Panacea or Pandora’s box

Penn study shows that computerized physician-order entry systems often facilitate medication errors

Health-care policymakers and administrators have championed specialty-designed software systems – including the highly-touted Computerized Physician Order Entry (CPOE) systems – as the cornerstone of improved patient safety. CPOE systems are claimed to significantly reduce medication-prescribing errors. “Our data indicate that that is often a false hope,” says sociologist Ross Koppel, PhD, of the Center for Clinical Epidemiology and Biostatistics at the University of Pennsylvania School of Medicine. “Good computerized physician order entry systems are, indeed, very helpful and hold great promise; but, as currently configured, there are at least two dozen ways in which CPOE systems significantly, frequently, and commonly facilitate errors – and some of those errors can be deadly.”

As reported in today’s Journal of the American Medical Association, Koppel and colleagues studied the day-to-day medication-ordering patterns and interactions of housestaff working in a tertiary-care teaching hospital, which, at that time, ran a popular CPOE system. In addition to a comprehensive survey of almost 90% of the housestaff who use CPOE, the researchers also shadowed the doctors and pharmacists, as well as performed interviews with the hospital’s attending physicians, nurses, IT and pharmacy leaders, and administrators. As a result, they identified 22 discreet ways in which medication-errors were facilitated by the CPOE system they studied.

The significance of their findings, notes Koppel, is to serve as a wake-up call to those who would believe that hospital IT systems — such as computerized physician order entry systems — represent a simple turn-key solution to patient safety; and, in particular, the reduction of medication errors. “Although we analyzed only one older CPOE system in a single setting, our findings reflect what is happening in health-care facilities across America that have adopted CPOE systems as a key patient-safety initiative,” said Koppel. “We show that CPOE systems need to be very carefully designed and implemented, as well as constantly evaluated and improved. Further, as these systems continue to be improved, designers should understand that their programs must seamlessly integrate into an institutional context of infinite complexity … one that operates 24/7, under great stress, and with a constantly-changing set of people, policies, and practices.”

“As vigorously as the nation’s administration pushes for IT solutions to reduce medication errors, so, too, must they push for research support in that area – so that IT systems can be constantly tested, evaluated, and modified, as necessary,” adds co-investigator Brian L. Strom, MD, MPH, Professor of Medicine at Penn and Chair of its Department of Biostatistics and Epidemiology.

Two Groups of Errors

Introduced approximately 10 to 15 years ago, computerized physician order entry systems were designed to transform paper-based prescriptions into computerized orders sent directly the hospital’s pharmacy. Since then, published studies have credited CPOE systems with reducing medication errors by as much as 81%, notes Koppel, principal investigator of this landmark study. However, while illegible handwriting may have been resolved satisfactorily by CPOE systems, other risks of medical-errors are accentuated.

After identifying 22 ways in which medication errors were facilitated by the CPOE system analyzed, Koppel and his research team grouped error types into two main categories: information errors; and human-machine interface flaws. Information errors, explains Koppel, result from fragmentation of data and information, or when there is a failure to fully integrate a hospital’s multiple computer and information systems. Examples of these errors are when a physician orders the wrong dose of a drug because the CPOE system displays pharmacy warehouse information that is misinterpreted by the physician as clinical-dosage guidelines or when warnings about antibiotics are placed in the paper chart and not seen by physicians who are using only the computerized system. Human-machine interface flaws reflect machine rules that do not correspond to work organization or usual behaviors. For example, within the CPOE system studied, up to 20 screens might be needed to view the totality of just one patient’s medications – thereby increasing the risk of selecting a wrong medication. “To be effective, a CPOE system must articulate well with the work-flow within the organization,” emphasizes Koppel.

“We seem to think that we can just wrap people and organizations around the new technology, rather than make the technology responsive to the way clinicians and hospitals actually work,” adds Koppel, who also teaches in Penn’s Sociology Department.

Recommendations

As CPOE systems continue to be implemented and enhanced, Koppel advises institutions and governments to diligently consider the errors caused by such systems as much as the errors prevented. Indeed, he and his colleagues suggest, among other things, that IT-assistance programs focus primarily on the organization of work in an institution, rather than on the technology itself. “Computers do some things brilliantly, and people do many things brilliantly – but substitution of technology for people is a misunderstanding of both,” he says. Indeed, as the 1957 Spencer Tracy / Kathryn Hepburn comedy Desk Set illustrated so well, a blind faith in technology is always misplaced.

Koppel and his colleagues also call for an aggressive examination of the technology in use: in other words, hospitals should perform an in-depth review and analysis of the way technology is actually used by physicians and nurses, rather than on how manufacturers expect the technology to be used. In addition, the researchers recommend that continuous revisions and quality improvement be part of all medical IT programs.