prescribing error per patient per day. In this study, 19.5 percent of the errors were adjudged serious—preventable or potential ADEs.
Turning to medication administration errors, according to several international studies, administration errors (excluding wrong-time errors) are frequent, with error rates per dose ranging from 2.4 to 11.1 percent (Dean et al., 1995; Taxis et al., 1999; Barker et al., 2002; Tissot et al., 2003; Lisby et al., 2005). The U.S. study in this group found an administration error rate of 11 percent, excluding wrong-time errors (Barker et al., 2002). This study employed an observation-based method for detecting medication administration errors that has been used by the Centers for Medicare and Medicaid Services (CMS) for almost 20 years as a quality indicator for nursing homes. It was carried out in Colorado and Georgia in 36 different facilities (12 accredited hospitals, 12 nonaccredited hospitals, and 12 skilled nursing facilities). There was no significant difference in error rates (regardless of whether wrong-time errors were included) by type of facility. For the 36 facilities, the administration error rate (excluding wrong-time errors) ranged from 0 to 26 percent, with 8.3 percent as the median value. The 36 institutions studied were selected at random primarily from the Atlanta, Georgia, metropolitan statistical area and the Denver-Boulder-Greeley, Colorado, consolidated statistical area. Each facility had to agree to participate in the study. Twenty-six selected facilities declined to take part in the study. Most did not give reasons for not wishing to participate; of those that did, many expressed concerns about poor scores and wanting to improve their performance first (Barker et al., 2002). Thus the authors concluded that the error rates reported likely represent a lower bound.
A study in five intensive care units (ICUs) in U.S. tertiary teaching facilities (Calabrese et al., 2001) found an administration error rate of 3.3 percent—lower than that reported in the above study. The ICU study identified administration errors for a group of high-alert medications using a similar observational technique. The authors of this study commented that the rates they obtained were lower than those found in a comparable French ICU study (Tissot et al., 1999), and suggested that this difference might be due to varying methods of observation and pharmacist participation in patient care in the U.S. study. The committee believes these results—while the best available for large ICUs in the United States—are not generalizable to non-ICU hospital care and that the study by Barker and colleagues (2002) represents the best estimate of administration error rates in U.S. hospitals for non-ICU care.
Much higher rates of administration errors were observed in two studies that focused on intravenous medications—34 per 100 in a joint U.K./ German study (Wirtz et al., 2003) and 49 per 100 in a U.K. study (Taxis and Barber, 2003).
On the basis of the Barker et al. (2002) study and assuming a patient in