Quality Rounds
Cyber Safety Net
Can web-based systems help prevent misdiagnoses?
By Scott D. Smith
In the past, the smartest doctors were the ones who could recite from memory a long list of symptoms and the diseases associated with them. With the body of medical knowledge already too vast for any one mind to contain, some physicians are looking for help from web-based diagnostic systems.
One physician who is taking a chance on such a system is Phillip Kibort, M.D., chief medical officer of Children’s Hospitals and Clinics of Minnesota. Last year, Children’s purchased the relatively new Isabel system.
Kibort says he decided to try the web-based system in part because of a recent case at Children’s in which a child died. Although he can’t share details, he says the treating physicians got caught up in believing the child’s diagnosis was one thing when it turned out to be something else.
The case illustrates the significant harm diagnostic errors can cause, and recent research has attempted to assess the magnitude of the problem. In the article “Overconfidence as a Cause of Diagnostic Error in Medicine,” which appeared in the American Journal of Medicine in May, authors Eta Berner, Ed.D., and Mark Graber, M.D., report rates of diagnostic error ranging from less than 5 percent in the perceptual specialties (pathology, radiology, dermatology) to 10 percent to 15 percent in most other areas. In the Harvard Medical Practice Study published in 1991 in the New England Journal of Medicine, which involved a review of 30,195 hospital records, diagnostic errors accounted for 17 percent of adverse events. In a follow-up study published in Medical Care in 2000, a review of 15,000 records from a random sample of hospitals in Colorado and Utah found that diagnostic errors contributed to about 7 percent of adverse events.
Diagnostic errors have many causes. Some, such as lab specimen mix-ups or lost test results, are largely outside a physician’s control. But according to Berner and Graber, sometimes misdiagnoses occur because physicians fall into cognitive traps such as confirmation bias, where they only seek confirming evidence for a diagnosis, or premature closure, where they truncate the diagnostic process and settle on a diagnosis too quickly.
“One of the most dangerous things we can do to a patient is to assign a diagnosis because we stop thinking,” says Emily Chapman, M.D., graduate medical education director for Children’s. “You must consider the broad differential diagnosis and not be wedded to just one.”
Kibort hopes Isabel will help physicians, residents, and medical students do just that by serving as a cognitive aid, a memory jogger, and a reference.
Computer-Aided Diagnosis
The idea of the computer-empowered diagnostician is not a new one—think Bones in the original Star Trek. But widespread use of such systems has never really caught on. Commercial products such as Meditel, Quick Medical Reference, and Iliad have come and gone. Two others, Problem Knowledge Couplers and DXplain, have been around for more than 10 years. However, physicians already confident in their diagnoses have been reluctant to use such systems because of a perception that entering the data and filtering through the results is an inefficient way to spend their time, according to Berner and Graber.
Isabel is a relative newcomer to the market. A pediatric version designed by British pediatrician Joseph Britto, M.D., was first launched in 2002. Now, there is also an adult version, and about 20 medical centers in the United States are using the product. Children’s and Luther Midlefort in Eau Claire, Wisconsin, which is part of Mayo Health System, are two of them. The annual cost for using the system is $180 per hospital bed. The system is available to medical staff at Children’s Minneapolis and St. Paul facilities.
Physicians log on to Isabel through a portal in Children’s electronic health record system. They type in a patient’s symptoms, vital statistics, lab results, and any other relevant information. The system then searches a database of 11,000 diagnoses to arrive at up to 30, highlighting the 10 most relevant ones.
In a demonstration, Chapman types in the symptoms fever, listlessness, rash, and hepatitis. She notes the patient’s age and gender. Isabel then produces a list of diagnoses Chapman might consider, ranging from herpes simplex virus (the one Chapman was actually thinking of) to lupus to the obscure Q fever, a zoonotic disease caused by Coxiella burnetii, with the most common diagnoses listed first.
Chapman can click on a disease name to learn more about it. She also can access UptoDate, a popular peer-reviewed information resource, and journal articles through the same portal.
Encouraging Broad Thinking
The system promises to rid doctors’ offices of shelves full of medical reference books. “In the old days, I would have written down every disease I could think a case could possibly be, and would have gone to my books and started looking things up. Today, I still need the initial list of diseases and the information about those diseases. This just gets it faster,” Kibort says.
Chapman uses Isabel when the house staff reviews cases. It helps medical students generate a broad differential of possible conditions for a set of symptoms, and it helps them to research that differential and identify a likely suspect.
In fact, according to Chapman, it has changed the student-teacher dynamic. Students now research cases in advance and ask her if the patient could have Q fever rather than look to her for a list of possible causes.
But do experienced physicians welcome the system? And, more important, has it prevented any tragedies?
Kibort says he isn’t sure whether Isabel has prevented any diagnostic errors at Children’s. But he does know that the system has been accessed about 100 times a month on average (including Chapman’s case reviews with house staff) since it was implemented and that practicing physicians turn to it when they encounter a case that puzzles them.
Although Kibort would like physicians to use the system more, he acknowledges a possible downside—it may cause them to spend too much time considering unlikely diagnoses and to order unnecessary tests and procedures.
So what is the answer to reducing the rate of misdiagnosis? In a commentary accompanying the research article in the American Journal of Medicine, Graber points to improved training and education about cognitive pitfalls, creating systems that double-check the delivery of test results to providers and patients, and providing point-of-care access to reference texts and journals and access to second opinions and experts. He cites web-based diagnostic systems as being only one mechanism for helping physicians avoid making the wrong diagnosis.
Says Kibort of his decision to purchase Isabel, “If it helps one or two patients a year, then I think it’s worth it.” MM
Scott Smith is an MMA staff writer.