Reaching and communicating a diagnosis

Pretest probability

When doctors are unable to recognise patterns in presentations quickly, various refinement strategies are used to arrange the possible diagnoses in order of probability. The pretest probability of a disease is the proportion of people in a population at risk who have the disease. For an individual with a new symptom, the pretest probability of disease depends on the context in which the symptom has appeared because the prevalence of disease varies between populations. In general practice populations, the incidence of serious disease, for example, colorectal cancer, is much lower than in hospital populations, although serious conditions still usually need to be excluded. In practice, the pretest probability of a disease is the clinician’s judgment of the likelihood of a particular disease based on the information gathered to date and their understanding of the context in which they work. Clinicians need a mental map of how likely different diseases are and how those probabilities shift as they gather and synthesise information. This may involve identification of ‘red flag’ or ‘alarm’ symptoms and signs of serious disease, for example, or the use of clinical prediction rules, such as the Wells score for deep vein thrombosis. Positive ‘alarm’ features or above-threshold prediction scores increase the probability of a disease in individuals and generally trigger further investigation. Clinicians also rely on understanding the sensitivity, specificity and predictive value of symptoms for the diagnosis of a particular condition in the population with which they work. For example, chest pain is a highly sensitive symptom in the diagnosis of acute coronary syndromes (ACS) as a high proportion of people who have an eventual diagnosis of ACS experience chest pain. However, it is not a specific symptom as many people who do not have an eventual diagnosis of ACS will also have chest pain. If the presence of chest pain alone were used to diagnose, ACS would therefore be overdiagnosed. The predictive value of symptoms is more useful in clinical practice than either sensitivity or specificity as it predicts the likelihood that a person with a particular symptom has the associated condition. Like pretest probability, it is affected by the prevalence of the disease in the population. For example, the positive predictive value of rectal bleeding in the diagnosis of colorectal cancer is higher in older populations than in younger ones.

Additional factors affecting the pretest probability of disease in patients with the same presenting symptoms include age, gender, past medical history, family history and lifestyle. Few doctors use formal probabilistic reasoning in making diagnoses, but most know the relationship between these factors and the likelihood of a specific disease and use this understanding intuitively to select likely diagnoses to subject to hypothetico-deductive reasoning ( Fig. 20.1 ). Initial history, examination and investigation results are used to develop a list of possible diagnoses – the hypotheses. Further history, examination and investigation are used to support or refute each of these putative diagnoses until a final diagnosis is determined. Returning to clarify the history or re-examine matters when signs are ambiguous allows an iterative approach and more accurate diagnosis.

Rare diseases

While diagnosis by probability works in most cases, rare diseases also occur, and to the affected patients and their families, they are not rare. Avoid the trap of thinking that all patients have common conditions, and symptoms that do not fit with common diagnoses are less important. Indeed, occasional patients with a credible and consistent history of unusual symptoms may actually merit more, not less, investigation. The art is to listen carefully, keep an open mind and pick up the uncommon situation when the usual patterns of presentation really do not fit the facts of a case.

Multimorbidity

The application of clinical skills in diagnosis is complicated when patients have multiple morbidity. New symptoms arise in the context of existing physical and psychological illness and may represent new manifestations or complications of a known condition, of more than one known condition, or of a new disease altogether. Typically, patients with multiple morbidities do not experience their diseases discretely and therefore may report symptoms in an indistinct or incoherent way. Furthermore, their symptoms might interact with each other, and present differently compared to a single disease. Faced with this atypical pattern of symptoms, it is not easy for clinicians to reach distinct diagnoses.

Diagnostic error

Diagnosis is not easy, and all clinicians, irrespective of expertise, make diagnostic errors. Errors are more common when presentations are atypical or nonspecific, when patients have comorbidities, or when the underlying condition is rare. Most errors occur because of defects in diagnostic thinking, of which the most common error is heuristic-based thinking. Heuristics are cognitive shortcuts used to solve problems; they are quick and reflexive and used to generate an approximate answer to a reasoning question. They are used, for example, in spot diagnosis, pattern recognition and hypothetico-deductive approaches but are prone to produce error by disproportionately diagnosing conditions that are at the forefront of the clinician’s mind. This could be due to:

• seeing several recent cases

• missing a diagnosis

• settling on a hypothetical diagnosis without gathering enough information to confirm or refute it

• interpreting new information in a way that supports rather than refutes a hypothetical diagnosis, or

• using stereotyping or profiling in clinical reasoning, for example, deciding that a drug-using patient presenting with back pain is seeking drugs rather than investigating the cause of their pain.

Many strategies have been proposed to debias diagnostic thinking. Metacognition is one such strategy. It promotes awareness and understanding of your own thinking as a way of recognising and minimising unconscious bias or errors. It encourages you to check for conflicting evidence and consider alternatives to the decision you have arrived at. For example, in reaching a diagnosis, it may be helpful to stop and ask yourself:

• ‘What else could this be?’

• ‘How much is my decision being influenced by the fact I am running late?’

• ‘How much is my decision being influenced by the patient I misdiagnosed last week?’

As a reflective process, metacognition can be learnt and practised.