Fatal ErrorsBy Larry Barrett | Posted 2003-02-10 Email Print
Eighty-five hundred people at St. Mary's Mercy thought they were still alive. But the hospital's computers were telling them they were not.
Last February, a 6-day-old boy died at Stony Brook University Hospital in Stony Brook, N.Y., after a nurse injected the child with a lethal dose of potassium chloride. The boy's parents claim the prescription for the drug read 35 milligrams instead of 3.5 milligrams, and that neither the nurse practitioner nor the nurse who administered the drug recognized that it was a lethal dose for a child that size.
Stony Brook University Hospital is now installing an $18 million computer system that will include a state-of-the-art computerized pharmacy system and physician order-entry system that—if properly installed and maintained—experts say can dramatically cut the risk of future medication errors. Analysts say less than 5% of U.S. hospitals have computerized pharmacy systems that interact with physician-entry systems.
"Many health care organizations are under enormous financial pressure at the same time that scrutiny on medical errors is increasing," Klein says. "The survival of many health care organizations will depend on their ability to harness available I.T. to the cost-effective delivery of health care services."
While none of the patients affected at St. Mary's Mercy died or were subjected to dangerous or unnecessary treatments, experts familiar with upgrading patient-management systems say blame ultimately must fall on the shoulders of the hospital and its I.T. staff.
"When you're doing a conversion or an update of a system like this, you have to absolutely get down to the tiniest details to make sure the data gets transferred accurately," says Linda Reino, CIO of Universal Health Services, a $2.6 billion hospital-management group based in King of Prussia, Pa.
Reino, who has supervised hundreds of patient-management system upgrades and conversions, says the remapping issues that plagued St. Mary's Mercy were likely a product of inadequate testing and attention to the steps of transferring data.
"For example, if a field like a patient's first name is going to be mapped to the new version as name/first rather than first/name in the old system, you have to make sure that every one of those thousands of data fields match up before you go live," she says. "Then you have to consider the data-valuing issue."
In programming a line of code, software developers establish a format and place where a number—the value of a piece of data—can be inserted and manipulated. In new software, programmers must make sure the descriptions for data refer to the same information as before.
Reino says one of the most common mistakes hospitals and other health care organizations make during upgrade and conversion cycles is incomplete data-value testing. If an old system used "D" to denote a patient's death and the new system uses "E" to stand for expired, the I.T. staff has to make sure every reference to "D" in the old system is recoded to equal "E" in the new system. "A lot of people don't understand the degree of specificity required to make sure something like this doesn't happen," she says. "And it gets worse when other departments start accessing the system. That's when you get the error multiplying itself throughout the system."
She says McKesson Corp.'s CareEnhance, Siemens Medical's Invision, Medical Information Technology Inc.'s Medtrack and GE Medical Systems' GE Centricity programs are among the most widely used patient-management systems in the industry. Which product St. Mary's Mercy was using, though, "doesn't really matter," Reino says. "If you're not incredibly diligent throughout the entire process, you will have big problems no matter which vendor you use."
The Institute of Medicine study estimated the cost of nonfatal medical errors is between $17 billion and $19 billion each year, and that between 2.9% and 3.7% of all patients admitted suffer some type of injury as a result of medical mismanagement.
"This situation came down to the customer not finding all the places where the values were changing," Klein says. "It's not negligence, but better testing by the customer would have prevented something as embarrassing as filing insurance claims for people who were already identified as dead."