Aims: This study compared different methods of assessing self-reported alcohol use among emergency room patients in order to explain the variations in reported prevalence rates. Methods: Alcohol use prior to patient's injury or illness was assessed in one hospital by a self-report questionnaire in three different ways: (i) administered by emergency room staff, (ii) administered by research staff, or (iii) sent to the patient's home by post. Results: Results show variations in self-reported alcohol use 6 h prior to the injury or illness ranging from 4.6 to 9.1%; these variations may be explained by sample selection bias and characteristics of the included study populations. When self-report is combined with staff judgement the corresponding prevalence rates are 6.8% for research staff and 16.2% for emergency room staff. This shows that the latter judge the patient's alcohol use more efficiently than the research staff. Using research staff 24 h a day resulted in almost no sample bias. Data collection via emergency room staff leads to the highest alcohol use prevalence rates and to the highest sample bias; this was influenced by the emergency room characteristics. A retrospective mail survey results in an older sample with age-related (lower) alcohol use and emergency room characteristics related to this age group. Conclusions: Future studies using patient self-report among emergency room samples should consider carefully the influence of sample selection bias. The combination of the research staff handing out the questionnaire and the emergency room staff giving their judgement on the patient's alcohol use seems to be a useful method.