Computerized physician order entry and medication errors: Finding a balance

Koppel et al s recent report on computerized physician order entry (CPOE) evaluated its role in facilitating medication errors [1]. Such studies are important because they can inform us regarding how to improve technologies like CPOE. This evaluation is also valuable because the authors studied a vendor-built product, and vendor-developed applications are the predominant type used in the U.S. today. To date, many of the studies of CPOE have been done in organizations like Wishard Memorial and Brigham and Women s Hospital, with ‘‘home-grown’’ applications [2–4]. The performance characteristics of these home-grown applications may be different from vendor systems in domains such as medication safety. This is both because the decision support has been improved over years in these institutions, and because the software has been targeted to the workflow in the individual institution, although the extent to which these represent an issue is uncertain. A main limitation of Koppel et al s study is that it did not count errors or adverse events, but instead measured only perceptions of errors, which may or may not correlate with actual error rates. Further, it did not count the errors that were prevented. As such, it offers no insight into whether the error rate was higher or lower with CPOE. Unfortunately, however, the press interpreted the study as suggesting that CPOE increases the medication error rate. While the authors did not state this, a press release put out by JAMA did so. In their paper, Koppel et al. were critical of other studies for ‘‘focusing only on its advantages.’’ In fact, other studies that we performed at Brigham and