Patients with liver disease often require drug therapy. Drug pharmacokinetics and pharmacodynamics can vary dramatically in these patients and careful consideration is needed in drug choice and dosage to avoid adverse effects. In this study, the choice of drugs and their dosage were evaluated in a population of patients with severe liver disease. A retrospective study, using routinely collected hospital and pharmacy data, was conducted among adult patients diagnosed with decompensated liver cirrhosis. Drug choice and dosage were evaluated when patients were admitted to the hospital (at least 3 months after diagnosis). Recommendations in the summary of product characteristics (SPC), reference-books and medical literature were used for comparison. Medication errors were divided into the following categories: ‘contra-indicated drug’, ‘wrong dose’, ‘required monitoring not executed’, and ‘wrongly discontinued’. Forty-one patients were included in this study. Their mean age was 59 years and 78% of patients had a history of alcohol abuse. One-third of patients had decompensated liver cirrhosis at the defined evaluation moment. Seventy-three medication errors were identified in 355 prescriptions (22%). Most medication errors were of the categories ‘contra-indicated drugs’ and ‘wrong dose’. Contra-indicated drugs mainly consisted of oral antidiabetics (35%), benzodiazepines (17%), statines (13%) and slow release iron formulations (13%). Wrongly dosed drugs were proton pump inhibitors (65%), paracetamol (15%) and tramadol (8%), mainly prescribed by protocol. Although recommendations on drug choice and dose in patients with severe hepatic dysfunction were present, 1 out of 5 medication orders were incorrect. Most medication errors were made when drugs were prescribed by protocol. Severe hepatic dysfunction was barely taken into account in these cases. Recommendations on drug choice and dose are available to a large extent and, if not, can be easily deducted with basic knowledge of human pharmacology. This population of patients seems particularly suited for the application of personalized medicine.