Using a direct observation approach, we found a higher incidence of potential and actual, preventable adverse drug events and an increased ratio of potential to actual, preventable adverse drug events compared with studies that used chart reviews and solicited incident reporting. All of the potential adverse drug events and approximately two thirds of the actual adverse drug events were judged to be preventable. There was one preventable error for every five doses of medication administered; most errors were due to dose omission, wrong dose, wrong drug, wrong technique, or interactions.

Among the interventions performed and documented by a clinical pharmacist in an ICU, patient care rounds and chart-review activities were associated with the greatest number of interventions and the greatest potential cost avoidance.