This article describes the implementation of a simple method of drug surveillance set up at a hospital emergency ward. From a total of 48678 patients admitted, the medical records of those presenting with one or more of a pre-established list of admission diagnoses (n = 7728; 15.8%) were checked. Of these 554 (1.1%) were diagnosed as experiencing an adverse drug reaction. When the medical record suggested an adverse drug reaction, drugs taken before admission were ascertained by interviewing the patients with a structured questionnaire. After excluding upper gastrointestinal bleeding (226 cases) and certain bone marrow blood dyscrasias (42 cases), 286 patients with drug-induced events leading to hospital admission were identified in 2 years. Fatal adverse drug reactions, previously undescribed reactions, and some specific examples, such as digoxin-amiodarone interaction, drug-induced pancreatitis, nicardipine-induced AV block, severe skin reactions, and NSAID-induced bronchospasm, are described. Basically, this method consists of assembling series of cases systematically, and is therefore devoid of selective bias. In addition, it allows a more indepth clinical and anamnesic study of specific diseases, as compared with voluntary reporting. © 1991, Adis International Limited. All rights reserved.