washingtonpost.com  > Technology > Special Reports > Software

Quick Quotes

Not Quite Fail-Safe

Computerizing Isn't a Panacea for Dangerous Drug Errors, Study Shows

By Sandra G. Boodman
Washington Post Staff Writer
Tuesday, March 22, 2005; Page HE01

Computerized drug ordering systems have been regarded as essential in reducing medication errors, the most prevalent and preventable kind of mistake that experts say affects an estimated 770,000 hospitalized patients annually. A review of death certificates from 1993 found that drug errors killed nearly 7,400 patients, according to the Institute of Medicine (IOM).

But a new study of a computerized physician order entry (CPOE) system manufactured by Eclipsys Corp. of Boca Raton, Fla., and used between 1997 and 2004 at the Hospital of the University of Pennsylvania, identified 22 types of persistent errors such systems are supposed to prevent.


Pharmacist Michael C. Cotungo is testing a computer program at Brigham Women's Hospital that will enable doctors to prescribe medicine on line. (David Brown - The Washington Post)

Among the potential or actual mistakes researchers found occurred weekly: incorrect doses prescribed for patients; patients who failed to get medication in a timely manner because of computer-related problems; and difficulty determining which patient was supposed to get a drug that had been prescribed.

A CPOE system requires doctors to enter medication orders into computers installed throughout a hospital, rather than scribbling them on paper or telling a nurse which drugs to order, as has been done traditionally. At Penn these electronic drug orders, which are sent directly to the hospital pharmacy, are reviewed by clinical pharmacists who check for harmful interactions and catch mistakes, Koppel said.

"We are so enamored of the technology that we are not making it responsive to the way people work," said Ross Koppel, a medical sociologist and lead author of a study, published in the March 9 issue of the Journal of the American Medical Association. Koppel said the errors his team uncovered by interviewing and shadowing 261 doctors were "stunningly frequent." Because Koppel's team did not measure errors before CPOE was installed, it is not known how often they occurred under the paper system.

During a three-month period the study found that 27 percent of doctors reported that a few times each week antibiotic administration was delayed because of system-related lapses getting such drugs re-approved, while 12 percent of physicians reported that several times each week they had trouble telling which patient they were ordering drugs for because of the poor quality of the display.

Rick Mansour, the medical information officer of Eclipsys, said in a written statement that the technology Koppel studied has been replaced by a "state-of-the-art system . . . designed to address the types of human errors described in the study."

Computerized drug ordering systems, first devised in the late 1970s, received a major boost in 1999 when the IOM endorsed them as a key way of reducing medical errors, which are estimated to kill as many as 98,000 hospitalized patients annually. Drug errors are the most common kind of medical mistake, and some studies have found that computerized systems can reduce medication errors caused by illegible handwriting and faulty transcription.

The Leapfrog Group, a consortium of large employers who have banded together to influence health care issues, made the implementation of CPOE systems one of its primary indicators of hospital quality. So far, however, less than 10 percent of the nation's hospitals have installed CPOE systems, which can cost from $2.5 million to more than $10 million to buy and operate. "I'm not opposed to CPOE, which holds extraordinary potential, but I'm opposed to dumb CPOE," Koppel said. Some errors his team reported were the result of design flaws that could have been easily corrected, Koppel said, while others reflected a fundamental mismatch between the functions they were supposed to perform and the way doctors actually work.

Drugs doses, for example, were listed according to the increments in which they are stocked by the pharmacy, not according to clinical prescribing guidelines. And if a patient was moved to another room while the computer system was down during one of its frequent crashes or for maintenance, that information was not updated in the computer, making it easier to dispense drugs to the wrong patient.


CONTINUED    1 2    Next >

© 2005 The Washington Post Company