A non-interventional retrospective study in ambulatory patients was conducted at the emergency department of the Division of internal medicine. In 2 months, 266 suspected adverse drug reactions (ADRs) were identified in 224/3453 patients (6.5 %). In 158/3453 patients (4.6 %), an ADR was the reason for emergency department visit and in 49 patients (1.4 %), ADRs led to hospitalisation. A causality assessment algorithm was developed, which included Naranjo algorithm and levels of ADR recognition by the treating physician and the investigators. Using this algorithm, 63/266 ADRs (23.7 %) were classified as “certain”, whereas using solely the Naranjo score calculation, only 19/266 ADRs (7.1 %) were assessed as “probable” or “certain”, and the rest of ADRs (namely, 247/266 = 92.9 %) were assessed as “possible”. There were 116/266 (43.6 %) ADRs related to potential drug-drug interactions (DDIs), stated in at least one of the literature sources used. Based on the causality relationship, the rate of the clinically expressed DDIs was 19.0 %, or 12/63 “certain” ADR cases. Of these, 10 cases presented serious DDI-related ADRs. In summary, ADR causality assessment based exclusively on Naranjo algorithm demonstrated low sensitivity at an ambulatory emergency setting. Additional clinical judgment, including the opinion of the treating physician, proved necessary to avoid under-rating of the causality relationship, and enabled the determination of clinically expressed DDIs.