Channel Insider content and product recommendations are editorially independent. We may make money when you click on links to our partners. Learn More.

Computerized systems that reduce certain medication errors increase the risk of others, concludes a study published Wednesday in the Journal of the American Medical Association.

CPOE (computerized physician order entry) is widely hailed as an important solution for reducing medical errors. However, the study, led by Ross Koppel at the University of Pennsylvania, listed over twenty ways that CPOE made medical errors more likely to happen. In particular, medicines could be ordered for the wrong patient, sent to the wrong place, or delayed for more than 24 hours.

Koppel’s original intention was not to study CPOE, but the sources of medical prescribing errors made by young physicians in hospitals. He told he was surprised when the doctors kept bringing up the CPOE system, until he looked at it himself.

“I shadowed people, and looked at the system, and interviewed them, and I understood what they meant. It was a clunky, clumsy system that could easily facilitate errors,” Koppel said.

The problem is that vendors and administrators expect doctors to shape their practices to the software rather than the other way around, Koppel said. “[When] software doesn’t work the way information flows in the hospital, they want doctors and hospital processes to wrap themselves like pretzels around the software.”

A study shows that health care providers with access to EHRs (electronic health records) are more likely to have complete patient information. Click here to read more.

Based on a survey of 88 percent of clinicians using the system, Koppel’s study found several ways CPOE could encourage errors, including several listed below.

  • If a physician enters orders after midnight, say after evening rounds, medication intended for the coming morning will not arrive until the next morning.
  • Names of patients are listed alphabetically in small font, making it easy to select the wrong patient.
  • Clinicians are unsure of patient’s medications because all medications cannot be shown on a single screen.
  • Some reminders are kept in paper charts and are often ignored, leading to gaps in medication.
  • Drug dosage levels displayed by the CPOE do not reflect guidelines, but pharmacy inventory.
  • Medications suspended for surgery must be re-entered and activated individually. Though nurses can change a patient’s status to “active,” this will not affect medications.
  • CPOE crashes and shut-downs delay orders and, if patients are moved to a different room when the system is out of service, cause medications to be sent to the wrong room.
  • When medical procedures are ordered, cancelled, or modified, all accompanying medicines must be stopped, re-ordered, or modified separately. Also, ordering non-standard drugs requires a separate process that may be ignored, causing delays.

Though clinicians are ultimately responsible for clinical errors, Koppel said, clinical software must make it harder, not easier to make mistakes.

He could not recall any improvements made to the system during the time of his study, from 2002 to 2004. “I saw doctors trying to work around the software and getting very frustrated that it hadn’t been adjusted to make it responsive to their needs.”

Nonetheless, Koppel said he was in favor of CPOE systems. “There are some dumb programs and really poor integrations, but on balance they reduce hand-writing errors, speed up orders and are instantly available from anywhere.”

Next Page: Koppel’s study reflects an increasing awareness of IT issues in the health care sector.

Though CPOE has been endorsed by high-profile groups like the Leapfrog Group, only about one in twenty U.S. hospitals have a system in place. Of those, fewer than half of physicians use the system for the majority of medication orders, according to information posted on the TEPR 2005 (Toward an Electronic Patient Record) Web site.

Koppel said that vendors needed to be willing to investigate how their products were working in hospitals and to make “constant tweaks” to adjust the software to the clinicians, rather than the other way around. “Installing the software and teaching people how to use it is only the first of many, many steps.”

Koppel’s study looked at an older CPOE system (Eclipsys’s TDS) in one hospital, the University of Pennsylvania Health System. Rick Mansour, Eclipsys’s Chief Medical Information Officer, said the company had already corrected flaws pointed out by the study in their latest product.

“With the evolution of technology since the first-generation CPOE-based systems were developed, Eclipsys has incorporated numerous features into the design and build of its Sunrise Clinical Manager that dramatically reduce the potential for human errors.”

Other vendors contacted by were not immediately available for comment or said they could not comment because their system had not been evaluated.

Two clinical informaticians contracted by said the problems cited in Koppel’s study seemed universal.

To read Monya Baker’s opinion about the promises and dangers of health IT, click here.

“Physicians and nurses have been concerned about the new [kinds of] mistakes for a long time, and it’s great to see someone documenting on that,” said Russ Cucina, a practicing physician and professor at University of California, San Francisco.

Joan Ash, a professor of medical informatics and clinical epidemiology at Oregon Health and Science University, has studied unintended consequences of CPOE in three different countries, and told her research has come to many of the same conclusions.

She said part of the problem was lack of awareness by those that pick the systems. Hospitals tend to get information from vendors and to hire consultants more inclined to study technical specifications than how a technology would fit into an organization.

The study reflected an increasing and welcome consideration of how new tools work within health care organizations, she said. “I don’t think anyone should be depressed by these results. I think it’s a wake up call, and a really good thing.”

Cucina said that user interfaces for clinical systems have lagged far behind those of other commercial applications partly because the clinicians using the system were not the ones purchasing the systems, but also because of the hectic, unpredictable pace of a clinician’s work.

“The software is written from an engineering mentality where work is linear and uninterrrupted,” Cucina said, “but clinicians don’t have linear, uninterrupted time.”

Check out’s for the latest news, views and analysis of technology’s impact on health care.