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.

Our medication error rate was similar to that of previous pediatric ICU studies that used the direct observation method for reporting but higher than the rates in previous studies using other detection techniques such as voluntary incident reporting. Periodic direct observation and other ongoing data collection methods such as voluntary incident reporting have the potential to be complementary approaches to medication error and ADE detection.