The most award winning
healthcare information source.
TRUSTED FOR FOUR DECADES.
Using better analytics and training staff in specific improvement strategies can reduce diagnostic errors. Physicians often need to broaden their differential diagnoses.
Reducing diagnostic errors requires a combination of strategies that address the reason most of these errors occur and the application of the latest data analytics.
When analyzed retrospectively, most diagnosis failures will be traced to thought errors, says David Kashmer, MD, MBA, MBB, FACS, a trauma and acute care surgeon, and associate director of The Surgical Lab, a quality improvement-focused healthcare group in Center Valley, PA.
“They’re not usually errors with our hands or test results being wrong, though those things happen,” Kashmer says. “They’re errors of how we think about a situation. It’s often failure to have a broad enough differential diagnosis when patients present to the emergency department or an acute situation of any sort. We didn’t think of all the possibilities it could be, didn’t take the time, or weren’t sensitive enough to it.”
For instance, 20% of patients in an ICU typically present with adrenal insufficiency, but it often is not diagnosed because physicians just don’t think about it, Kashmer says. When teaching young surgeons, Kashmer encourages them to think broadly about diagnoses with the mnemonic “VINDICATE.” It reminds them of all the diagnoses they should entertain:
Kashmer encourages the physicians to take the time to consider each possibility and prioritize what is most common and which one you have the least time to catch. They also need to think in terms of, “how can I be less wrong?”
“I’ve found that gets it on their radar, including things you might not think about as often, like adrenal insufficiency and right-sided heart attack, which is often missed in the ICU,” Kashmer says. “I also teach to think about how they would manage uncertainty. How likely does it need to be that someone has something before you treat for it? That’s a very different way to think about diagnoses than we’re taught in medical school, where either they have it or they don’t. It’s very black and white.”
Kashmer advocates for more use of “decision science,” which is more often seen in the business world. This involves better decision-making with decision trees for things like pulmonary emboli, which will help determine how likely the diagnosis needs to be before the risk of heparin will equal the benefit of heparin.
The strategic use of big data also can help reduce diagnostic errors, Kashmer says. Big data involve extremely large data sets that are analyzed to reveal patterns, trends, and associations.
Big data can help eliminate or reduce diagnostic uncertainty or determine which tests to run, for instance.
“That can include decision trees and other tools that might tell you not to bother running this test because it’s not going to change your mind about whether they have a certain disease because of how accurate the test can be,” Kashmer says. “That’s really powerful and comes to us only from decision science. Those things are typically not taught in medical schools.”
A branch diagram, for instance, can even be used to predict the outcomes of certain decisions, including the potential financial cost for the hospital. The diagram might show that if you make this particular decision and it’s wrong, the patient ends up losing five years of life.
“You can do the same thing with money,” he says. “You can say the typical payout in our state for this situation is a million dollars if it goes to trial, and if settled, this is the typical amount. You can almost do triage and say if you miss inflammatory breast cancer, that’s a big one and will pay out this much. Missed fracture in a trauma patient, maybe not as big a deal and a lower payout. So you can develop an understanding that you really can’t afford to miss inflammatory breast cancer, so you adjust your decision-making in light of that.”
The time is right to apply big data for improving diagnoses, says Mark Wolff, PhD, chief health analytics strategist with SAS, a data analytics firm based in Cary, NC.
Big data analysis is increasingly useful because so much data are now available in a digitized form, making it possible to analyze faster and more thoroughly, he explains. Technological improvements also make it possible now to analyze massive amounts of information, Wolff says.
“We can now begin improving the process of population health analytics, which then allows us to better diagnose individuals and to actually make a prediction as to what will or won’t work, what the outcomes might be, what drugs will work best and in what way, and what adverse events might be encountered,” he says.
With big data, information on hundreds of millions of patients can be analyzed to look for patterns in symptoms and diagnosis, for instance.
“Technology will provide data to support a clinical diagnosis, and we have very good technology to analyze a disease state. But one of the challenges are limitations of human beings,” he says. “Cancer researchers have said that it is unethical for human beings to diagnose cancer without the aid of technology, because we have approached the limit of human cognitive capabilities in terms of understanding the amount of information available and what is relevant in diagnosing and treating disease.”
The point, Wolff says, is that the availability of data only makes it possible to improve diagnostic accuracy. For those improvements to actually come to fruition, the problem of information overload must be addressed with computational technology.
“Reducing errors relies on technology, but we still have people who don’t trust technology,” Wolff says. “Physicians sometimes don’t want computers making diagnoses for their patients, but we’re at a point where the technology can do that and save the complex cases for a physician’s judgment.” n
Financial Disclosure: Author Greg Freeman, Editor Jill Drachenberg, Assistant Editor Jonathan Springston, and Nurse Planner Maureen
Archambault report no consultant, stockholder, speaker’s bureau, research, or other financial relationships with companies having ties to this
field of study. Physician Editor Arnold Mackles, MD, MBA, LHRM, discloses that he is an author and advisory board member for The Sullivan
Group and that he is owner, stockholder, presenter, author, and consultant for Innovative Healthcare Compliance Group.