Computerized provider order entry (CPOE) systems prevent prescribing errors by warning clinicians about medication interactions or contraindications. However, extensive research has shown that clinicians ignore many warnings, especially those perceived as clinically inconsequential. In this randomized trial, investigators created a "hard stop" warning that essentially prevented co-prescribing of warfarin and trimethoprim-sulfamethoxazole (a combination that exposes patients to severe bleeding risks). Although the hard stop was much more successful than a less stringent warning at preventing co-prescribing, the trial was stopped and the warning abandoned because several patients experienced delays in needed treatment with one of the drugs. The accompanying editorial by Dr. David Bates points out that this study vividly illustrates the unintended consequences of CPOE, a persistent issue that has slowed the pace of CPOE implementation.

Patients who received specific instructions (from physicians, nurses, or pharmacists) when prescribed the anticoagulant warfarin experienced fewer hospitalizations due to bleeding complications. The Agency for Healthcare Research and Quality (AHRQ) has published a patient information guide for warfarin therapy.

Accurate identification of medication errors poses methodological challenges. This study analyzed orders entered and discontinued within 2 hours as a trigger for inexpensive and rapid real-time evaluation. Investigators discovered that two thirds of orders discontinued within 45 minutes were viewed as inappropriate.

The author summarizes the current system of monitoring pharmaceutical safety in the United States, along with its limitations, and recommends an approach for improvement.

While computerized physician order entry (CPOE) is expected to significantly reduce medication errors, systems must be implemented thoughtfully to avoid facilitating certain types of errors. This AHRQ-funded study identified 22 situations in which the CPOE system increased the probability of medication errors. According to the study, these situations fell into two categories: information errors generated by fragmentation of data and hospitals' many information systems, and interface problems where the computer's requirements are different than the way clinical work is organized. The study looked at clinicians' experience in using one CPOE system at a major urban teaching hospital.