substantially influence the estimate of error rates. Flynn and colleagues (2002) collected parallel data on 2,557 doses using direct observation, chart review, and voluntary incident reporting. Rates of error detected by the three methods across all study sites (including 24 hospitals and 12 nursing homes) were 17.8 percent, 0.1 percent, and 0.003 percent, respectively.
Three studies investigated error rates by stage of the medication-use process. Using a cohort design involving chart review and stimulated reporting, Gurwitz and colleagues (2000) detected most errors in the prescribing and monitoring stages. Among the 464 preventable ADEs identified in the study, errors occurred most often in the prescribing stage (315 errors, 68 percent of ADEs) and the monitoring stage (325 errors, 70 percent of ADEs). Errors were rare in the documentation/transcription, dispensing, and administration stages.
Similar results were found in a later study by the same research team using similar chart review methods (Gurwitz et al., 2005). Errors associated with the 338 preventable ADEs were more likely to occur at the prescribing (59 percent of ADEs) and monitoring (80 percent of ADEs) stages. Errors were less common at the dispensing and administration stages. Forty-six percent of preventable ADEs involved errors at two stages of the medication-use process, and 5 percent involved errors at three stages.
Handler and colleagues (2004) analyzed incident reports at one long-term care facility; they found an average of 4.7 reports per month, while residents averaged 11.2 medications per day. A process analysis indicated that the same stages of medication use occur in the nursing home and hospital settings. Consistent with hospital reporting, most incident reports in this study were filed by nurses; 68 percent of reported errors occurred at the administration stage, 20.4 percent at the dispensing stage, and 11.6 percent at both the administration and dispensing stages.
Gurwitz and colleagues (1994) published a retrospective review of incident reports from one 703-bed academic nursing home for 1 year to identify adverse and unexpected events. After falls, medication-related events (n = 180) were most common, at 26 per 100 beds. Errors in dosing (72.2 percent of reports) were more common than adverse drug reactions (26.7 percent). A more recent study based on incident reports during 21 months at a single 126-bed long-term care facility identified 98 errors, but no denominator was used to compute error rates (Handler et al., 2004). Authors of both of these studies acknowledged that the findings under-