TABLE 3-6 Rates of Preventable ADEs in Hospitals


Preventable ADE Rate

Proportion of ADEs Preventable (No. of ADEs in study)

ADE Rate

Classen et al., 1997

1.2 per 100 admissions

About 50% (2,227)

2.4 per 100 admissions

Bates et al., 1995b

1.8 per 100 admissions

28% (247)

6.5 per 100 admissions


3.2 per 1,000 patient-days


11.5 per 1,000 patient-days

Jha et al., 1998

5.7 per 1,000 patient-days

27% (617)

21 per 1,000 patient-days

NOTE: ADE rates usually are not reported in the medical literature by categories such as renal failure, hypotension, or bleeding. On the other hand, severity levels are often quoted— for example, mild (self-limited); moderate (requiring treatment); severe (life-threatening, disabling, or markedly prolonging hospitalization) (Classen et al., 1991); or fatal, life-threatening, serious, or significant (Bates et al., 1995b).

In the study at LDS Hospital (Classen et al., 1997), ADEs were identified using computerized surveillance of medical records through the use of various automated signals (for example, drug stop orders, antidote orders) plus voluntary reporting. Among the 2,227 ADE patients, 42 percent of the ADEs arose from excessive dosage of a drug for a patient’s weight and calculated renal function, 4.6 percent from drug interactions, and 1.5 percent from known drug allergies. All these ADEs were thought to be potentially preventable, particularly through the application of computer-based programs that monitor drug use for appropriate selection and dosage.

In the first Brigham and Women’s Hospital study (Bates et al., 1995b), ADEs were identified by stimulated self-reports by nurses and pharmacists and daily review of charts by nurse investigators. Relative to the LDS Hospital study, this study reported a higher ADE incidence rate (6.5 ADEs per 100 nonobstetric admissions versus 2.4 ADEs per 100 admissions) and a lower proportion of ADEs identified as preventable (28 percent versus almost 50 percent). Among the preventable events (preventable ADEs and potential ADEs) in the first study at Brigham and Women’s Hospital (Bates et al., 1995b), 49 percent of primary errors occurred in the ordering stage, 11 percent in the transcription stage, 14 percent in the dispensing stage, and 26 percent in the administration stage. The leading types of ordering errors—wrong dose, known allergy, wrong frequency, and drug–drug interactions—were all thought to be potentially preventable by computer-

The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement