted using Gray’s test of cumulative incidence curves and exact Chi-square tests, respectively. Adjusted associations between TLC attendance and outcomes were examined using a propensity score-adjusted competing risk model.¹⁶ The propensity score included all clinically-relevant covariates that could influence TLC attendance irrespective of statistical significance. To assess the adequacy of the propensity score, we fit a logistic regression model using the propensity score variables to predict TLC attendance and assessed its predictive ability by the area under the receiver operating characteristic curve (AUROC).

For hypothesis-generating purposes, clinically-relevant subgroups were examined with the main effect of TLC across subgroups, along with their interactions. To assess whether the association between TLC and readmission was maintained for those with timely TLC attendance, sensitivity analyses were also performed, excluding patients who attended the TLC >14 days after discharge. We also performed landmark analyses, where exposure to the TLC was ascertained at 7 days and patients with outcomes prior to that time were excluded. All analyses were performed in SAS 9.4 and R 3.6.2 with significance level 0.05.

RESULTS

498 patients were seen by hepatology and discharged without hospice. Of these, 146 were discharged before the TLC was established in April 2019 (Figure 1). Five were discharged in late March and attended the TLC in April. Between April and June, 14% attended the TLC. Attendance increased to 34% after July, when nurse phone calls were instituted, and among those contacted, attendance was 63%. Three were admitted in December 2019, but not discharged until January 2020. Of non-attendees, 4% refused referral, and 13% were “no-shows,” although not all reasons for non-attendance could be ascertained. The median time between discharge and TLC visit was 10 days (interquartile range 8–14).

Figure 1.

Implementation phases of the TLC during the 2019 calendar year and the corresponding attendance rates. *Five patients were discharged in late March and attended the TLC in April. Three patients were admitted in December and discharged in January 2020.

Patient Characteristics

The mean age was 52, 46% were female, 91% were white, and 92% had a support person (Table 1). Reasons for admission varied widely; the most common underlying liver diseases were alcoholic liver disease (35%) and viral hepatitis (23%). 58% had cirrhosis, the mean MELD-Na was 16, and 28% had AKI. The mean length of stay was 8 days, and 82% were discharged home. Baseline characteristics were largely similar in those who did and did not attend TLC. Patients on the primary hepatology service were less likely to attend. CKD was more common in the non-TLC group, and mental health disorders were present in 14% of attendees compared to 23% of non-attendees (p=0.06). 88% of attendees were discharged to home, compared to 81% of non-attendees (p=0.07). Comparisons according to attendance within different periods are shown in Supplementary Table 1. TLC attendance was less common on the primary service and in those with CKD and mental health disorders across time periods.

Table 1.

Patient Characteristics

Characteristic	Overall (N=498)	TLC (N=98)	No TLC (N=400)	p-value
Age, yrs, mean (SD)	52.0 (14.7)	52.1 (15.0)	52.0 (14.6)	0.92
Female sex, n (%)	228 (45.8)	43 (43.9)	185 (46.3)	0.67
White race, n (%)	451 (90.6)	89 (90.8)	362 (90.5)	0.92
Insurance, n (%)				0.17
Private	131 (26.5)	34 (35.1)	97 (24.4)
Medicare	171 (34.5)	30 (30.9)	141 (35.4)
Medicaid	171 (34.5)	28 (28.9)	143 (35.9)
None	22 (4.4)	5 (5.2)	17 (4.3)
Support person, n (%)	453 (92.1)	93 (95.9)	360 (91.1)	0.27
Distance to medical center, mi, mean (SD)	58.4 (54.3)	57.7 (90.9)	58.6 (40.8)	0.88
Reason for admission, n (%)				0.32
AKI/electrolyte disturbance	18 (3.6)	6 (6.1)	12 (3.0)
Altered mental status/HE	58 (11.7)	13 (13.3)	45 (11.3)
Acute viral hepatitis	70 (14.1)	9 (9.2)	61 (15.3)
Alcoholic hepatitis/alcohol related	51 (10.3)	13 (13.3)	38 (9.5)
Ascites/anasarca/hydrothorax	57 (11.5)	12 (12.2)	45 (11.3)
Elevated liver enzymes	69 (13.9)	16 (16.3)	53 (13.3)
Gastrointestinal/variceal bleeding	55 (11.1)	8 (8.2)	47 (11.8)
Infection/sepsis	51 (10.3)	12 (12.2)	39 (9.8)
Other	68 (13.7)	9 (9.2)	59 (14.8)
Hepatology primary service, n (%)	167 (34.1)	21 (21.4)	146 (37.2)	0.003
Liver disease, n (%)				0.83
Alcohol	173 (34.7)	33 (33.7)	140 (35.0)
Viral hepatitis	114 (22.9)	21 (21.4)	93 (23.3)
NAFLD	86 (17.3)	19 (19.4)	67 (16.8)
Cholestatic/Autoimmune	35 (7.0)	9 (9.2)	26 (6.5)
Other	90 (18.1)	16 (16.3)	74 (18.5)
Cirrhosis, n (%)	286 (57.5)	55 (56.1)	231 (57.9)	0.75
Ascites, n (%)	231 (46.4)	44 (44.9)	187 (46.8)	0.74
Hepatic encephalopathy, n (%)	213 (42.9)	40 (40.8)	173 (43.4)	0.65
Albumin, g/dL, mean (SD)	2.9 (0.6)	3.0 (0.5)	2.9 (0.6)	0.04
MELD-Na, mean (SD)	16.3 (6.4)	16.9 (6.4)	16.2 (6.4)	0.35
Acute kidney injury, n (%)	138 (27.8)	25 (25.5)	113 (28.3)	0.58
Chronic kidney disease, n (%)	62 (12.4)	6 (6.1)	56 (14.0)	0.03
Congestive heart failure, n (%)	37 (7.4)	9 (9.3)	28 (7.0)	0.44
COPD, n (%)	46 (9.3)	7 (7.2)	39 (9.8)	0.44
Mental health disorder, n (%)	104 (21.1)	14 (14.3)	90 (22.8)	0.06
Substance use disorder, n (%)	149 (30.4)	30 (30.6)	119 (30.4)	0.96
Length of stay, days, mean (SD)	7.9 (10.8)	6.2 (5.2)	8.3 (11.7)	0.08
Discharge to home, n (%)	406 (82.0)	84 (88.4)	322 (80.5)	0.07

Abbreviations: AKI, acute kidney injury; COPD, chronic obstructive pulmonary disease; HE, hepatic encephalopathy; MELD-Na, Model for End-Stage Liver Disease; NAFLD, nonalcoholic fatty liver disease; SD, standard deviation; TLC, Transitional Liver Clinic.

Outcomes

30-day cumulative incidence of readmissions for TLC attendees was 12%, compared to 22% for non-attendees (SHR 0.50; 95% CI, 0.28–0.91; Table 2). This reduction in readmissions was seen across the 30-day follow-up (Figure 2). In a sensitivity analysis excluding patients whose TLC visit was >14 days after discharge, the association between TLC and readmission was attenuated and not significant (SHR 0.61; 95% CI, 0.33–1.13; p=0.12). Patients who were successfully contacted by phone and attended the TLC had numerically lower 7- and 30-day cumulative incidences of readmissions (1.8% vs. 6.6% and 10.9% vs. 21.2%, respectively), but this difference was not statistically significant (SHR 0.48; 95% CI, 0.21–1.08; p=0.07). In a landmark analysis discounting patients with outcomes before day 7, 30% of readmissions were excluded, and the association between TLC and readmission was not significant (SHR 1.75; 95% CI, 0.63–4.90). 30-day mortality was similar between the groups (2.0% TLC vs. 4.3% non-TLC; p=0.29; Table 2). Readmission and mortality comparisons were similar when stratifying according to implementation period (Supplementary Table 1).

Table 2.

Outcomes According to Patient Characteristics

Characteristic	30-day readmission SHR (95% CI)	p-value	30-day mortality HR (95% CI)	p-value
TLC	0.50 (0.28, 0.91)	0.02	0.46 (0.11, 1.07)	0.29
Age, yrs	0.99 (0.98, 1.00)	0.22	1.07 (1.03, 1.11)	<0.001
Female sex	0.82 (0.55, 1.22)	0.33	0.54 (0.20, 1.41)	0.21
White race	0.75 (0.42, 1.36)	0.34	0.89 (0.21, 3.85)	0.87
Private insurance	1.10 (0.68, 1.81)	0.93	NE
Medicare	0.96 (0.59, 1.55)
Medicaid	Ref
No insurance	0.88 (0.32, 2.40)
Support person	1.03 (0.46, 2.28)	0.95	1.27 (0.17, 9.57)	0.82
Distance to med center, mi	1.00 (0.99, 1.00)	0.95	1.00 (1.00, 1.01)	0.50
Reason for admission		0.39	NE
AKI/electrolyte disturbance	1.21 (0.51, 2.85)
Altered mental status/HE	0.89 (0.45, 1.77)
Acute viral hepatitis	0.51 (0.23, 1.12)
Alcoholic hepatitis/alcohol related	0.72 (0.34, 1.54)
Ascites/anasarca/hydrothorax	0.82 (0.41, 1.66)
Elevated liver enzymes	0.40 (0.18, 0.90)
Gastrointestinal/variceal bleeding	0.66 (0.30. 1.43)
Infection/sepsis	0.74 (0.34, 1.62)
Other	Ref
Hepatology primary service	1.28 (0.86, 1.93)	0.23	0.51 (0.17, 1.54)	0.23
Liver disease		0.55		0.68
Alcohol	1.42 (0.80, 2.50)		0.91 (0.27, 3.11)
Viral hepatitis	1.06 (0.56, 2.03)		0.38 (0.07, 2.09)
NAFLD	0.91 (0.45, 1.84)		1.31 (0.35, 4.87)
Cholestatic/Autoimmune	1.14 (0.47, 2.72)		0.63 (0.07, 5.63)
Other	Ref		Ref
Cirrhosis	1.40 (0.93, 2.11)	0.11	6.44 (1.49, 27.87)	0.01
Ascites	2.08 (1.39, 3.12)	<0.001	2.55 (0.97, 6.72)	0.06
Hepatic encephalopathy	1.07 (0.72, 1.58)	0.75	1.35 (0.53, 3.39)	0.53
Albumin, g/dL	0.61 (0.42, 0.90)	0.01	0.53 (0.23, 1.23)	0.14
MELD-Na	1.05 (1.02, 1.08)	<0.001	1.12 (1.05, 1.18)	<0.001
Acute kidney injury	1.41 (0.93, 2.14)	0.10	2.35 (0.96, 5.79)	0.06
Chronic kidney disease	1.58 (0.97, 2.57)	0.07	2.57 (0.93, 7.14)	0.07
Congestive heart failure	1.32 (0.67, 2.60)	0.42	2.48 (0.72, 8.49)	0.15
COPD	1.93 (1.11, 3.36)	0.02	2.72 (0.90, 8.18)	0.08
Mental health disorder	0.83 (0.49, 1.38)	0.47	0.70 (0.20, 2.40)	0.57
Substance use disorder	0.96 (0.62, 1.48)	0.84	0.61 (0.20, 1.82)	0.37
Length of stay, days	1.00 (0.99, 1.01)	0.89	1.00 (0.95, 1.05)	0.85
Discharge to home	0.77 (0.48, 1.25)	0.29	0.46 (0.18, 1.22)	0.12

Abbreviations: AKI, acute kidney injury; COPD, chronic obstructive pulmonary disease; HR, hazard ratio; HE, hepatic encephalopathy; MELD-Na, Model for End-Stage Liver Disease; NAFLD, nonalcoholic fatty liver disease; NE, not estimable; SD, standard deviation; SHR, subhazard ratio; TLC, Transitional Liver Clinic

Figure 2.

Cumulative incidence of readmission accounting for the competing risk of death, according to TLC attendance.

Reasons for readmission were evenly distributed (Supplementary Figure 1). TLC had numerically lower readmissions for AKI/electrolyte disturbance, altered mental status/hepatic encephalopathy, ascites/anasarca/hydrothorax, and “other,” though these differences were not significant (p=0.08; Figure 3).

Figure 3.

Reasons for 30-day readmission according to TLC attendance. In TLC attendees, readmissions were numerically lower for AKI/electrolyte disturbance, altered mental status/HE, ascites/anasarca/hydrothorax, and “other,” but the differences were not statistically significant (p=0.08).

Those who died were older, but age was not associated with readmission (Table 2). Readmissions were similar in those on the primary and consult services. Both readmissions and mortality were associated with ascites and MELD-Na, and both were also greater in those with comorbidities, though most differences were not significant. Length of stay and discharge to a facility (vs. home) were not associated with either outcome. Mortality associated with insurance and reason for admission were not estimable due to sample size.

In a propensity score-adjusted model, TLC remained associated with reduced readmissions (SHR 0.52; 95% CI, 0.27–0.997; p=0.049). The AUROC of the covariates in predicting attendance was 0.77, suggesting moderately good predictive power for the score. 90-day mortality was also similar between the groups (11.2% TLC vs. 18.5% non-TLC; p=0.10).

Subgroup Effects

In subgroup analysis, TLC was associated with a trend toward reduction in readmissions in women as compared to men (p=0.07; Table 3). TLC had a greater impact in those without CKD (p=0.02). Other subgroup differences were not significant. The association of TLC with readmission was similar in those with and without cirrhosis, and across disease severity. It also did not vary with comorbidities or discharge location. Subgroup analyses according to reasons for admission and underlying liver disease were not estimable due to sample size.

Table 3.

Association Between TLC and 30-Day Readmission Across Subgroups

Subgroup	SHR (95% CI)	p-value for interaction
Age < 50 years	0.32 (0.12, 0.89)	0.25
Age ≥ 50 years	0.68 (0.33, 1.41)
Female	0.18 (0.04, 0.75)	0.07
Male	0.78 (0.40, 1.50)
Race	NE
Private insurance	0.17 (0.04, 0.70)	0.26
Medicare	0.90 (0.35, 2.30)
Medicaid	1.06 (0.13, 8.60)
No insurance	0.17 (0.04, 0.70)
Support person	NE
Distance to med center < 50 mi	0.55 (0.25, 1.22)	0.73
Distance to med center ≥ 50 mi	0.45 (0.18, 1.10)
Reason for admission	NE
Hepatology primary service	0.35 (0.09, 1.41)	0.48
Consult service	0.62 (0.32, 1.20)
Liver disease	NE
Cirrhosis	0.32 (0.10, 1.04)	0.35
No cirrhosis	0.62 (0.31, 1.23)
Ascites	0.56 (0.27, 1.13)	0.69
No ascites	0.43 (0.15, 1.22)
Hepatic encephalopathy	0.72 (0.34, 1.56)	0.23
No hepatic encephalopathy	0.35 (0.14, 0.88)
MELD-Na < 15	0.30 (0.07, 1.24)	0.42
MELD-Na ≥ 15	0.56 (0.29, 1.08)
Acute kidney injury	0.81 (0.35, 1.89)	0.19
No acute kidney injury	0.36 (0.16, 0.84)
Chronic kidney disease	1.96 (0.68, 5.66)	0.02
No chronic kidney disease	0.41 (0.21, 0.82)
Congestive heart failure	0.39 (0.05, 2.92)	0.85
No congestive heart failure	0.47 (0.25, 0.90)
COPD	0.65 (0.15, 2.76)	0.73
No COPD	0.49 (0.26, 0.94)
Mental health disorder	0.34 (0.05, 2.42)	0.67
No mental health disorder	0.53 (0.28, 0.99)
Substance use disorder	0.60 (0.21, 1.70)	0.73
No substance use disorder	0.47 (0.23, 0.97)
Length of stay < 5 days	0.33 (0.10, 1.06)	0.34
Length of stay ≥ 5 days	0.64 (0.32, 1.26)
Discharge to home	0.50 (0.26, 0.96)	0.76
Discharge to facility	0.63 (0.17, 2.43)

Abbreviations: AKI, acute kidney injury; CI, confidence interval; COPD, chronic obstructive pulmonary disease; HE, hepatic encephalopathy; MELD-Na, Model for End-Stage Liver Disease; NAFLD, nonalcoholic fatty liver disease; NE, not estimable; SD, standard deviation; SHR, subhazard ratio; TLC, Transitional Liver Clinic

DISCUSSION

The TLC is a patient-centered intervention leveraging a multidisciplinary team (PA, nursing, social work, nutrition, pharmacy) to improve the lives of patients with liver disease. By intervening early after discharge, the TLC guides patients through the inpatient/outpatient transition, a time when changes and handoffs can lead to poor outcomes.¹⁷ This study demonstrates a potential reduction in readmissions for those attending TLC, suggesting that such interventions may have a beneficial role in hepatology.

In non-liver populations, interventions to reduce readmissions have had mixed results.^18,19 Effective interventions are bundled, complex, delivered by multiple providers/interactions, and enhance patient capacity for self-care.¹⁹ The TLC leverages these ingredients by incorporating different providers, and it enhances self-care capacity by linking patients to additional resources, including substance use disorder treatment. Elsewhere, alcohol rehabilitation has reduced alcoholic hepatitis readmissions.⁵ Beyond alcohol, other liver disease features make this population vulnerable to readmission. These features include frequent comorbidities, ascites, and hepatic encephalopathy, which are common causes of readmission.^1,2,7 At discharge, patients with cirrhosis often feel unprepared to manage their needs.¹⁷ The TLC addresses these challenges with the backbone of an experienced PA, providing expert care after discharge, and this expertise is seen in the trend toward reduced readmissions for ascites, hepatic encephalopathy, AKI, and electrolyte disturbance. Elsewhere, PAs have been associated with improved quality of care and outcomes in cirrhosis.²⁰

Interventions for patients with liver disease have met with mixed success. A disease management program improved quality of care, but trended toward increased hospitalizations.⁸ Likewise, early follow-up in the VA was associated with increased readmissions.²¹ A telephone-only intervention also did not reduce readmissions.⁹ Two intensive programs did reduce readmissions, but these relied on same-day availability of radiology and endoscopy, which may not be scalable.^10,11 In contrast, the TLC incorporates routine visits, phone calls, and services such as nursing, nutrition, and social work that can be mobilized broadly. The TLC also leverages CMS incentives, which provide payment for transitional care management (TCM).²²

TCM payments were established in 2013, with a goal of encouraging care coordination.²² They require several components, including communication within 2 business days, medication reconciliation, a face-to-face visit within 7–14 days, and non-face-to-face services.²³ TCM programs have reduced readmissions in other populations, though the evidence remains limited.²³ Our study offers evidence that TCM payments can encourage effective programs, and it does so in a high-risk population. Additionally, our finding that the phone call improved attendance shows that early communication is key to its effectiveness.

TLC was associated with reduced readmissions in patients with and without cirrhosis, regardless of comorbidities and disease severity. These findings support the TLC’s effects beyond cirrhosis. Despite these data suggesting a broad impact, subgroup analyses were exploratory, and more work is needed to target the highest risk population. Emerging data show excessive readmissions in alcoholic hepatitis, and TLC has the potential to positively impact this group.^3–5 The benefit of the TLC may be greater in those without CKD, highlighting the challenge of volume management in comorbid liver and kidney impairment, where diuretic therapy is limited. Volume overload and AKI were among the most common reasons for readmission. Reduced readmissions with TLC in women compared to men was unexpected. Reasons for this differential effect may relate to sex-based differences in comorbidities and disease etiologies. Readmissions among patients with cirrhosis are easier to predict in women compared to men.²⁴

We acknowledge several limitations. Although improved attendance with the phone call highlights its importance, we were unable to deconstruct the TLC to assess which components were responsible for improved outcomes. Additionally, as an observational study, the results must be interpreted with caution; we were unable to ascertain all reasons for non-attendance from the EHR. We attempted to limit confounding through a propensity score-adjusted model; however, the potential for residual confounding remains. Immortal time bias is a concern since the clinic visit typically occurs 1–2 weeks post-discharge. Notably, the 7-day landmark analysis did not show improvement with TLC; however, this analysis was limited, as 30% of readmissions were excluded. In addition, we also note that the TLC is a bundled intervention including an early phone call (after July) and additional patient contacts as needed. Attendees who received a phone call had numerically lower 7-day readmissions that could help to account for the early signal favoring the TLC. These limitations highlight the need for prospective studies of transitional care interventions. In contrast to these limitations, our study benefits from a diverse sample, including those without cirrhosis, enhancing generalizability. In addition, mortality was not associated with TLC (but was associated with with other expected variables), supporting internal validity.

The TLC is a novel, scalable intervention that may be associated with reduced readmissions in patients with liver disease. Prospective work is needed to establish its efficacy and to refine the intervention, focusing on its most effective aspects to allow for dissemination and localization to other healthcare systems that care for large populations of patients with liver disease.

CLINICAL SIGNIFICANCE.

• Patients with liver disease have high rates of hospital readmission.

• Effective, scalable interventions to reduce this risk are lacking.

• A Physician Assistant-led Transitional Liver Clinic may be associated with reduced readmissions.

• The effects of the Transitional Liver Clinic are similar across liver disease severities and etiologies.