A Flaw In Human Diagnosis

Posted by
Noise invisibly impacts the way we see and interpret facts (photo credit: freepik.com)

Doctor’s decisions can be disrupted by “noise” – invisible factors like weather, mood, stress, or even hunger. If not recognized, noise can lead to life-threatening medical errors. So what can be done to improve the accuracy of human judgments in the diagnosis and treatment?

At a glance:

  • Whenever there are random variabilities in judgment, there is noise.
  • Two patients with the same disease can get different medications from the same doctor. Likewise, one patient can get two different diagnoses from two different doctors.
  • Occasion noise occurs when a decision is influenced by factors that should not be relevant to the decision-making process, like mood, weather, time of the day, etc.
  • Reducing noise can help reduce medical errors and improve patient safety.
  • Some doctors tend to be overconfident in their decision-making process. It’s objective ignorance.
  • Algorithms are free of noise. Therefore, by applying a mathematical formula, you will always get the same result, regardless of the number of trials and the circumstances of the calculation.
  • Noise is measurable and can be reduced using decision hygiene.

When you go to the doctor, you expect to receive advice consistent with current expertise—a precise diagnosis and a carefully selected treatment. With scientific progress, the development of new medical technologies, and easier access to information, the quality of medical decisions has significantly improved. But are they always fully rational, consistent, based on scientific facts? Not really.

In every decision, an important role is played not only by the available information but also by external environmental conditions. Sometimes these are subtle unconsidered determinants like mood, weather, or the time of day or year—noise that leads to random variability of judgments. It also happens to you—when you get fantastic news, this influences how you think, interpret information, react. Such fluctuations are unavoidable in everyday life and mostly harmless. But not in medicine.

Scientific research clearly shows that noise determines the outcome of treatment and endangers patient safety. In the recent book “Noise A Flaw In Human Judgment,” Nobel Prize winner Daniel Kahneman, together with Olivier Sibony and Cass R. Sunstein, takes a closer look at how to recognize and reduce bias and noise.

Decision-making as a matter of chance

The problem of noise occurs in professions where subjective judgment plays a significant role. Judges, executives, and doctors are often guided by experience or intuition rather than rigid rules. As a result, a patient consulting two doctors may receive different diagnoses, just as two patients visiting the same doctor at two different times do not always get the same prescriptions or recommendations.

Noise has been confirmed in many scientific studies. For example, one study analyzing pneumonia diagnoses found that 44% of the variation in diagnoses was due to differences in skills, which included education and clinical experience. A University of New York study showed that dermatologists misdiagnosed skin lesions in 36% of cases. When examining blockages in arteries using coronary angiography, 22 doctors disagreed 63% to 92% of the time. Yet another study showed that the later the time of day, the less likely patients are to receive referrals for bowel and breast screening. An extreme example is psychiatry, where—according to various studies—doctors agree on a diagnosis only about 50% of the time.

How to explain these differences? The cause can be occasion noise. In the case of doctors, many factors can be enumerated that increase the likelihood of occasion noise: stress, an enormous responsibility, time shortages, fatigue, the need to serve a large number of patients in a single day.

The noise present in daily decisions is very insidious and dangerous. It can lead to poor treatment efficacy, large numbers of medical errors, and even increased patient mortality that is difficult to explain.

Bias, noise, or human error?

Occasion noise occurs when a decision is influenced by factors that should not be relevant to the decision-making process. For example, the state of optimism after a lovely weekend, a disruptive feeling of hunger, haste.

But this is not the only type of interference that doctors face. There is also level noise, which occurs when two people make different decisions at the same organizational level. For example, the average number of antibiotics prescribed by two doctors working in the same clinic is different, although statistically, they see patients with similar diseases. There is also system noise, which occurs when two doctors make different diagnoses for the same patient. It is prevalent in the case of rare diseases. Studies show that the average rare disease patient in the USA visits several doctors and waits seven years before receiving a correct diagnosis.

When one doctor is more effective in treating children and another in treating the elderly, pattern noise—in this case, a repetitive regularity can be identified and removed.

It is also essential to distinguish between bias and noise. In some situations, they happen individually, but sometimes they overlap. The source of bias is always some fixed set of elements influencing the decision. For example, consider a doctor who consults patients remotely while unaware that their screen displays colors incorrectly. As a result, a misinterpretation of the patient’s skin health can occur because the image does not correspond to reality—a permanent, invisible filter is imposed on decision-making.

Doctors are unaware of many of the interferences mentioned. Even if they do their best to consult each patient in the best possible way, occasion noise cannot be avoided. It is simply an inherent part of human decisions. While it is not harmful in everyday life, it should not be left unaddressed in medicine.

Daniel Kahneman also points out another phenomenon in his book, which he calls objective ignorance. Some doctors tend to be overconfident in their decision-making process. As a result, they ignore available scientific facts and rely on their own intuition. Such tunnel-vision thinking leads to selecting information (diagnoses) that only confirms a preconceived thesis (diagnosis).

Unfortunately, this type of error is still part of the paternalistic healthcare model—the doctor is an authority whose opinion the patient, or even another doctor, should not question. On top of that, an authority who is used to making decisions based on their own beliefs is often reluctant to consult other professionals. 

This belief in the power of self-knowledge and experience can prove to be a trap.

Same judgment for the same prerequisites. Is this possible?

We are becoming increasingly effective at eliminating the factor of uncertainty in decisions. For example, since the introduction of checklists—the first tolls invented to structure decision-making—some medical procedures are always performed according to the same blueprint, regardless of circumstances, place, and time. This significantly increases patient safety.

Thanks to new information and communication technologies, doctors have access to a patient’s complete medical record, prescribed medications, test results, opinions of other specialists—the set of information necessary for precise diagnosis and effective treatment. In addition, clinical decision support systems provide insight into scientific studies and medical recommendations. These measures have helped reduce the scale of system noise, but not of occasion noise.

This is why Kahneman argues for the broader use of algorithms. Many studies confirm that decisions generated by even the simplest statistical algorithms are often more accurate than those made by human experts—even if experts have access to more information than algorithms.

Algorithms are free of noise. Using a mathematical formula, you will always get the same result, regardless of the number of trials and the circumstances of the calculation. Two plus two equals four, whether it’s 6 a.m or 6 p.m. or how many simultaneous operations the computer performs. Computer systems are not susceptible to political views, stress, or mood—they don’t get tired and aren’t prone to fluctuations in performance or precision.

Looking for truth in data

The good news is that noise is measurable and can be reduced using decision hygiene. The first step in identifying noise is to conduct a thorough noise audit. It’s not about pointing out errors but improving patient safety. Some results can be shocking and uncomfortable, leading to resistance from healthcare professionals who misunderstand the goal of noise audit. Conducting an internal “noise investigation” is not technically easy. It requires access to data and thorough analysis according to various criteria: time, the performance of other doctors, results, category of the patients, etc.

Doctors must also learn to take a critical look at their work.

One of the main reasons the noise remains invisible is that some professionals can’t even imagine alternative decisions. Tunnel vision—a reluctance to consider alternatives to one’s preferred line of thought—is in medicine deeply rooted.

Also, cultural factors can matter. Many decisions made by experts are not questioned because of authority what does not foster a work culture based on open dialogue, transparency, learning from mistakes. A doctor with less training will not dare to question the decision of a doctor with 30 years of experience. Whereas what Google suggests to the patient is wrong in advance and not worth checking. Is it patient-oriented care?

The factor of randomness in decision-making may be minimalized by introducing decision observers. Every time a medical case is more complex, doctors should consult specialists in a given field, while patients must have the right to ask for a second opinion. What’s more, the reasoning behind the diagnosis should be described in the electronic patient record. This will help understand the arguments/data behind the decision and compare different views.

Stop guessing. Start analyzing

Wherever possible, intuition and subjective viewpoints in the decision-making should be replaced by scientific guidelines, procedures, and objective scales. To be clear—medicine is a science, and only its practice is an art. And every science is based on facts and evidence, not conjecture.

The truth is doctors work under time pressure and stress—sometimes, they have to make quick decisions. However, every time there is such an opportunity, algorithms, checklists, clinical decision support systems should augment the doctor’s knowledge. They can perfectly guide doctors in diagnosis, help to choose the proper medication. They are created to analyze medical images pixel-by-pixel and carefully scan all notes in the medical records, regardless of the scope of data.

Reducing noise in medical decision-making is not about limiting doctors’ autonomy. It is about eliminating elements that can inadvertently lead to severe errors. Of course, high-quality medical education and continuous learning are always a priority—even the best algorithms are only part of a team with humans leading the way.

The doctor’s right to make subjective decisions based on their own experience and intuition ends where patient safety begins.

Imagine two doctors making different diagnoses for one patient or two judges in the same court giving significantly different sentences to people who have committed the same crime. The book “Noise A Flaw in Human Judgment” describes the variabilities in decisions that theoretically should be identical.

Can I ask you a favor?

Please support aboutDigitalHealth.com (€1+). It only takes a minute. With your donation all content will remain free and ad-free. Thank you!


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s