Clinical Reasoning Theories

Dual Process

Dual Process Theory is an off shoot from psychology, its an attempt to put a model to how we think. Simply it puts our reasoning into 2 camps, referred to as Types or Systems. The initial work comes from Danny Kahneman and can be read in his pop science book, Thinking Fast and Slow. I am sure we will come back to Kahneman and his colleague Amos Tversky later when we look at bias. The biography The Undoing Project is an excellent read if you are interested in their work

Type 1 Thinking

Intuitive, efficient and based on pattern recognition

Happens so quickly it is not recognized as a distinct cognitive process – a “gut feeling”

Type 2 Thinking

Hypothetico-deductive, analytical.

Time intensive and deliberate

Requires conscious thought.

 A little aside on this at this point, whilst it might seem that Type 2 would be more fatiguing, Type 1 allows for many more decisions to be made and so can cause a similar amount of Decision Fatigue to Type 2 but as it is more unconscious it may take the decision maker by surprise.

An example of Type 1 thinking – what diagnosis pops into your head when I tell you about a young lady on the COC comes in with unilateral leg swelling, Shortness of breath and pain in her chest when she takes a deep breath. That little voice that says PE is the your type 1 thinking.

Type 2 thinking would come with a more complex or atypical patient. For example, an elderly gentleman with an AKI, weight loss and back pain. The diagnosis isn’t suddenly apparent and we have to stop and think, could it be myeloma, could it be another malignancy, could it be a chronic inflammatory condition like RA/Ank Spond.

So which system do we use when? Now this is varied and individualized. People are more likely to use Type 1 when they are more experienced but also when their own Illness Script matches the patient’s presentation. Medical Students, with their under formed Illness Scripts and lack of experience use a lot of Type 2 thinking. This is why being on a new placement or starting a new job is exhausting. It seems that even seeing the same patient in a new environment will make it more likely to be using Type 2 thinking.

The idea that thinking is polarized is an oversimplification. In reality we are using both all the time, as more experienced clinicians we often form an initial impression from the presenting complaint and triage or other people’s notes and then check our thinking when we see the patient, unconsciously using Type 2 thinking to check our Type 1.

A lot of early medical training, particularly in undifferentiated patients is about seeing patients when you have the supervision and safety net of a seniors around to hopefully shield you from the bigger situation allowing new doctors to take the time to use their Type 2 thinking and develop the ‘database’ of patient presentations facilitating long term Type 1 thinking.

Now as doctors develop experience there seems to be potential for mistakes to occur. When you start to be able to pattern recognize, ie type 1 thinking, but you haven’t yet quite completed a sufficient database to avoid making mistakes. It seems to be stage all doctors go through and probably correlates a little with their positions on the Dunning-Kruger Curve. 

Now why do people make these errors? Both Type 1 and Type 2 thinking are prone to error and we will go into bias in detail later. Type 1 for example is prone to anchoring and not properly weighing up the information available if it doesn’t fit the pattern, leading to an incorrect diagnosis. Type 2 is prone to placing undue emphasis on a particular part of our exam, history or testing leading us to be misled. Remember its important to recognize that not all our tests are perfect, well none of them are, and need interpreting with the clinical context. The commonest error I see is when it comes to XRs, people rule out the fracture because there is nothing on xray, forgetting that the X-ray is to confirm our clinical diagnosis of fracture.

One of the ways we use these systems of thought to help our patients is when a junior colleague, doing something they lack experience in, discusses it with a senior. In Emergency Medicine there are some high risk presentations that have to be discussed, for example a traumatic chest pain in the over 30s and abdominal pain in the over 70s. Behind each of these patient groups there are potentially significant and catastrophic diagnosis but we reduce the risk for our patient by using the type 2 dominant thinking of the junior member of staff talking to a type 1 dominant senior member of staff, hence reducing the bias inherent to each system.

Illness Script

These are what our mental model of an illness looks like, it summarises everything we know on the topic, hopefully in a succinct way to help it be memorable. This model evolves over time with experience and further knowledge, and given that we all think and work in different ways, every clinician will have a different model of each disease.

 When we take a history and start talking to patients we checking our Illness script with their script, how they present. Through the process we form our differentials, working diagnosis and final diagnosis.

As these are mental models they can be quite complex but one way of presenting them can be with the 4Ws (Who, What, When and Why) and Don’t Miss:

-          Who

o   These are our risk factors. Eg demographics, health related behaviours, exposures (ie travel, drugs), Family history, past medical history

o   It sets up the risk of that patient

-          When

o   This is about the tempo of the illness

o   What is the onset like – days, weeks, months, minutes, hours?

-          Why

o   Mechanisms of the disease that increase our clinical suspicion

o   This is where our illness script will create specificity

o   Includes pathophysiology, Key syndromes and an anatomic approach

-          What

o   Have we got specific feature that lead us one way or another

§  Key lab results

§  Key Imaging

§  Key pathology results

§  Key examination findings

-          Never Miss Features

o   These are signs that are pathognomic or are significant markers of deterioration, increased mortality or disability

What we put in the model is important, it can easily become cluttered beyond use. It may be reasonable to put only disease features that are indicative of a disease condition, ie raised JVP and S3 sound in heart failure. It could also be useful, if dealing with undifferentiated patients, to make Presentation Scripts with presenting complaints instead of diseases, for example instead of heart failure, one for Shortness of Breath. A similar structure of 4Ws and Never Miss can be applied to this as well. This is a useful teaching tool to engage people at the start of a session.

The Problem Representation

The problem representation is a one sentence summary that highlights the defining features of a case.

It can be used as a way to focus a clinician, allowing thoughts to be processed or formed leading to a differential diagnosis. Ideally it should be addressing all three of the following:

1.       Who is the patient?

a.       the relevant demographics and risk factors

2.       What are the timings of the presentation?

a.       Acute, Chronic, Hyperacute? Ie how long has it been going on

b.       Tempo of the presentation? Improving, intermittent, resolved, progressive, stable etc

3.       What is the clinical situation?

a.       Key Signs and Symptoms

They work in a way of activating illness scripts and challenge the clinicians memory and current illness scripts with the current patient. In doing so there will be an almost automatic prioritisation of the differential diagnosis depending on the patient’s script/problem representation and the illness scripts the clinician has already formed. Part of forming a Problem Representation is taking the patient’s history and examination and turning it into the appropriate medical language.

With A+E departments being busy, we use problem representations as part of our handovers as they are succinct and full of useful information.

Here is an example of a problem representation:

A 60-year-old woman with rheumatoid arthritis presents with one day of left ankle pain and swelling as well one week of malaise. She has been on prednisone 20mg daily for the past 6 months. On exam, she is febrile and tachycardic, with left ankle oedema, erythema, and tenderness with active and passive range of motion. Blood work is significant for a WBC of 15.

A concise problem representation of this case would look like:

A 60yo immunocompromised woman presents with acute monoarticular arthritis and a systemic inflammatory response.

Just by presenting it this way allows us to immediately start forming a differential of what is going on. In summarising the salient feature and minimising distractors we reduce our cognitive load and facilitate problem solving. A large part of this is through minimising red-herrings and distractors but also the translation into medical language activates our illness scripts. simply we are turning a chaotic presentation that represents the real world into something that reflects our textbooks and lectures when we learn medicine.

Bayesian Thinking

Bayesian Thinking (from Thomas Bayes in the 1700, theorem and statistical principles) - https://www.upgrad.com/blog/what-is-bayesian-thinking-introduction-and-theorem/

-          The probability of a hypothesis being true is dependent  on 2 conditions:

1.       How reasonable is it based on what we already know?

2.       How well does it fit with the new evidence

Works from the basis of conditional probability (likelihood of an event or outcome based on the occurrence of a previous event or outcome). Calculated from multiplying the probability of the event by the probability of the subsequent or conditional event

Bayes Theorem allows us to update our beliefs and convictions based on new pieces of information. Ie if we are trying to work out the probability of someone having cancer we can assume the percentage of the population that has cancer. However extra evidence such as smoking, family Hx and change our perception (hence the probability).

This can be applied to most of A+E (undifferentiated patients with a wide differential on the presenting complaint). As we gather our history each extra part changes with probability/perception of what is most likely the problem. Our investigations then add to this with their own sensitivity.

With these structures and theories of how we think and how we form diagnoses we can understand the other headliners in clinical reasoning theories:

Hypothetic-deductive

Pattern Recognition

Both of these we have discussed when we touched on Dual Process Thinking. In more detail, Hypothetic-deductive describes how we form an idea and then challenge it with new information. We run several hypotheses in parallel as part of our differential formation and adjust the likelihood of each with further information we gather through our history and examination. As medical students some of us learnt to take histories by learning the associated symptoms with each presenting complaint that would allow us to challenge each differential and this is an example of a Diagnostic Scheme, which we will look at when we come to understanding how we develop clinical reasoning (for example I remember learning that when someone has Chest Pain its worth asking them if they have a cough or fever, if they are short of breath, have leg pain or swelling etc etc). This way of working reflects Baysean thinking and working on conditional probabilities in our heads with each piece of new information.

Pattern Recognition is when a you have a database of presentations from patients in which you have previously diagnosed and new histories and examinations fit into those patterns – the presentation but even the words and phrases used can lead you to suspect a diagnosis strongly. As you can see this fits with our Type 1 thinking well.