In recounting a Harris poll, April’s column noted that 65 percent of people surveyed said they were concerned that computerization of medical records might lead to an increase in medical errors. Although few, if any, of the poll respondents had seriously researched the issue, research does back up their concern.

A widely publicized study in the March 9 Journal of the American Medical Association found that a leading computerized physician order entry system, TDS, posed 22 specific risks that could lead to medication errors. The study caused a big stir because a variety of healthcare system stakeholders has looked to computerized medication ordering systems as a way to reduce errors and cut costs.

“The literature on CPOE, with few exceptions, is enthusiastic,” wrote the authors, a multidisciplinary group at the University of Pennsylvania. “Our findings, however, reveal that CPOE systems can facilitate error risks in addition to reducing them.”

The authors used interviews, focus groups, and observations to determine how hundreds of staff members at a hospital used their CPOE system. The study looked at only one system in one hospital and documents only situations that could lead to errors, rather than documenting actual errors. Still, the authors contend that their paper tells a cautionary tale: “As CPOE systems are implemented, clinicians and hospitals must attend to errors that these systems cause, in addition to errors that they prevent.”

The 22 risky situations arose from one of two main causes: 1) fragmented or poorly integrated data, or 2) the software’s flawed interface.

An example of how fragmented data can lead to errors is when the unnamed hospital’s staff looked to the CPOE system for guidance as to the “minimally effective” dose of a medication. The dosage data in the CPOE system came from the pharmacy’s warehousing and purchasing criteria, not from clinical criteria.

“For example, if usual dosages are 20 or 30 mg, the pharmacy might stock only 10 mg doses, so 10 mg units are displayed on the screen,” according to the authors. “Consequently, some house staff order 10-mg doses as the usual or ‘minimally effective’ dose.”

An error-prone situation could arise from the flawed interface because staff may need to view as many as 20 screens to see all of a patient’s medications. As a result, “72 percent of staff reported that they often were uncertain about medications and dosages.”

Many of the risky situations the authors identified seem fixable. Staff can be trained not to rely on the CPOE system for guidance on dosing. In developing subsequent versions of the software, the vendor can consolidate patient medicine information on far fewer than 20 screens. The hard part is identifying situations that could lead to errors, rather than fixing problems once they are found. Other facilities using other software, for example, might have different error-causing situations than the 22 described in the JAMA paper.

Anyone familiar with implementing technology should not be shocked by the study’s findings. Major software implementations seldom are perfect in any industry, especially when they have a major effect on how busy human beings do their work. The authors never suggest that the software doesn’t prevent many of the errors characteristic of paper-based systems. The study merely points out that new ways of doing business often have unforeseen consequences that must be identified and addressed. Funny how the lay public had already figured that out without a big study.

The author of this monthly column, David Orenstein, is a technology and business writer in Silicon Valley. If you have a question about a Computing Care column or to request a future topic, email him at davealli@comcast.net.