A study of patients with cirrhosis admitted to nontransplant general intensive care in the UK: prevalence, case mix, outcomes and evaluation of critical illness and disease-specific scoring systems

Introduction

Hospital admission rates attributable to cirrhosis are increasing in the UK. The evidence base for the outcome of critically ill patients with cirrhosis admitted to intensive care (ICU) is currently limited to data from tertiary transplant centres, which report mortality rates of between 45 and 74%. We hypothesised that the disease cohort experienced in secondary care is less severe than that reported in the literature and would subsequently demonstrate lower mortality. We report the prevalence, case mix and outcomes of a dual-centre ICU study in London, UK.

Methods

Data were collected prospectively from two large nontransplant general ICUs in London (St George's and St Thomas' Hospitals) over 20 months (November 2007 to June 2009). All ICU admissions were screened for cirrhosis. Clinical data were recorded to calculate and evaluate several critical illness and disease-specific scoring systems (APACHE II, SAPS II, SOFA, Child Pugh (CPS), MELD, UKELD, Glasgow Alcoholic Hepatitis Score and the Royal Free Hospital (RFH) score). Study endpoints were ICU and hospital mortality.

Results

Cirrhosis accounted for 3.3% (137/4,198) of ICU admissions. In 118 patients meeting inclusion criteria, ICU and hospital mortality rates were 38% and 47%. Median (IQR) age was 50 (43 to 59) years, 68% were male, median length of ICU stay was 4.5 (2 to 10) days and 72% had alcoholic cirrhosis. Median CPS was 10 (8 to 11) (45% Grade A/B), MELD 18 (12 to 24) and APACHE II 16 (13 to 22). Hospital mortality was 24% in CPS grade B and 66% in grade C. After multivariate analysis, factors independently predictive of mortality were bilirubin, INR, urea, bicarbonate and pO₂/FiO₂ ratio. AUROC for the scoring systems ranged from 0.76 (UKELD) to 0.81 (RFH) for the prediction of mortality. Renal failure (OR 3.4 (1.4 to 8.4)) and haemofiltration during admission (OR 9.6 (3.7 to 24.8)) had strong associations with mortality. Intubation for gastrointestinal bleeding OR 0.4 (0.1 to 1.3) was associated with a trend towards better outcome than intubation for respiratory failure OR 2.7 (1.0 to 7.0).

Conclusions

The mortality rates and disease stage reported here are lower than those described in the established literature. Data from specialist transplant centres should not be applied to the patient cohorts seen in general ICUs. We urge critical care physicians to carefully consider the individual clinical case when deciding on ICU admission and not apply potentially misrepresentative figures from previous datasets.