Early and late mortality following unscheduled admissions for severe liver disease across England and Wales

Summary Background There is a known shortfall in hepatology service resources across England and Wales. Aim To investigate early and late mortality following unscheduled admissions for severe liver disease, overall and by cause of death, and to determine how mortality is related to admissions to transplant centres, transplant surgery, hospital size, consultant specialty, patient socio‐demographics, seasonal and geographical factors. Methods Cohorts of people with a first unscheduled admission for severe liver disease across England and Wales from 2004, based on record linkage of national inpatient and mortality data. Findings Mortality for alcoholic liver disease and hepatic failure was 23.4% and 35.4% respectively at 60 days and 61.8% and 57.1% at 5 years. Standardised mortality ratios (SMRs) were extremely high at 60 days (184 and 117 respectively) and remained highly increased at 5 years (16.7 and 6.3). Mortality at 5 years was most elevated from liver disease, viral hepatitis and varices. The 60‐day mortality was significantly lower for patients seen by consultant hepatologists and gastroenterologists. Both early and late mortality were significantly reduced for patients admitted to transplant centres or larger hospitals, who received a liver transplant, or were resident in London. Early mortality was significantly higher for patients admitted in winter and autumn, while elevated mortality among the most vs least deprived quintile increased with longer follow‐up. Conclusions The study shows a very poor prognosis for people with unscheduled hospitalisation for severe liver disease. The findings suggest that access to specialist expertise and services improves survival, both in the short and long term.


| INTRODUCTION
There is a known shortfall in specialist liver services resources across the UK, [1][2][3]
The main objectives of this study were, firstly, to establish early and late mortality following unscheduled admission for severe liver disease across England and Wales, both overall and by cause of death. The second main objective was to establish how early and late mortality are associated with admissions to transplant centres, liver transplantation, the hospital size, consultant specialty, patient socio-demographics, seasonal factors and the geographical region of England and Wales.
To provide confirmatory evidence, the study was based on two independently run UK National Health Services for which similar but separate data have been collected across England and Wales. The main study hypotheses were firstly, that mortality for severe liver disease would be very high, with greatly increased mortality from infections, liver cancer, accidents and suicide as well as from liver disease itself. Further hypotheses were, that mortality would be greatly reduced following liver transplantation; that early mortality would be improved for patients managed by trained hepatologists or gastroenterologists rather than by other specialists and by admission to specialist centres; and that late mortality would be worse for patients with the highest levels of social deprivation.

| MATERIALS AND METHODS
The study was based on retrospective cohorts of people admitted unscheduled for severe liver disease across England and Wales. The first cohort, to investigate early mortality, included each person's first admission for severe liver disease from the start of the study in The study cohorts were based on national administrative inpatient data, Hospital Episode Statistics for England (population 53.5 million in 2012) and the corresponding Patient Episode Database for Wales (population 3.07 million). The inpatient data were linked systematically to mortality data from the Office for National Statistics and the Welsh Demographic Service to identify deaths that occurred while in hospital or after discharge from hospital. The data were compiled and accessed through the Secure Anonymised Information Linkage (SAIL) databank. 23,24 The ascertainment of mortality has been validated as > 98% accurate and the record linkage methodology, based on a unique anonymised, encrypted linking field for each patient, as > 99.8% accurate. 23 Severe liver disease was defined as alcoholic liver disease (International Classification of Diseases, ICD, 10th revision code, K70) or hepatic failure (K72) when based on the principal diagnosis on the discharge episode. Alcoholic liver disease was also differentiated according to the three aetiologies, alcoholic hepatitis (K70.1), alcoholic liver cirrhosis (K70.3) and alcoholic hepatic failure (K70.4).

| Study exposure factors
Mortality was compared according to whether patients were admitted to one of the six hospitals in England in which liver transplant centres were located, in Birmingham, Cambridge, Leeds, London (2 centres) and Newcastle. 1 For patients admitted to transplant centres, we also assessed mortality according to whether or not they were local patients, defined as resident in the same local authorities in which the transplant centres were located. We also assessed mortality according to whether patients received a liver transplant (OPCS Classification of Interventions and Procedures, OPCS-4 code J01) in their index cohort admission.
The size of the admitting hospital was categorised by the total number of beds in five bands from < 400, 400-599, 600-799, 800-999 to 1000+ beds). Consultant specialty was based, firstly, on whether or not the patients were seen by a hepatologist or gastroenterologist and compared with all other specialties (recorded during either the first or last episodes of the admission). Secondly, on whether, the patients were seen by a critical care specialist, compared with all other specialties. The season of admission was assessed by comparing winter months (December to February) with autumn (September to November), summer (June to August) and spring (March to May) as the reference category.
Social deprivation was measured using the widely used English Indices of Multiple Deprivation (IMD) for England, 25 and the similar Welsh Index of Multiple Deprivation (WIMD) for Wales. 26 The total IMD and WIMD deprivation scores for geographical Lower Super Output Areas (LSOAs) (average LSOA population = 1640 in England and 1580 in Wales) were ranked and categorised into quintiles (I = least deprived and V = most deprived quintile).
The regions of England and Wales were based on the patients' recorded residence and the conventional Government Office Regional classification which includes 10 regional categories. 27

| Outcome measures
The main outcome measures were 'early mortality', defined as all deaths within 60 days of admission and 'late mortality', all deaths within 5 years of admission. Secondary outcome measures were standardised mortality ratios (SMRs) at 60 days and 5-year follow-up and also relative survival at monthly intervals up to 5 years. They ROBERTS ET AL. | 1335 were used to compare mortality in the cohorts of patients hospitalised for severe liver disease with those in the corresponding general populations of England and Wales. Cause-specific SMRs at 5year follow-up were based on the underlying causes of death on death certificates.

| Methods of analysis
Age and sex adjusted SMRs were calculated using the indirect method, by applying age and sex specific mortality in the general adult resident populations of England of Wales to obtain the expected mortality and by then comparing observed and expected mortality. The age groups used were 18-19 years, 20-24, quinquennially up to 80-84 and then 85+ years. Relative survival was calculated as a ratio to compare the observed survival in the cohorts of patients hospitalised for severe liver disease with that expected in the corresponding (age-and sex-matched) general populations of England and Wales.
Multivariate logistic regression models were used to assess associations between the study exposure factors and subsequent mortality. In the models, mortality was adjusted for patient age (in 5-year groups from 35 to 85+ years, with < 35 years as the reference category), sex and 10 major patient co-morbidities (ischaemic heart dis- 18.5%, 13 531). Missing data were excluded from the analyses involving the respective factors. There were no missing data for patient age, residential region, admissions to transplant centres, cause of death and day of admission or death.
Additionally, through record linkage of the inpatient data, we identified subsequent emergency admissions for severe liver disease in the study cohort patients and compared survival according to the numbers of subsequent admissions. When assessing trends in early mortality for alcoholic liver disease and for hepatic failure, annual mortality rates were standardised using the direct method and the total populations of patients admitted as the standard populations.
Logistic regression modelling was used to obtain mean annual reductions over time in the age and sex adjusted early mortality rates for alcoholic liver disease and hepatic failure. Other methods used were unpaired t tests to compare patient ages, chi-squared tests to compare patient co-morbidities and socio-demographics and the Mann-Whitney test to compare lengths of inpatient stay. Statistical significance was measured at the conventional 5% level.

| RESULTS
In the first study cohort with 60-day follow-up to investigate early mortality, there were a total of 73 123 patients hospitalised with severe liver disease; mean age = 52.8 years (SD = 12.7) and 35.3% were female. In the second cohort with 5-year follow-up to establish late mortality, there were a total of 33 726 patients; mean age = 52.5 years (SD = 12.2) and 35.0% were female.

| Early and late mortality
For England and Wales combined, early 60-day mortality (cohort 1) for alcoholic liver disease and hepatic failure was 23.4% and 35.4% respectively ( Table 1). Late 5-year mortality was respectively 61.8% and 57.1% (cohort 2). When comparing mortality with the corresponding general resident populations of England and Wales, SMRs were higher throughout follow-up for alcoholic liver disease than for hepatic failure, they were extremely high at 60-day follow-up (184 for alcoholic liver disease and 117 for hepatic failure, compared with 1.0 in the general population) and remained highly elevated in the longer term (16.7 for alcoholic liver disease and 6.3 for hepatic failure at 5 years).
Early survival was substantially worse following admission with hepatic failure than for alcoholic liver disease, approximately equal by 3-year follow-up, but over longer term follow-up, it deteriorated much more severely for alcoholic liver disease ( Figure 1A). Of the three main categories of alcoholic liver disease, prognosis was worst for alcoholic liver failure, followed by alcoholic liver cirrhosis, but comparatively better for alcoholic hepatitis ( Figure 1B).
Early 60-day mortality fell significantly over time during the study period, although this was greater for hepatic failure than for alcoholic liver disease ( Figure 2). The mean annual reduction was 3.4% (95% CI = 1.8%-5.1%) for hepatic failure and 0.7% (0.0%-1.4%) for alcoholic liver disease. Table 2 shows cause-specific mortality (SMRs) for late mortality (5-year follow-up) separately for patients admitted with alcoholic liver disease or with hepatic failure. At 5 years, liver disease accounted for 72.2% of all deaths among those originally admitted with alcoholic liver disease and 51.0% of those admitted with hepatic failure (Table 2). Late mortality was extremely increased (SMR > 100) from most of the major causes of death from liver disease and also from viral hepatitis and varices (although the SMR for varices was lower at 71.7 in patients admitted for hepatic failure).

| Cause-specific mortality
Late mortality was also very highly increased (SMRs > 10) from liver cancer, acute and chronic pancreatitis, septicaemia, infectious diseases generally, mental and behavioural disorders due to alcohol and from accidental poisoning ( Table 2). Late mortality was also very highly increased (SMRs > 10) among patients admitted with alcoholic liver disease, but not with hepatic failure, from certain gastrointestinal diseases (oesophageal ulcer, perforated peptic ulcer and peritonitis and herniae), accidental falls, mental and behavioural disorders generally and due to drugs.
There was relatively little or no increased mortality from cancers overall and from the major types of cancer, with the exceptions of mouth and throat cancer in people admitted with alcoholic liver disease (SMR = 7.9) and pancreatic cancer in those admitted with hepatic failure (SMR = 5.5; Table 2). Among patients admitted for hepatic failure, there was little or no increased late mortality from circulatory diseases generally, ischaemic heart disease, respiratory diseases generally, COPD and pneumonia.

| Factors that may influence prognosis
Early 60-day mortality but not late 5-year mortality was significantly lower among patients seen by hepatologists or gastroenterologists (by 22.3% compared with other specialties; Table 3). Both early and late mortality were significantly much lower among patients admitted to specialist transplant centres (by 26.5% and 38.6% respectively), who received liver transplants (by 4-and 7-fold respectively; Table 3; Figure 3) and who were resident in London compared with all other major regions of England and Wales. Both early and late mortality were also significantly reduced among patients admitted to larger compared with smaller hospitals, although this pattern was stronger for early mortality.
Early but not late mortality was significantly higher among patients admitted during winter and autumn compared with other months (Table 3). Among women compared with men, early mortality was significantly increased (by 6.3%) but late mortality was reduced (by 10.8%). Mortality was also significantly worse among patients with the highest levels of social deprivation compared with the lowest (quintile V vs I) and this differential increased from early mortality (9.2%) to late mortality (21.0%). There were no other significant differences between deprivation quintiles.
Both early and late mortality were greatly increased among patients seen by critical care consultants, by 9-fold and 5-fold respectively ( Table 3). Compared with all other patients admitted acutely with severe liver disease, those seen by critical care specialists were less often: diagnosed with alcoholic liver disease (73.8% vs 87.9%; P < 0.001), male (59.4% vs 64.8%; P = 0.002), and (non-significantly) socially deprived (54.6% vs 58.3% in the two most deprived quintiles; P = 0.067) but were of similar ages (mean = 52.2 vs 52.8 years; P = 0.201). They also had longer inpatient stays (median = 14.0 vs 9.0 days; P < 0.001).
T A B L E 1 Early and late mortality rates and standardised mortality ratios (SMRs) following unscheduled admissions for severe liver disease in England and Wales overall and according to aetiology

| Patients admitted to transplant centres
Among patients admitted to transplant centres, both early and late mortality were substantially higher among those resident locally, compared with patients admitted from other local authorities ( Of the major co-morbidities, local patients had higher levels of Year of admission

| DISCUSSION
The study shows a very poor overall prognosis following unscheduled hospital admissions for severe liver disease. Mortality was increased by more than 100-fold in the short term and remained highly elevated throughout 5 years of follow-up. Mortality was most extremely elevated from liver disease, viral hepatitis and varices.
Early 60-day mortality was substantially lower for patients seen by consultant hepatologists or gastroenterologists. Both early and late mortality were reduced for patients admitted to specialist transplant centres and larger hospitals, those who received liver transplantation, or were resident in London. Early mortality was higher for patients admitted in winter and autumn months, while late mortality was increased with higher levels of patient social deprivation.
Our findings of high early mortality and very poor long-term prognosis for people hospitalised with severe liver disease are con- Long-term prognosis following admissions for alcoholic liver disease is known to be very poor, largely since many cases present with decompensated liver disease when first seen, and subsequent abstinence rates can be low. 8,28 Although early mortality was much higher for (non-alcoholic) hepatic failure than for alcoholic liver disease, by 6 months most of the excess mortality for hepatic failure T A B L E 3 Early and late mortality following unscheduled admissions for severe liver disease in England and Wales in relation to consultant specialty, hospital size, admissions to liver transplant centres, geographical residential region, season and socio-demographics in patients who received liver transplantation and this reduction widened with longer term follow-up.

No. of admissions
As expected, both early and late mortality were greatly increased in patients seen by critical care specialists. The highest mortality in winter months may also be linked to access as resources become more strained in winter due to increases in admissions for seasonal illnesses, although it may also reflect seasonal variation in alcohol consumption.
We found that late but not early mortality was significantly worse among patients with higher levels of social deprivation and also in Wales compared with England overall. This would reflect more general social inequalities in health, 29,30 and have previously been reported for severe liver disease, 16,28 with increases in equalities over the longer term rather than soon after hospital admission and treatment.

| Strengths and limitations
Major strengths are that this study is national, one of the largest investigations on prognosis following admission for severe liver disease and it provides new evidence on factors that influence both early and late mortality. The study uses systematic validated record linkage methodology that has been used extensively in previous publications and it covers more than 70 000 admissions for severe liver disease and over 20 000 subsequent deaths. Importantly for confirmatory purposes, it is based on independently collected but similar information sources covering two different populations. The inpatient data sources are based on public hospitals, but these would account for almost all unscheduled admissions for severe liver disease in the two populations.
Limitations are that the national administrative inpatient data used in this study lack detailed information about disease history, disease severity, any therapeutic treatments and also alcohol consumption. Although the study cohorts are based solely on each patient's first admission for severe liver disease during the study period, the information sources do not provide adequate details of how they relate to any possible long-term previous history of liver disease. We assessed patients who received liver transplants during their study admissions, but were not able to identify from our data patients who subsequently received transplants electively. The principal diagnosis used to determine admissions for severe liver disease is also not accurate in all cases, 31 (1788) hepatology separately from gastroenterology. However, this should still have enabled ascertainment of most of the patients who were managed by hepatologists, gastroenterologists or critical care specialists. Late 5-year mortality would also be affected by cohort attrition, mainly through population emigration. However, emigration from England and Wales was less than 2.8% per annum during the study period, 32,33 and it would probably be substantially lower among people with severe liver disease than among the general population. For these reasons, and also since our findings are similar to those from previous cohort studies, [20][21][22] cohort attrition should be small.
In summary, the study shows a very poor prognosis for people admitted unscheduled for severe liver disease, and several factors that are strongly associated with survival both in the short and long term. In the longer term, prognosis for (non-alcoholic) hepatic failure is considerably better than for alcoholic liver disease. The study suggests that better access to expertise and specialist services improves survival, both in the short and long term.