Dr. Art Papier is the CEO of VisualDx, which is a visual clinical decision support system specialized in medical image recognition and analysis. The company’s tagline is “quality care begins with an accurate diagnosis,” which to me is so perfect, it almost comes across as a little ironic or tongue-in-cheek. Because of course, good care requires a good diagnosis, right? And yet, that misdiagnosis is a real problem in our healthcare system today. So we’re going to get into that and what it means for both individual patients and the system as a whole.

THE PROBLEM

Start by setting up the problem of misdiagnosis.

Patients come to physicians and advanced practice professionals with symptoms. And these complaints lead to the clinician taking a patient history, doing a physical exam and the clinician assessing the problem and coming up with diagnostic possibilities. It turns out that between 10 and 20 percent of all diagnoses are wrong, and that’s been the case for years. It just hasn’t been on the radar in the healthcare debate. We’ve been talking about problems like hospital-based infection, hand-washing, falls in hospitals, wrong site surgery. The National Academy of Sciences issued report two years ago called Improving Diagnosis in Healthcare that made this incredible statement in nearly every person […] will experience a diagnostic error in their lifetime.

What proportion of that 10 to 20 percent rate is comprised of really serious mistakes?

A lot of people think it’s for surgical mistakes or obstetric mistakes, but it's actually diagnostic error that is the majority of the lawsuits Click To Tweet

A lot of people think it’s for surgical mistakes or obstetric mistakes, but it’s actually diagnostic error that is the majority of the lawsuits and head the list with cancer […] and then problems such as missed sepsis – which would be a life-threatening infection – a pulmonary embolus, stroke. Those are the kind of diseases that cause serious harm, but that doesn’t necessarily make up the majority of these diagnostic mistakes because there are mistakes of non life-threatening diseases [like skin rash].

What are some of the causes of diagnostic errors?

Cognition and bias:

The research shows that the majority of mistakes are cognitive mistakes, meaning that a patient comes to a provider and gives their symptoms and their history, the problem is analyzed, and the clinician – between 10 and 20 percent of the time – jumps to the wrong diagnosis. So there are cognitive mistakes such as premature closure where you’re immediately jumping to what you think it is without including all the possibilities, or anchoring, where you latch onto a prior diagnosis.

Education and human limitations:

We teach medical students to memorize prototypical disease and illness scripts of how a disease presents, but it’s impossible as a primary care physician or urgent care physician to memorize all the variants of presentations of each disease.

In medicine we have this saying, “common things happen commonly, and if you’d hear hoofbeats think horses, not zebras.” So we’re trying to teach medical students not to go way out on a limb and think about all these rare diseases. We’re trying to encourage them to rule out and think about the common diseases first. And unfortunately, this can get you into trouble. If you memorize common presentations, well, common diseases often have variant presentations. […] In a memory based system of education we’ve asked students and residents to memorize the classics, which is incredibly efficient and it works and 80 percent of the time. But you know, you wouldn’t get on the airplane if the pilot said they were getting it right 80 percent of the time.

Structure of the healthcare system:

There’s a trade-off called the thoroughness-efficiency trade-off. And because we’re a fee-for-service medical system, we’re valuing speed over thoroughness. So your physicians are being reimbursed in a fee-for-service world, we’re seeing a high volume of patients. If you’re high volume, if you’re fast, you can’t be fully thorough. If you’re fully thorough, you can’t be efficient. So there’s a sweet spot in the middle, and the way we’re reimbursing medicine is not valuing quality and thoroughness.

What’s the cost to the system?

It depends. If you have a full hospital, you don’t have enough beds and you’re wasting beds on patients who really don’t need to be there it’s costing a lot of money. So as an example we know that close to 30 percent of the patients admitted with cellulitis do not have cellulitis. […] Each admission for cellulitis is about $10,000. Then you say, of the 30 percent that don’t need to be here, probably 20 percent of those we should err on the side of caution. So there’s a certain percentage of these admissions that don’t need to be here. And you could do the math on the hospital side and see that if you just dial back that one problem, you save a boatload of money. Just for one diagnosis.

THE SOLUTION

Define clinical decision support

Clinical decision support means that you’re giving information to the clinician [that she couldn’t do from memory] at the point in time of decision making in context to the unique individual that’s in front of you. […] The analogy really is to have a cockpit of information just like a pilot has. The pilot has instruments and checklists.

Talk about the role of technology in clinical practice

We started to see a change with the advent of the smartphone. So once the iPhone came out we started to see doctors and students and residents grabbing information from apps on their phone. They answer questions like “what’s the dose?” from an app on their phone. We launched VisualDx on the iPhone 10 years ago, and now about half our traffic comes through mobile devices and the other half comes through the desktop and now connection through an electronic health record.

I think what we’re saying is the largest technology companies in the world are getting involved in healthcare and you know, obviously there’s a lot of hype around AI and machine learning, but some of it is, is justified. So, you know, we actually are doing machine learning and can take a photograph of a skin rash and categorize that physical exam or ready. So whenever that meeting, so I’ll try to see what’s new and ai and machine learning and what’s the technology companies are certainly, apple was very progressive. Things would bring in the electronic record data into the health app on the phone. So it’s amazing. It’s really assisting the movement towards open notes and making sure that people have their health data secure and on their phones, but they control the data. So I’m very, very interested in a consumerism in healthcare because we know as a company we don’t believe that there shouldn’t be a separate consumer health company in a separate professional health company. We, we see that there’s a real need to have everybody on the same platform. So we’re very interested in platforms and companies that are innovative from a technology.

On the need for a broader view of health information technology

A really important point [is that] patients see their doctors using the electronic health record, and there’s a lot of frustration with the clinician becoming a typist, becoming a bureaucrat filling out screens instead of interacting with the patient. That is not fully the category of health information technology. Click To Tweet So now we have digital radiology on our screen, which means your orthopedist has the x-ray right with you in the exam room and point to that x-ray. We couldn’t do that 10, 15, 20 years ago. […] So there are many digital improvements in healthcare. But the dissatisfaction with the electronic health record is causing people not to think about the advantages of tools that do decision support.

How can the healthcare system drive adoption of these tools?

We’re now seeing the shift where, in particular, the students and the residents are grabbing these tools and actually teaching the teachers who aren’t comfortable with them. A really important point is that medicine is not like aviation where all pilots talk to the tower in the same way and follow the same processes. In medicine, it’s all over the map, where people have been trained in different medical schools by different mentors, and they have different ways of solving problems and different workflows. The promise of clinical decision support is to put these tools across the enterprise and everybody starts to perform at a higher level.

[One way to do this is] continuing medical education credits. VisualDx has been accredited for point of care CME, meaning that every search a clinician does in VisualDx is tracked and they can get a half a credit of CME. And the idea is that when you look up information in a resource like a Visual Dx for your patient that’s in an exam room, that’s stickier learning than sitting passively in a classroom, which is what CME has traditionally been.

And so [a] liability insurer actually gives a discount on the liability insurance to the clinician for using diagnostic clinical decision support. The liability insurer realizes that diagnostic errors are the number one reason to be sued and they’re saying “we want clinicians to be using these tools.“

Do you ever see these tools shifting expertise to the point where people can self-diagnose?

In a limited way, I see that happening fairly soon. But there’s so much complexity I don’t think this is going to be a man-versus-machine debate. I think that gets a lot of headlines, you know, “Big Blue can beat the chess masters, therefore computers are better than the people.” I really feel it’s the computer augmenting the brain. So we describe AI as augmented intelligence, that these tools are going to be our co-pilot as patients and they’re going to be our co-pilot as physicians. I don’t think it’s really worth discussing in 2018, are we gonna just have no doctors. There’s a real human component to medicine where people want to interact with people and have them understand their problems and listen to the emotional and psychological component of a lot of medical issues.

Share This