Regional variations in inpatient decompensated cirrhosis mortality may be associated with access to specialist care: results from a multicentre retrospective study

Abstract

Introduction

Specialist centres have been developed to deliver high-quality Hepatology care. However, there is geographical inequity in accessing these centres in the United Kingdom (UK). We aimed to assess the impact of these centres on decompensated cirrhosis patient outcomes and understand which patients transfer to specialist centres.

Methods

A UK multicentred retrospective observational study was performed including emergency admissions for patients with decompensated cirrhosis in November 2019. Admissions were grouped by specialist/non-specialist centre designation, National Health Service region and whether a transfer to a more specialist centre occurred or not. Univariable and multivariable comparisons were made.

Results

1224 admissions (1168 patients) from 104 acute hospitals were included in this analysis. Patients at specialist centres were more likely to be managed by a Consultant Gastroenterologist/Hepatologist on a Gastroenterology/Hepatology ward. Only 24 patients were transferred to a more specialist centre. These patients were more likely to be admitted for gastrointestinal bleeding and were not using alcohol. Specialist centres eliminated regional variations in mortality which were present at non-specialist centres. Low specialist Consultant staffing numbers impacted mortality at non-specialist centres (aOR 2.15 (95% CI 1.18 to 4.07)) but not at specialist centres. Hospitals within areas of high prevalence of deprivation were more likely to have lower specialist Consultant staffing numbers.

Conclusions

Specialist Hepatology centres improve patient care and standardise outcomes for patients with decompensated cirrhosis. There is a need to support service development and care delivery at non-specialist centres. Formal referral pathways are required to ensure all patients receive access to specialist interventions.

WHAT IS ALREADY KNOWN ON THIS TOPIC.

• Previous reports have demonstrated that care for UK inpatients with decompensated liver disease is suboptimal with low numbers of patients being managed by specialists on specialty wards.

• Hepatology specialist centres have been developed to deliver high-quality hepatology care but there are regional inequities in access to these services.

• To address this, regional ‘hub-and-spoke’ networks have been developed.

WHAT THIS STUDY ADDS

• Few patients transferred to a more specialist centre with the main indication for transfer being gastrointestinal bleeding.

• While specialist centres did not improve overall adjusted mortality, they reduced regional variations in mortality which were present at non-specialist centres.

• Non-specialist centres in areas of deprivation were more likely to have lower Consultant Gastroenterologist/Hepatologist numbers.

• Mortality at non-specialist centres was associated with lower Consultant Gastroenterologist/Hepatologist staffing.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• Ongoing work is required to support service development and care delivery at non-specialist centres with a focus on areas of deprivation.

• There is a need to establish formal referral pathways to ensure equitable access to specialist care.

Introduction

Liver disease is the second leading cause of working years of lives lost in Europe and is associated with increased standardised mortality in the United Kingdom (UK).^{1 2} Previous reports have demonstrated that less than half of UK inpatients with alcohol-related liver disease (ARLD) receive good care.³ Ensuring all patients with decompensated cirrhosis receive optimal care has become a priority to reduce mortality.

Specialist Hepatology centres, which provide services, including transjugular intrahepatic portosystemic shunts (TIPS), hepatocellular carcinoma (HCC) management, liver transplantation, and have dedicated Hepatology inpatient and outpatient services,⁴ have been developed to deliver high-quality care for patients with liver disease. However, these centres have not been developed to reflect patient need either by geographical access or by disease burden.^{5 6} While efforts have been made to improve the standard of care for all patients with decompensated cirrhosis,⁷ distance from specialist centres has been demonstrated to adversely impact outcomes in a number of settings, including liver transplantation.^{8 9} The concept of ‘hub-and-spoke’ care provision has developed over the past decade to provide better access to specialist care via regional networks.¹ However, these relationships remain mostly informal with criteri for referral and transfer ill-defined. There is a need to understand whether these networks are used and the impact they have on inpatient outcomes.

In this study, our primary aim was to evaluate the impact of specialist Hepatology centres on mortality of inpatients with decompensated cirrhosis. Our secondary aims were to explore the difference in care provision between specialist Hepatology centres and non-specialist centres and to define how frequently referral networks are used to provide access to specialised services.

Methods

Study design

This was a UK multicentre, retrospective, observational cohort study, including patients emergently admitted to acute hospitals with decompensated cirrhosis in November 2019. November 2019 was selected due to the lack of UK public holidays, junior doctor changeovers and predating the COVID-19 pandemic. Trainees were invited to participate in this study via the Trainee Collaborative for Research and Audit in Hepatology UK network.¹⁰ Sites were invited by email through society mailing lists (British Association for the Study of the Liver, British Society of Gastroenterology (BSG), Scottish Society of Gastroenterology, and Welsh Association for Gastroenterology and Endoscopy) and through Twitter via the @uk_torch account. We used the National Health Service (NHS) Health Research Authority application, which did not consider the study research and was therefore exempt from ethical approval. The study was registered as an audit at all participating sites (Host site King’s College Hospital—audit reference number LIV16062021). Patient hospital admissions were identified via ICD-10 codes used by NHS England Cirrhosis Quality Dashboard.¹¹ Admissions were excluded if they were for patients; less than 18 years old, not admitted as an emergency, not felt to have decompensated liver disease (histologically, radiologically or clinically) or admitted for alternate non-liver pathology, included in clinical trials during that admission, transferred from an alternate hospital, previous liver transplant, known active non-liver cancer, metastatic liver cancer or were pregnant (online supplemental figure 1). Reporting of the analysis of this study complies with Strengthening the Reporting of Observational studies in Epidemiology guidelines for the reporting of cohort studies.¹² Pseudonymised data were transferred in a standardised password-protected Microsoft Excel data collection spread sheets via encrypted NHS email addresses.

Data collection

Each registered site was issued with guidance to assist with data collection (online supplemental methods). Data were acquired for each hospital regarding whether it was designated as a specialist Hepatology training centre (determined by the BSG, see online supplemental methods ⁴). Hospitals were grouped by NHS region in England with hospitals in Northern Ireland, Scotland and Wales grouped into Health and Social Care Northern Ireland, NHS Scotland, and NHS Wales, respectively.¹³ Subregional data regarding estimated local population per full-time equivalent (FTE) Consultant Gastroenterologists and Hepatologists were provided by the Royal College of Physicians.¹⁴ This was dichotomised to above/below the UK mean for 2019–2020 (44 155 people/FTE Consultant Gastroenterologist/Hepatologist). English NHS Trust data regarding social deprivation were acquired from NHS Digital in the form of proportion of patient admissions at each trust within deprivation quintile 5 (least deprived cohort).¹⁵ This was dichotomised to above or below the national mean for November 2019.

Patient notes were interrogated for demographic, clinical and laboratory data at admission and clinical outcome data following the first 24-hour postadmission. Aetiology of liver disease was recorded as per documentation within the clinical notes. Dual aetiology of liver disease was permitted and patients were categorised within each cohort (eg, non-alcoholic fatty liver disease (NAFLD) and concomitant ARLD). The primary presenting complaint leading to hospital presentation was recorded (eg, a patient presenting with a large volume gastrointestinal bleed but was jaundiced with low-grade hepatic encephalopathy and an acute kidney injury (AKI) was designated as a gastrointestinal bleed). Laboratory and clinical data were recorded for calculation of the Child-Pugh, Model for End-stage Liver Disease (MELD) and UK End-stage Liver Disease Scores.^16–19 Time and date of patient admission were recorded, which allowed admissions to be classified as in hours (Monday to Friday between 09.00 and 17.00) or out of hours (OOH). Further clinical data after the first 24 hours of care were recorded, including management on a specialist Gastroenterology/Hepatology ward, whether they were under the care of a Consultant Gastroenterologist/Hepatologist, transfer to a specialist Hepatology centre, admission to Critical care and survival till hospital discharge.

Data analysis

Comparisons were made between specialist and non-specialist Hepatology centres, NHS regions, hospitals within areas of higher and lower deprivation and between patients transferring to more specialist centres and those who did not. If a patient admission resulted in a transfer to another centre, they were excluded from the in-hospital mortality analysis due to lack of information regarding survival/mortality from that encounter. Admissions to transplant centres were excluded from analysis of transfers to a more specialist centre. As being managed on a Gastroenterology/Hepatology ward was potentially a competing variable with Critical care admissions, this was analysed, including and excluding Critical care admissions separately. Continuous demographic, clinical and laboratory variables were analysed for normality using the D’Agostino and Pearson tests. All data were non-normally distributed and analysed using Mann-Whitney U tests (two groups) with results reported as median (IQR). Categorical data were analysed by Fisher’s exact tests (two groups) or χ² test (3+ groups) and results were reported as number (%).

Multiple logistic regression was used to adjust for patient age, MELD score and critical care admission to understand association of specialist and non-specialist Hepatology centres, England NHS trust deprivation data and Consultant staffing with mortality. For assessment of NHS regions impact on mortality, regions were excluded from analysis if there was only one hospital at specialist/non-specialist level to not highlight individual hospital data. Variables within each model were recorded as an adjusted OR (aOR) and p value. Variables which achieved an r² of >0.5 with other variables within the model were excluded to reduce the risk of collinearity. Goodness-of-fit was recorded by pseudo-r², area under the curve (95% CI), Hosmer-Lemeshow statistic and log likelihood ratio statistic. Complete case analysis was used, excluding individuals with missing data. Correction for multiple comparisons was performed using the Benjamini-Hochberg procedure with a false discovery rate (FDR) set at 0.05.²⁰ All univariable and multivariable analyses were performed using Prism V.9.2.0 (GraphPad, San Diego).

Results

Comparison of specialist and non-specialist hepatology centres demographic, clinical and outcome data

Completed data sets were received from 104 hospitals across the UK. A total of 11 045 patient admissions were identified from coding. After application of the exclusion criteria (online supplemental figure 1), 1224 patient admissions from 1168 patients were included in the final analysis. The number of sites participating varied regionally (online supplemental table 1).

The median number of admissions across all centres was 9.00 (IQR 6.25–16.00) (figure 1A). While there was no significant variation in median number of admissions across the UK, there was significant regional variation in the proportion of admissions managed in specialist hepatology centres (p<0.0001*) (figure 1B). Admission numbers at specialist centres were higher than that at non-specialist centres (16.00 (IQR 10.25–20.00) vs 8.00 (5.00–11.00), p<0.0001*) (figure 1C,D).

Figure 1.

Distribution of patient admissions across specialist and non-specialist Hepatology centres and NHS regions. (A) Map demonstrating regional median admissions (see online supplemental table 1 for further details). No significant variation was demonstrated following Kruskall-Wallis testing (p=0.52). (B) Variation in proportion of regional admissions to specialist Hepatology centres across the UK (p<0.0001*). There were significantly more admissions to specialist Hepatology centres (n=36) (C) compared with non-specialist centres (n=68) (D) following Mann-Whitney U testing (16.00 (IQR 10.25–20.00) v 8.00 (5.00–11.00), p<0.0001*). Statistical significance signified by*. NHS, National Health Service.

There were no significant differences found between admissions at non-specialist and specialist Hepatology centres for patient age or sex, proportion of patients with ARLD, NAFLD or currently using alcohol, proportion of patients with known liver disease or previous decompensation or admissions OOH. Similarly, there were no significant differences found between non-specialist and specialist Hepatology centres for proportion of patients admitted primarily for AKI, ascites, encephalopathy, gastrointestinal bleeding, jaundice or sepsis. Significantly more patients were admitted to specialist Hepatology centres primarily for a reason not encompassed by the above categories. There were no significant differences between specialist and non-specialist centres for prognostic scores on admission (table 1A).

Table 1.

Comparisons between (A) specialist centres and non-specialist centres and (B) admissions resulting in patient transfer to a more specialist centre and those which did not

(A)
Variable	N	Non-specialist Hepatology centres, N=68	N	Specialist Hepatology centres, N=36	P value
Age	619	59.00 (48.00–68.00)	605	57.00 (48.00–67.00)	0.16
Male sex	619	369 (59.61%)	605	385 (63.64%)	0.16
Alcohol included in aetiology	619	472 (76.25%)	605	443 (73.22%)	0.24
NAFLD included in aetiology	619	88 (14.22%)	605	88 (14.55%)	0.87
Current alcohol use	455	246 (54.07%)	480	278 (57.92%)	0.26
Previously known liver disease	619	518 (83.68%)	605	523 (86.45%)	0.20
Previous known decompensation	619	421 (68.01%)	605	401 (66.28%)	0.54
Known HCC	619	36 (5.82%)	605	28 (4.63%)	0.37
Admitted out of hours	607	379 (62.44%)	601	384 (63.89%)	0.63
Primary reason for admission
Acute kidney injury	619	23 (3.72%)	605	21 (3.47%)	0.88
Ascites	619	223 (36.03%)	605	185 (30.58%)	0.05
Encephalopathy	619	105 (16.96%)	605	106 (17.52%)	0.82
Gastrointestinal bleeding	619	87 (14.05%)	605	93 (15.35%)	0.52
Jaundice	619	97 (15.67%)	605	88 (14.55%)	0.63
Sepsis	619	40 (6.46%)	605	39 (6.45%)	>0.9999
Other	619	44 (7.11%)	605	73 (12.07%)	0.004*
Prognostic scores
MELD score	548	17.00 (12.00–21.00)	554	17.00 (12.00–20.25)	0.33
UKELD score	548	57.00 (52.00–62.00)	554	56.00 (52.00–61.00)	0.18
Child Pugh score	526	9.00 (8.00–11.00)	545	9.00 (8.00–10.00)	0.71
Bundle usage
Any admission bundle use	619	72 (11.63%)	605	68 (11.24%)	0.86
After the first 24 hours
Managed by a Gastroenterologist/Hepatologist for admission	619	425 (68.66%)	605	476 (78.68%)	<0.0001*
Predominately managed on a Gastroenterology/Hepatology ward	614	305 (49.67%)	605	385 (63.64%)	<0.0001*
Critical care admission during stay	619	66 (10.66%)	605	66 (10.91%)	0.93
Predominately managed on a Gastroenterology/Hepatology ward   (excluding critical care admission)	548	282 (51.46%)	539	358 (66.42%)	<0.0001*
Admission mortality	601	108 (17.97%)	599	79 (13.19%)	0.03*

(B)
Variable	N	Transferred	N	Not transferred	P value
Age	24	57.50 (50.25–70.75)	1098	58.00 (48.00–67.25)	0.83
Male sex	24	14 (58.33%)	1098	683 (62.20%)	0.68
Alcohol included in aetiology	24	17 (70.83%)	1098	821 (74.77%)	0.64
NAFLD included in aetiology	24	5 (20.83%)	1098	159 (14.48%)	0.39
Current alcohol use	14	3 (21.43%)	848	478 (56.37%)	0.01*
Previously known liver disease	24	19 (79.17%)	1098	937 (85.34%)	0.38
Previous known decompensation	24	16 (66.67%)	1098	737 (67.12%)	>0.9999
Known HCC	24	1 (4.17%)	1098	55 (5.01%)	>0.9999
Reason for admission
Acute kidney injury	24	1 (4.17%)	1098	38 (3.46%)	0.58
Ascites	24	6 (25.00%)	1098	369 (33.61%)	0.51
Encephalopathy	24	3 (12.50%)	1098	192 (17.49%)	0.78
Gastrointestinal bleed	24	9 (37.50%)	1098	161 (14.66%)	0.006*
Jaundice	24	4 (16.67%)	1098	163 (14.85%)	0.77
Sepsis	24	0 (0.00%)	1098	64 (5.83%)	0.39
Other	24	1 (4.17%)	1098	111 (10.11%)	0.50
Prognostic scores
MELD score	22	21.50 (13.25–29.00)	985	17.00 (12.00–21.00)	0.05
UKELD score	22	59.00 (52.00–64.25)	985	56.00 (52.00–61.00)	0.15
Child Pugh score	20	10.00 (8.25–11.00)	967	9.00 (8.00–11.00)	0.11

Results from Mann-Whitney U tests presented as median (IQR). Results of Fisher’s Exact tests presented as number (%). Admissions for patients transferred between centres were excluded from this analysis.

*Statistical significance after correction for FDR.

FDR, false discovery rate; HCC, hepatocellular carcinoma; MELD, Model for End-stage Liver Disease; NAFLD, non-alcoholic fatty liver disease; UKELD, UK End-stage Liver Disease.

After the first 24 hours of care, admissions for patients at specialist Hepatology centres were significantly more likely to be managed by a Gastroenterologist/Hepatologist (78.68% vs 68.66%, p<0.0001*) on a Gastroenterology/Hepatology ward (63.64% vs 49.67%, p<0.0001*). This remained significant after exclusion of admissions, resulting in Critical care admission. There was no significant difference between specialist and non-specialist centres in proportion of patients admitted to Critical care (table 1A).

Comparison of patients transferred to a more specialist centre to those who were not

Twenty-four admissions resulted in a transfer to a more specialist centre. Despite this small number, admissions were more likely to result in transfer if they were for patients not currently using alcohol (21.43% vs 56.37%, p=0.01*) or patients admitted for gastrointestinal bleeding (37.50% vs 14.66%, p=0.006*). There were no significance differences between patients transferred and those not for age, sex, ARLD/NAFLD aetiologies and previous history of liver disease. While patients transferred to specialist centres did have higher MELD scores (21.50 (IQR 13.25–29.00) vs 17.00 (IQR 12.00–21.00), p=0.05), this was not statistically significant after correction for FDR (table 1B).

Impact of specialist Hepatology centres, Gastroenterology/Hepatology Consultant numbers and deprivation on admission mortality

Admissions to specialist Hepatology centres were associated with a lower mortality than those to non-specialist centres on univariable analysis (13.19% vs 17.97%, p=0.03*) (table 1). However, following adjustment for patient age, MELD score and critical care admission, specialist Hepatology centres were not significantly associated with reduced admission mortality (aOR 0.70 (95% CI 0.48 to 1.01)) (figure 2). There was no significant regional variation in adjusted admission mortality at specialist Hepatology centres (figure 3A). However, at non-specialist centres, there was significant regional variation in adjusted mortality with NHS North East and Yorkshire (aOR 2.37 (95% CI 1.25 to 4.43)) and NHS South West (aOR 3.36 (95% CI 1.41 to 7.74)) admissions associated with an increased likelihood of mortality, and NHS East of England (aOR 0.25 (95% CI 0.05 to 0.68)) and NHS Midlands (aOR 0.33 (95% CI 0.09 to 0.89)) admissions associated with a reduced likelihood of mortality (figure 3B).

Figure 2.

Impact of Hepatology specialist centres on mortality after adjustment for patient age, MELD score and critical care admission. OR plot demonstrating Hepatology specialist centres association with admission mortality when adjusted for patient age, MELD score and critical care admissions (n=1080, AUC 0.78 (95% CI 0.75 to 0.82), pseudo r²=0.22, HL Statistic 5.00 (p=0.76), log-likelihood ratio statistic 195.90 (p<0.0001*)). See online supplemental table 2 for variable performance within models. AUC, area under the curve; HL, Hosmer-Lemeshow; MELD, Model for End-stage Liver Disease.

Figure 3.

Assessment of regional variations in mortality across levels of centre following adjustment for patient age, MELD score and critical care admission and the impact of Consultant numbers and deprivation. (A) OR plot demonstrating adjusted mortality of specialist Hepatology centres at individual NHS regions when adjusted for patient age, MELD score and critical care admissions, (B) OR plot demonstrating adjusted mortality of non-specialist centres at individual NHS regions when adjusted for patient age, MELD score and critical care admissions, (C) OR plot demonstrating the lack of association of a lower sub-regional Consultant Gastroenterologist/Hepatologist: local population ratio at specialist centres with admission mortality when adjusted for patient age, MELD score and critical care admission (n=548, AUC 0.76 (95% CI 0.70 to 0.82), pseudo r²=0.16, HL Statistic 7.74 (p=0.46), log-likelihood ratio statistic 68.28 (p<0.0001*)). (D) OR plot demonstrating the association of a lower sub-regional Consultant Gastroenterologist/Hepatologist: local population ratio at non-specialist centres with admission mortality when adjusted for patient age, MELD score and critical care admission (n=532, AUC 0.81 (95% CI 0.76 to 0.86), pseudo r²=0.29, HL Statistic 9.87 (p=0.27), log-likelihood ratio statistic 131.50 (p<0.0001*)). (E) OR plot demonstrating the impact of deprivation on mortality of admissions at specialist centres. (n=441, AUC 0.73 (95% CI 0.66 to 0.80), pseudo r²=0.13, HL Statistic 7.18 (p=0.52), log-likelihood ratio statistic 44.96 (p<0.0001*)). (F) OR plot demonstrating the impact of deprivation on mortality of admissions at non-specialist centres. (n=469, AUC 0.81 (95% CI 0.76 to 0.86), pseudo r²=0.29, HL Statistic 6.10 (p=0.64), log-likelihood ratio statistic 116.00 (p<0.0001*)). See online supplemental tables 3,4,6,7,10 and 11 for variable performance within models. AUC, area under the curve; HL, Hosmer-Lemeshow; HSCNI, Health and Social Care Northern Ireland; MELD, Model for End-stage Liver Disease; NHS, National Health Service.

Admissions to subregions with a lower Consultant Gastroenterologist/Hepatologist:local population ratio were not significantly more likely to result in admission mortality on univariable analysis (online supplemental table 5). At specialist Hepatology centres, a lower Consultant Gastroenterologist/Hepatologist:local population ratio was not associated with a significant difference in adjusted mortality (aOR 0.89 (95% CI 0.51 to 1.55)) (figure 3C). However, at non-specialist centres, a lower Consultant Gastroenterologist/Hepatologist:local population ratio was associated with an increased likelihood of adjusted admission mortality (aOR 2.15 (95% CI 1.18 to 4.07)).

Comparisons were then made between centres within deprived areas (defined as a lower proportion of patients in deprivation quintile five compared with the UK mean) compared with centres who were not (defined as a higher proportion of patients in deprivation quintile 5 compared with the UK mean). The only significant difference between cohorts was that centres within deprived areas were more likely to have a lower Consultant Gastroenterologist/Hepatologist:local population ratio (86.54% vs 41.67%, p<0.0001*) (online supplemental table 8). This analysis was repeated at non-specialist centres only and similar results were demonstrated with a lower Consultant Gastroenterologist/Hepatologist:local population ratio the only variable which was significantly different between groups (88.89% vs 54.55%, p=0.005*).

On univariate analysis, deprivation did not impact admission mortality across the cohort (64.94% vs 58.64%, p=0.15) (online supplemental table 8). Admission to a hospital in an area of deprivation had no impact on adjusted mortality at specialist centres (aOR 1.07 (95% CI 0.59 to 1.95) (figure 3E). However, admission to a hospital in an area of deprivation was associated with increased adjusted mortality at non-specialist centres (aOR 1.91 (95% CI 1.05 to 3.61)) (figure 3F).

Discussion

In this study, we have provided a detailed snapshot of care provision for patients with decompensated cirrhosis across the UK. We demonstrate that specialist Hepatology centres have more admissions than non-specialist centres, and patients at these specialist centres are more likely to be managed by a Consultant Gastroenterologist/Hepatologists on dedicated specialist wards. We also demonstrate that there were no significant regional variations in mortality adjusted for patient age, MELD score and critical care admission at specialist Hepatology centres, which is in contrast to non-specialist centres. Specialist Consultant staffing significantly impacts adjusted mortality at non-specialist centres, whereas it does not impact adjusted mortality at specialist centres. While deprivation impacts on patient outcomes at non-specialist centres, this may be in part due to lower specialist Consultant staffing. However, referral networks to transfer inpatients to more specialist Hepatology services are rarely used.

The first Lancet Commission Report into liver disease in the UK highlighted that only 20% of patients with liver disease had an interaction with a Consultant Gastroenterologist/Hepatologist and few were managed on specialty wards during hospital admissions.¹ Our findings may suggest that there has been increased focus on providing specialist care across UK centres. It is unsurprising that specialist Hepatology centres have significantly more patients managed by Consultant Gastroenterologists/Hepatologists given that this was one of the purposes of their design. The lack of regional variation in adjusted mortality at specialist centres combined with the lack of impact of specialist Consultant staffing at these centres likely reflect the design of these services to optimise outcomes for patients with decompensated cirrhosis. In contrast, non-specialist centres have to allocate both resource and expertise to manage patients across the breadth of Gastroenterology. Local expertise and specialist Consultant staffing are likely to have greater impact on care provision within these services. Worryingly, our data confirm previous findings from nearly 20 years ago demonstrating regional variations in mortality at non-specialist centres suggesting limited progress in addressing this healthcare inequity.¹ However, a number of initiatives have been recently developed to increase global standard of care. The Improving Quality in Liver Services (IQILS) accreditation programme has been developed to standardise care across the UK through service development and quality improvement.²¹ However, involvement with the IQILS accreditation programme is currently optional for trusts, which may need to be reviewed going forwards. Decompensated cirrhosis admission care bundles have been developed which have been shown to improve care for patients in the first 24 hours of admission.^{7 22} However, we have previously demonstrated that these admission care bundles are currently infrequently used across the UK.²³ Understanding and addressing the barriers to use are essential to maximise their potential impact. Finally, increasing Consultant Gastroenterologist/Hepatologist numbers at non-specialist centres remains a challenge. Nearly half of UK Consultant Gastroenterologist/Hepatologist advertised posts remained unfilled in 2020, and a recent survey demonstrated that less than one-third of Hepatology trainees would want to be a Consultant in a non-specialist centre.^{24 25} Innovative job planning, including roles split across specialist and non-specialist centres particularly in areas of deprivation, may address this. Ongoing work is required to support service development and care delivery at non-specialist centres with a focus on centres in areas of deprivation.

The low number of patients that transferred to a more specialist centre is concerning and may suggest that certain patients who may benefit from specialist interventions (eg, TIPS or transplantation) are not receiving the same care that they would receive at a specialist centre. The development of formal referral networks has long been used to improve access to specialist interventions. In Cardiology, the developments of networks to deliver percutaneous coronary interventions (PCI) following acute coronary syndrome are well established and have been associated with increased utilisation of PCI.²⁶ For patients with malignancy, formal referral networks have been demonstrated to reduce time waiting for investigations and interventions.^{27 28} Within Hepatology, formal referral networks for transplant assessment have been demonstrated to improve patient selection.²⁹ It is likely that developing formal networks with standardised referral criteria would improve equity in access for patients with decompensated cirrhosis to specialist interventions. It is also notable that admissions for gastrointestinal bleeding were significantly more likely to result in a transfer between centres, probably reflecting the need for emergency TIPS. The role of liver transplantation in acute-on-chronic liver failure (ACLF) continues to expand with a survival benefit demonstrated even among those with ACLF grade 3.³⁰ While only a small number of patients would be appropriate for life-saving transplantation and less than those requiring emergency TIPS, the low number of transfers may suggest that this is currently rarely considered for patients outside transplant centres. While there are challenges in transferring patients due to a combination of bed pressures and infection control measures, there is a need to formalise and standardise referral networks to ensure equitable high standards of care for patients with decompensated cirrhosis.

There are a number of limitations of this study, including the retrospective design. Given the ‘real-world’ nature of this study, clinical diagnoses were taken from the records as opposed to specific diagnostic criteria. The use of a single month may highlight anomalies in care provision and outcomes, which are not consistent over a longer period. While data were submitted from all NHS regions, some regions had a lower proportion of hospitals represented, which may have impacted on regional performance and lead to underpowered regional subgroup analysis. As trainees were invited to participate in this study, this may have led to selection bias, explaining why specialist Hepatology centres were over represented within the final data set. However, the design led to wide participation, including sites which are not typically captured in UK observational studies and provide more clinical detail than studies utilising national databases. Future studies should be prospectively designed utilising proformas with validated diagnostic criteria and include comparable populations from all NHS regions across a broad time period. Admissions were selected by coding, therefore incorrectly coded admissions would not have been detected. However, we used established NHS England coding and applied strict exclusion criteria to ensure quality of admissions included.¹¹ We have demonstrated that coding used for the NHS England Cirrhosis Dashboard lacks specificity for admissions with decompensated cirrhosis. There is significant variation in performance of conventional coding sets to detect cirrhosis.³¹ The NHS England Cirrhosis Dashboard uses a broader code set than most clinical studies, which may increase the sensitivity of detecting patients with cirrhosis but impact on specificity.¹¹ Further work is required to evaluate and optimise these codes to ensure the accuracy of data recorded. While we have excluded patients with previous liver transplants, non-liver cancer and metastatic HCC, we do not have data pertaining to non-liver non-cancer comorbidities. Given the younger age of this cohort and impact of liver disease severity, it is likely that the cause for inpatient mortality was secondary to patients’ underlying liver disease. Further work is required to understand the impact of non-liver comorbidities in this cohort of patients. Future studies should also aim to evaluate the impact of Hepatology Consultants compared with Gastroenterology Consultants, advanced care planning, specialist liver nurses and alcohol care teams, as well as other relevant healthcare providers, on inpatient outcomes with site-specific data. Utilising trust-level data for deprivation may not reflect patient-level deprivation data. Future studies should aim to collect data regarding deprivation on an individual level. Future studies should aim to capture outcomes of an appropriately powered sample of patients transferred to specialist centres to understand the impact of ‘hub-and-spoke’ care for inpatients with decompensated cirrhosis. However, the strengths of the study should also be acknowledged. This is a large-scale multicentred study with data submitted from all NHS regions with multiple data points encompassing patient care across each admission. Multivariable analyses were used to evaluate individual variables’ impact on admission mortality with relevant sensitivity analyses which reflect patient care provision.

In conclusion, specialist Hepatology centres provide inpatient care leading to comparable risk-adjusted mortality across regions. This is in contrast to non-specialist centres where specialist Consultant numbers significantly impact risk-adjusted mortality. Supporting service development and care delivery at non-specialist centres, with a focus on areas of deprivation, remains a priority. Very few patients transfer between centres which reflects a need to standardise and formalise referral pathways. This will allow equitable access for specialist life-saving interventions including TIPS and liver transplantation.

Data availability statement

Data are available upon reasonable request.

Ethics statements

Patient consent for publication

Not applicable.