High rates of adverse drug events related to medication selection, dosage and monitoring were recorded at the Department of Veterans Affairs Medical Center in Salt Lake City even though the facility used a computerized physician order-entry system that automatically checked prescriptions for drug allergies and interactions, according to a study in the May 23 issue of the Archives of Internal Medicine.
The center experienced 483 clinically significant adverse drug events -- defined as events that required a change in treatment plan -- in 937 hospital admissions studied over the final 20 weeks of 2000. The number of patients experiencing at least one adverse drug event was 241 (26%) and many experienced more than one, according to the report.
Of the 483 events, 45 (9%) were classified as serious and led to prolonged hospitalization, permanent harm, death or a near-death experience. The most common drugs involved in adverse events were narcotic analgesics, diuretics and cardiovascular or renal agents, according to the report. The most common outcomes of the events were constipation, hypokalemia and hypotension.
Some 27% of the events were the result of errors, with the most frequent being failure to provide prophylaxis for expected adverse drug reactions. Nevertheless, the report's lead author, Jonathan Nebeker, M.D., an attending physician at the VA's Salt Lake City Health Care System, said the study was not about medication errors but the need to develop computerized decision support that targets the most troublesome types of adverse drug events.
"Many studies have shown that human beings can only simultaneously consider four independent factors in decisions," Nebeker said in an e-mail interview. "Although physicians think they are superhuman, they are subject to the same limitations. It is common to have to consider 10 to 20 simultaneous factors in patient medication decisions."
He said these factors include: multiple medication interactions, disease processes, patient history with medications and patient preferences.
While praising the "excellent work" that has been done on developing computer systems that reduce medication errors, Nebeker noted that these errors rarely result in patient harm. So, what is needed, he said, are systems that better model the clinical experience and present physicians with more useful information, he said.
Nebeker's report also offered this advice to healthcare organizations that are looking to buy prescribing systems: Do not expect "generic" CPOE and bar-code systems to dramatically lower adverse drug events. Instead, consider systems that offer automated recommendations for prophylaxis or monitoring strategies. For instance, when loop diuretics are prescribed, suggestions for potassium supplements and monitoring potassium levels should appear.
Read the study abstract.