Introduction Liver function tests (LFTs) are commonly measured upon admission to hospital. LFT derangement on admission not only prompts investigation of pathology but also predicts short-term mortality. Longer-term implications of short-term changes in these tests are less well-studied. We sought whether fluctuations in LFTs during admission predict 12-month mortality and, if so, which LFTs were the best predictors.

Method We included patients admitted non-electively for a period of at least 7 days to General Medicine at an urban Australian hospital from 2005 to 2012. LFTs (bilirubin, albumin, gamma glutamyl transferase, alkaline phosphatase, alanine aminotransferase (ALT) and aspartate aminotransferase (AST)) were measured on admission and repeated at least once with the greatest change in each noted. Only the first relevant admission of each patient was analysed. Patients were excluded if they died during admission, were admitted to Intensive Care or carried a diagnosis of malignancy or liver disease. A logistic regression model was developed to determine 12-month mortality incorporating patients’ age, gender, co-morbidity index, heart failure diagnosis and admission LFTs. A second model also included intra-admission changes in LFTs.

Results Of 4160 patients analysed, median age was 79 years (IQR 66–86), 25% had heart failure and 12-month post-discharge mortality was 17%. Admission albumin was low in 39% of patients; admission ALT was high in 14% of patients. Admission albumin predicted 12-month mortality better than other LFTs at admission. 12-month mortality was greater by: 37% (p = 0.02) with a standing diagnosis of heart failure; 4% (p < 0.001) with a 1g/L intra-admission decrease in albumin; and 6% (p = 0.05) with a 100 IU/L intra-admission increase in ALT. Comparison of the two models showed that the intra-admission change model was superior to the admission-value model at predicting 12-month mortality. Of intra-admission changes in LFTs, a drop in albumin was the best individual predictor of 12-month mortality; a rise in ALT added predictive power.

Conclusion Changes in LFTs predict long-term mortality better than a single value. In the absence of known liver disease, albumin predicts long-term mortality of elderly inpatients better than transaminases. This may reflect systemic influences on albumin such as nutritional status. Patients whose albumin falls during admission are at high risk of death within a year even if admission albumin levels are normal. Attention should be directed to reduce these patients’ mortality risk or consider planning end-of-life care.

Disclosure of interest None Declared.