Home  |   Subscribe  |   Resources  |   Reprints  |   Writers' Guidelines

June 25, 2007

Detecting Diagnostic Errors: Have You ‘Isabeled’ It?
By Aggie Stewart
For The Record
Vol. 19 No. 13 P. 8

What types of medical mistakes occur more frequently—medication or diagnostic errors? Given all the media attention, the obvious answer is medication errors, right? Wrong. Some estimate that errors in diagnosis account for 10% to 30% of all medical errors.1 What’s more surprising is
that, given its frequency, diagnosis errors haven’t received more attention.

Researchers probing this neglected area of patient safety believe diagnostic errors are challenging to detect and dissect, in part because of the difficulty in coming to an agreement or even recognizing that an error has occurred, followed by greater difficulty determining its causes and consequences.

While this may be the case for clinicians, for patients and their families, it’s another story. Anyone who’s experienced a missed or delayed diagnosis knows something went wrong and that it had consequences—sometimes very serious consequences.2

That was the case for 3-year-old Isabel Maude and her parents when the youngster started experiencing what her doctors thought were the normal side effects of the chicken pox—fever, vomiting, and diarrhea. “They had tunnel vision and shrugged everything off as chicken pox,” says Isabel’s father, Jason Maude.

Three days later, Isabel’s symptoms had not only worsened, she had also developed swelling around her groin and the skin had turned a purplish color. Maude and his wife, Charlotte, immediately took Isabel to the emergency department (ED); this was their third trip to the ED in as many days. Ten minutes after arriving, Isabel went into multisystem failure related to toxic shock syndrome caused by a secondary infection to her chicken pox. That secondary infection was later diagnosed as necrotizing fasciitis, a bacterial infection caused by what is commonly known as flesh-eating bacteria, a variety of the bacteria that cause strep throat.

Isabel required several surgeries, first to remove the infected skin, and then to reconstruct the areas where the infected skin had been removed. Her two-month stay in the hospital was divided between the pediatric intensive care unit and a high dependency unit, where she was cared for by Joseph Britto, MD, a pediatric intensive care specialist. During that time, Isabel’s survival was never certain. If she did survive, there was concern she may suffer brain and other organ damage. Fortunately, that wasn’t the case, and she eventually made a full recovery.

Today, Isabel is a healthy, bright, and active 11-year-old. She is also the inspiration behind the Isabel System, a Web-based clinical diagnosis decision support system designed to improve the quality of diagnosis decision making at the point of care.

Conceived by Britto after he cared for Isabel through her recovery, the Isabel System is composed of two components: the Isabel Diagnosis Reminder System, a database containing more than 10,000 diagnoses, including those associated with bioterrorism, and 4,500 drugs; and the Isabel Knowledge Mobilizing System, which provides diagnosis-specific knowledge to help answer clinical questions and improve the quality of decision making. The database is continually updated and the quality of its results validated each week by a designated team of clinicians using a spectrum of clinical cases. The Isabel System is produced by Isabel Healthcare, a company cofounded by Britto and Maude.

Here’s how the Isabel System works: A clinician enters a patient’s clinical signs and symptoms, including relevant lab and test results. Using proprietary natural language processing (NLP) software, Isabel checks the data against its vast database and produces a list of possible diagnoses and/or drugs that may cause the patient’s signs and symptoms. For each listed diagnosis and drug, Isabel marshals clinical information from authoritative and widely read medical textbooks, such as **Griffith's 5 Minute Clinical Consult## and **Oxford Textbook of Medicine##, and journals, such as **The New England Journal of Medicine##. The clinical information returned also includes annotated images and treatment guidelines.

NLP recognizes and matches concepts from dissimilar terms, giving it an advantage over search engines that operate by keyword searches alone. “When you take something that clever [NLP] and apply it to the mountains of medical knowledge that exist in textbooks and on the Internet, you get a very clever diagnostic tool, which is what we’ve developed,” explains Maude.

The system is designed to augment and enhance a clinician’s pattern recognition and cognitive skills by providing a reminder checklist of likely diagnoses that should be considered. By acting as an instant reminder system, it aids the diagnosis process without interfering with the clinician’s autonomy or responsibility for determining which diagnoses to investigate and treat.

According to Maude, the Isabel System’s design can help with something called premature closure, a type of cognitive error in which a physician stops considering reasonable alternatives after formulating an initial diagnosis that appears to fit the facts. Research has shown that premature closure is the single most common cause of diagnostic errors. Maude believes premature closure was a factor in his daughter’s misdiagnosis, causing a delay that led to Isabel’s severe deterioration.3

“The outcome of any disease or condition is always going to be much worse, take longer to treat, and be far more expensive to treat when there is a delay in diagnosis,” says Maude. Creating a decision-support system that could help reduce delays due to diagnostic errors became a mission for him and Britto. The Isabel System accomplishes this not only through its content but through the Web-based technology that delivers the content. Queries to Isabel take only seconds to return results, reducing the need to cross-reference symptoms from multiple external sources.

In addition to marshalling key diagnosis-related knowledge and putting it at a physician’s fingertips, the Isabel System functions within a physician’s normal workflow. When implemented in a hospital or physician’s office that uses an electronic medical record (EMR), Isabel can be interfaced with the EMR system. The interface enables data from preassigned fields (eg, age, gender, presenting/chief complaints) to be submitted to Isabel with a mouse click. Isabel then returns a list of potential diagnoses and/or drugs for consideration—without disrupting the physician’s routine workflow. The system can also be used on a PDA with wireless Internet connectivity.

Although originally created for pediatrics, the Isabel System was expanded in 2005 to accommodate adult diagnoses. According to Maude, the system is most useful to those who see a wide range of patients and conditions, such as family practitioners who may see a newborn boy followed by a 90-year-old woman. “This incredible range means that it can be very easy for [the doctor] to miss the odd things that come through,” explains Maude. “The system is also very useful for physicians in internal medicine, starting with emergency medicine, who, again, see a broad range of conditions and ages, and those in general pediatrics. If you look at the figures on misdiagnosis, it’s in these areas where most of the problems occur.”

Several independent multicenter collaborative studies have shown Isabel’s ability to reduce diagnosis error and improve patient safety and quality of care. In fact, a November 2005 study found that Isabel provided the correct diagnosis in the list of possible diagnoses 96% of the time when key clinical features from 50 challenging Clinical Pathology Conference cases reported in The New England Journal of Medicine were entered into the system.

The Isabel System has attracted users across the globe. In the United States, it counts among its customers institutions such as Kaiser Permanente, Yale–New Haven Children's Hospital, and the Bernard Becker Medical Library at Washington University School of Medicine in St. Louis. Its EMR partners include Allscripts, NextGen, Cerner, Misys, and Eclipsys.

“The challenge in all of this was figuring out how to get the knowledge, in a useful way, into the hands of the people who first assess patients” says Maude. “Isabel does this.”

— Aggie Stewart is a freelance writer and editor specializing in HIM and HIT. She also serves as consulting editor of Health Information Management Manual, 2nd edition. She can be contacted at s-p-s@earthlink.net.


1. Schiff GD, Kim S, Abrams R, et al. “Diagnosing Diagnosis Errors: Lessons from a Multi-institutional Collaborative Project.” Advances in Patient Safety, Volume 2. Agency for Healthcare Research and Quality. 2005.

2. Schiff GD, Kim S, Abrams R, et al. “Diagnosing Diagnosis Errors: Lessons from a Multi-institutional Collaborative Project.” Advances in Patient Safety, Volume 2. Agency for Healthcare Research and Quality. 2005.

3. Graber ML, Franklin N, Gordon R. Diagnostic error in internal medicine. Arch Intern Med. 2005;165(13):1493-1499.