Each year, more than 770,000 hospitalized patients die from an adverse drug event. Adverse drug events are most often caused by prescribing errors—even though there are many checks in place to catch such errors as the prescriptions move between the doctors who write them, the pharmacists who fill them, and the nurses who deliver them to the patients. Prescribing errors include things such as prescribing the wrong medication; prescribing the wrong dose of the right medication; or prescribing a drug that interacts dangerously with another drug the patient is already taking.

In the past several years, some hospitals have implemented computerized physician order entry (CPOE) systems in an effort to reduce prescribing errors. They introduced many new checks into the system such as red-lighting dangerous drug-drug interactions and eliminating the need to decipher physicians’ handwriting. The early reports of CPOE have been very positive, suggesting that they reducing prescribing errors by up to 81%.

But in the March 9, 2005 Journal of the American Medical Association , researchers report that computerized prescription systems might make certain types of medication errors more likely. They found that CPOEs facilitate 22 specific types of medication errors, and that many of these types of errors occur frequently in the hospital setting.

About the Study

The researchers studied a 750-bed teaching hospital that had implemented a widely used CPOE system.

They surveyed 261 house staff members (interns, residents, and fellows), who regularly use the CPOE system, about working conditions and potential sources of error and stress. They then conducted focus groups and held one-on-one interviews with house staff to learn about sources of stress and actual prescribing errors. The researchers also conducted interviews with the chairperson of the surgery department, the pharmacy and technology directors, five nurses, and five attending physicians. Finally, for a four-month period, they shadowed four house staff members, three attending physicians, and nine nurses during patient care and CPOE use, and observed three pharmacists reviewing orders.

The Findings

The researchers found that the CPOE system they studied facilitated 22 types of medication errors. They categorized these errors as either information errors or human-machine interface flaws.

Information Errors

Information errors occurred when data was scattered or because all of the hospital’s computer and information systems were not integrated.

For example, house staff often used the CPOE system to determine the minimally effective dose of a medication. If the minimally effective dose for a medication was 30 milligrams (mg), but the pharmacy stocked the medication in 10 mg doses, then only 10 mg doses were displayed on the screen. As a result, the staff may have assumed that 10 mg was the minimally effective dose. Seventy-three percent of the house staff reported using the CPOE system to establish minimally effective doses for medications they were not familiar with.

Another example of an information error was the failure to renew antibiotics. At the hospital studied, antibiotics were generally prescribed for three days, and they had to be renewed before the third day for continued use. However, reapproval stickers were placed on paper charts on the second day, but not in the CPOE system. Over a three-month period, 83% of the house staff observed gaps in antibiotic therapy due to unintentional delays in reapproval.

Human-Machine Interface Flaws

Human-machine interface errors occurred when the computer system did not interface effectively with the people using it.

For example, on the CPOE system, patient names appeared in small font and did not appear on all of the screens. Fifty-five percent of house staff reported that they had a difficult time identifying which patient they were ordering medication for because of the way in which this information was displayed.

In addition, 84% of house staff reported that medication orders were delayed because of CPOE system shutdowns. Forty-seven percent said that these shutdowns occurred anywhere from several times a week to more than once a day.

This study was limited because even though the researchers discussed actual prescribing errors during the focus groups and interviews, the surveys dealt with the likelihood of medication errors, not actual adverse drug events. As a result, the survey analysis reflected potential sources of errors rather than actual errors. In addition, the researchers looked only at one hospital’s experience with its CPOE system, so their results cannot necessarily be applied to other such systems.

How Does This Affect You?

This study identified 22 situations in which a CPOE system increased the risk of prescribing errors. These errors were observed by a majority of the house staff in one hospital and tended to occur frequently.

This study was designed to examine the errors facilitated by a CPOE system (not the errors that were actually caused or prevented by it), and should be interpreted in this context. Many of the problems revealed by this study have easy solutions, and the potential value of the CPOE concept can be preserved if these flaws are fixed.

Computerized order entry systems have been shown to prevent many types of prescribing errors. These systems eliminate the need to decipher messy handwriting, they can link directly to drug-drug interaction warnings, and they can be easily integrated into computerized medical records. Like any other systems—technological or otherwise—improvement is always possible if flaws are identified and corrected. Hopefully, this study will be part of that inevitable process.