The Anatomy of a Medical Diagnosis

By Gerrit Van Wyk


There is what is supposed to happen during a medical diagnosis, there is what can make it better, and there is what we believe about what happens during it.

When you attend medical school, they teach you a formula to diagnose that looks something like this.

You take a case history, during which you collect information about a patient’s complaint, the patient’s past illness history, and social information for locating the patient and complaint within a social context. Next, based on that, you perform a physical examination, which, in my day, was a full body examination, with an emphasis on the systems of the complaint. Next, there was something called in-office side room examinations, which included things like urine tests, an EKG, basic blood tests, etc., which could be used to gather additional information, and then, once you collected all this information, you made a list of possible causes for the complaint, called a differential diagnosis, ranked from most to least likely. You then used the differential diagnosis as a guide to run additional laboratory tests and X-Rays to confirm, or exclude, each of the provisional diagnoses, until you arrived, by testing, at a final diagnosis, which becomes the basis for treatment, which could be advice, a prescription, or procedure of some kind.

If you cannot arrive at a diagnosis, or the condition is too complex for you to manage, it is appropriate to refer the patient to a physician with the appropriate training and expertise who can do so. This is what is supposed to happen, but during my career, I didn’t see physicians follow that protocol to diagnose and refer a lot.

Another way to look at a medical diagnosis is from the perspective of solving a problem. The patient is there because it detected a health issue and don’t know what it is, and therefore don’t know whether it should be concerned about it or not. The patient, directly or indirectly, pays you to diagnose the condition, so one can determine whether something must be done about it or not.

Since medicine claims to be a health “science”, it is reasonable to approach a diagnosis like a scientific problem. In other words, the purpose of taking a medical history, doing a physical examination, and side-room examinations, is to gather information leading up to a hypothesis about the problem (the differential diagnosis), which you test with your special examinations to arrive at a conclusion (diagnosis), which is tested by an intervention (advice, prescription, or procedure), which is the equivalent of a scientific experiment. It means, unlike the traditional diagnostic formula, treatment is not the endpoint, it is what proves or disproves your theory about what is going on; if it proves to be correct, you learn from it, and if it disproves it, you must revise your theory by collecting more information, including about why the experiment failed, revise your hypothesis, and retest it. A scientific approach to diagnosis is therefore circular, not linear like the traditional protocol. I rarely saw physicians use the latter, and never the former.

What becomes immediately clear, is the accuracy of a diagnosis, and therefore the likelihood it will solve a medical problem, critically depends on the amount and quality of information gathered. Next time you go to your doctor, or speak to a nurse, notice how often the bulk of the diagnostic protocol or scientific cycle is bypassed to jump straight to the solution. That is the statistical equivalent of making a lucky guess, or buying a lottery ticket.

Secondly, notice, if there is a physical examination, how brief and perfunctory it is, often through your clothes, which is mind boggling. As a young urologist, I did my own microscopic examinations on urine etc., urodynamic testing, etc. Those days are long gone, hence the side-room as a source of information no longer exists. Finally, notice how often the laboratory tests and x-rays you go for substitute for the lack of information from the absent proper physical examination, as a fishing trip casting a big net roping in all kinds of creatures in the hope, amongst them, you’ll find the right one. In other words, trying to improve the likelihood your initial guess is correct by trawling with technology.

As much as 40% of medical diagnoses are wrong, and wrong diagnoses lead to wrong treatments and mistakes, and we know a shocking number of them happen. You don’t have to be a rocket scientist to figure out guesswork based on minimal information within an industry as complex as healthcare is risky.

Anatomy is the study of the structure of the human body, physiology the study of the function of the human body. The traditional and scientific approaches are the anatomy of diagnosis, the social game within which it plays out the physiology of it. To understand why doctors and nurses ignore the anatomy of diagnosis, you must study its physiology.

Side-rooms disappeared because there is no reward for the time and effort spent within them. For that, we can thank third party payers. I read this week large multi-specialty medical practices in the US, which is the catnip of healthcare, are dying out because they can no longer be run profitably. Long, leisurely consultations and interdisciplinary referrals are anathema to the cost cutting and efficiency craze amongst healthcare administrators. In their own words, they rely on rapid turnover, which means short consultations, which is an incentive to butcher information needs to the bone, and for financial incentives securing that outcome. As it turns out, in addition to the doctors, we should also blame administrators and policy makers for the mistakes that harm us and we die from.

Another major problem is the healthcare shift to technology, and the unsupported belief of many that technology is accurate and rarely fails.  Hence the idea that technology will fish up what is missed during the history and lack of physical examination. What everyone forgets is the technology is used by and interpreted by humans, who are fallible along many dimensions.

Many technology believers have a fantasy that artificial intelligence will supplant fallible doctors and nurses. Since labor makes up the bulk of healthcare expenses, that thought must make the hearts of managers and policy planners go pitter patter and eyes gleam, because of the massive cost saving. But the technology must be programmed to be perfect, and no matter how you do it, it takes time to collect the necessary information from patients, and technology can’t overcome that limitation. Of course, AI machines don’t examine patients, they, even more so than what we do today, rely on the belief information from a physical examination and side room can be bypassed to directly go to laboratory and x-ray technology for diagnosis. Somehow, there is a belief machines can make our current inefficiency efficient, but that philosophy is wrong and the outcome remains inefficient.

Behind healthcare in Canada lurks a highly legalistic regulating system. All doctors will tell you receiving a letter marked Personal and Confidential is the cause of several days of diarrhea, and sooner or later, we all receive them. That has two consequences for the technological approach. The first is it becomes a defense against complaints about error, and the second is, those who interpret them never commit to a proper diagnosis. Early in my career, one sent a patient to a radiologist or pathologist with the question: with these symptoms and signs, could you please use your technology and tell me your diagnosis? Today you receive a report describing a seven-vehicle pileup without calling it an accident, to avoid the risk of being wrong and getting a complaint. So much for the accuracy of technology.

Doctors and nurses have high status in society as medical experts, and with it comes hubris. When you believe you are an expert, you also believe you know a lot, and that you don’t make mistakes. No-one in healthcare can today know more than a fraction of the available healthcare knowledge base, yet we act as if we do. In 2002 the Harvard Business Review reported there are 10,000 diseases and syndromes, 3,000 medications, 1,100 laboratory tests, etc.; no physician can know them all in their different combinations and permutations. The only way to avoid that problem is to rigorously follow the traditional diagnostic or scientific diagnostic protocol, but, because of the hubris, we don’t. Instead, as Tversky and Kahneman showed, doctors and nurses use the same mental shortcuts as average people to diagnose, and, consequently, often get things wrong. They rail against Doctor Google but then, oddly enough, do things the same way.

Take prostate cancer as an example. A recent study shows doing a rectal examination is useless for diagnosing prostate cancer, which is a useless finding. You get a lot of information from doing a rectal examination; is the prostate enlarged, is it symmetrical, are there any abnormal areas within it? The point is, a nodule is suggestive of cancer, a normal examination doesn’t rule it out. For that, you need additional information, such as a Prostate Specific Antigen (PSA) blood test; if it is abnormal, and your rectal examination is abnormal as well, there is a significant chance you have cancer, for which you need more information; a prostate biopsy. Bypassing the examination, as most doctors do today, or using AI, makes you go straight to a PSA test, which is not diagnostic of cancer, which leads to a biopsy, which in some cases have serious complications, and two more problems. A negative biopsy does not confirm you don’t have cancer, just there is no cancer in the tiny sample of a much larger organ. Secondly, this approach diagnose many cancers that mostly have a benign course, which triggers anxiety, unnecessary treatment, and even more complications. I would argue this example makes a strong case for a scientific approach to diagnosis, and dismantles the argument for instant-on 10-minute diagnoses and AI.

Research shows most doctors today qualify in Canada without ever doing a rectal examination, and those who did, don’t know what the examination means. When I qualified, we were examined on patients with real pathology; if you didn’t do the examination and missed the diagnosis, you failed. Today, to ensure “fairness”, actors are used pretending to have the symptoms, and when you get to “the examination”, they whip out a cue card with the finding. You don’t get points for what you know and find, you get points for showing empathy during the mock examination. For this deficiency we can blame our healthcare education system.

I’m now a retired boomer, which means I became a consumer, and am no longer a provider of healthcare. I long for the fairy-tale healthcare land existing when I became a doctor, when the anatomy of diagnosis made sense, but that, long ago, was bulldozed and paved over by the industrial management turn in healthcare, and I must accept it will never return. As a patient, the new philosophy based on the physiology of diagnosis as the outcome of a socio-technological game scares me. It lacks the precision of the past.