Drug-error risk at hospitals tied to computers

Hospital computer systems that are widely touted as the best way to eliminate dangerous medication mix-ups can actually introduce many errors, according to the most comprehensive study of hazards of the new technology. The researchers, who shadowed doctors and nurses in a Philadelphia hospital for four months, found that some patients were put at risk of getting double doses of their medicine while others get none at all.

Doctors at the Hospital of the University of Pennsylvania identified 22 types of mistakes they have made because of difficulty using computerized drug-ordering, such as failing to stop old medications when adding new ones or forgetting that the computer automatically suspended medications after surgery. Some doctors interviewed for the study said they made computer-related mistakes several times a week.

The findings underscore the complexity of improving safety in US hospitals, where the Institute of Medicine estimates that errors of all kinds kill 44,000 to 98,000 patients a year.

The University of Pennsylvania researchers stressed that computers hold great potential, but said many systems are overhyped and hard to use, prompting one Los Angeles hospital to turn off its drug-ordering system altogether.

''We can make [hospitals] safer, but we can't do it from an imperial throne," said Ross Koppel, the sociologist who led the research, published in today's Journal of the American Medical Association. ''You've really got to get in there and find out what is happening on the floor."

Dr. David Bates, chief of general medicine at Brigham and Women's Hospital in Boston and a leading proponent of computerized drug-ordering, said the Pennsylvania study primarily applies to older software, although the system studied is one of the most widely used. Updated systems such as the one at Brigham and Women's are simpler to use and give background information to help doctors make decisions, he said. Still, Bates said, the findings are an important caution about the hope for a quick fix for the estimated 770,000 medication mistakes at US hospitals each year.

''Computerized physician order entry is still a good thing . . . but it requires a lot of work and attention and continuous updating," said Bates, who has led several studies showing that computer drug-ordering can reduce medication mistakes by 55 percent or more.

The computer debate speaks to a larger issue in the patient safety movement: figuring out what makes patients safer. The emphasis on reducing medical mistakes is only about a decade old, and the science behind many promising ideas remains sketchy, or the studies were done by researchers biased by the fact that they made the safety improvements.

Some ideas turn out to be simply impractical. For example, Dr. Kaveh Shojania, a Canadian patient safety researcher at the Ottawa Hospital-Civic Campus, described how a program to prevent hip fractures among elderly hospital patients fell apart because patients didn't want to wear special pads all the time. ''Not surprisingly, the elderly, on top of feeling awful to be in the hospital, don't want to wear hockey pads while they're lying in bed," he said at a Washington, D.C., conference on medical errors in December.

As for computerized drug-ordering, Shojania said, the evidence that it saves lives ''is really quite weak," and based on studies that don't look at new types of mistakes or that focus on the errors prevented rather than number of patients saved. ''I have been calling for a little bit more restraint in going out and recommending [new safety technologies] so quickly" before the evidence is in, he said in an interview last week.

Nevertheless, hospitals are under mounting pressure to replace paper-based drug-ordering systems with computers, eliminating the confusion caused by physicians' sloppy handwriting, and greatly improving the ability to track patients' medication history. The software requires doctors to be very specific about the drug and dose they want to prescribe, then automatically warns them if it might cause an allergic reaction or interact with other medication. The Leapfrog Group, a powerful coalition of 160 companies and organizations that buy healthcare, has made use of computerized drug-ordering one of its main measures of hospital quality.

Only about 10 percent of US hospitals have fully installed computerized drug-ordering, including just four of the 65 acute care hospitals in Massachusetts. Partly, that's a reflection of costs -- the systems can cost several million dollars to install -- but it's also a measure of doctors' wariness of switching from the familiar prescription pad.

Doctors successfully pressured Cedars-Sinai Medical Center in Los Angeles to turn off the system two years ago. They complained that computers slowed them down and sometimes lost their orders. Hoping to avoid similar problems, Beverly Hospital in Massachusetts set up a ''war room" late last year where doctors in the maternity ward could test the drug-ordering system before it was activated.

Sociologist Koppel said he discovered the drawbacks of computerized drug-ordering by accident while hearing from young doctors about the stresses in working long hours at a 750-bed Philadelphia teaching hospital. ''The thing I thought would be the least stressful, this computer that is supposed to be the solution to all medication errors, kept on emerging as both a cause of stress and a cause of error," he said.

Koppel said some of the mistakes occurred because doctors didn't fully understand the system. For example, residents would sometimes confuse inventories of the drug dosages available in storage with recommended doses. And 83 percent of the junior doctors in the study said they knew of patients unintentionally cut off from antibiotics because doctors didn't renew the prescription every three days as the computer required.

Other errors reflected design flaws in the 1997 vintage system, which is among the most widely used in US hospitals, Koppel said. For example, patients with complex problems would require up to 20 computer screens to display their medication history, making it difficult for doctors to avoid duplication or drug interactions.

All 22 sources of errors identified in the study, Koppel said, were affecting patients, potentially either slowing their recoveries or harming them, although he had not heard of any deaths because of the mistakes. However, he said he was told that one organ transplant patient at another hospital using a similar computer system failed to get his antirejection drugs for six days because of a data entry mistake. That patient did survive, said Koppel.

Bates, of Brigham and Women's Hospital, who served as an adviser for the study, said that Koppel's research didn't look at the error rates before the computer system was installed, making it impossible to know whether the overall error rate rose or fell. By contrast, Bates's research at Brigham and Women's found substantial drops in medication errors when computers were introduced.

Bates also noted that the University of Pennsylvania bought a new computerized drug-ordering system in 2004 to replace the one Koppel studied. The new system fixes many flaws in the earlier version, he added.

Scott Allen can be reached at allen@globe.com.