bar coding—especially when linked to an electronic medication administration record—hold substantial promise, although the evidence for their efficacy is less strong than is the case for CPOE (AHRQ, 2005). Bar coding will likely be especially important for medications taken orally, although it will probably also be important for intravenous medications. The latter are particularly risky because they are especially potent, have rapid action, and are given to critically ill patients, and because it is easy to give an intravenous dose that exceeds the norm by 10 (or more) times. Another technology that appears likely to have an impact on improving the safety of intravenous medications is “smart” pumps. These pumps can be programmed according to the medication being given, warn the nurse if the dose is too high, and record what happens if the dose is overridden. The first large trial of this technology demonstrated that it could be used to identify many instances in which doses were too high, but it did not reduce the rate of serious medication errors because nurses often ignored even important warnings (Rothschild, 2004).

Although dispensing appears to be relatively safe compared with other steps in the medication-use process, it, too, can be made safer by bar coding (Poon et al., 2005). The use of robots for filling prescriptions in pharmacies may also have the potential to improve dispensing (Bates, 2000). Automated dispensing devices on clinical units may improve safety as well, although the one study evaluating their impact failed to demonstrate any benefit (Barker, 1995; Barker et al., 1998).

Monitoring should also benefit from computerization in two ways. One is that the standard approach to identifying ADEs—spontaneous reporting on paper—should be replaced by on-line reporting, which has many advantages (Bates, 2002). Specifically, on-line reporting makes it possible to use branching logic; makes it easier to update the reporting scheme; and facilitates collection of the data in standard formats, which in turn facilitates analysis. Much more important, though, is implementation of computerized ADE monitoring (Classen et al., 1991) that uses signals to identify situations in which an ADE has occurred or is likely to occur. In many instances, it may be possible for someone—usually a pharmacist—to intervene either before the event occurs or before it becomes as severe as it might otherwise have been.

There have been a few studies of override problems related to technologies designed to reduce medication errors. As noted above, a smart pump can fail to reduce serious medication error rates because nurses ignore important warnings (Rothschild, 2004). At the prescribing stage, Hsieh and colleagues found that overrides of allergy warnings were common, and 1 in 20 of these overrides, while appearing to be clinically appropriate, resulted in an ADE (Hsieh et al., 2004). At the pharmacist order entry stage, Grisso