The first key steps in managing a patient's problem consist of accurately assessing the situation, making a clear diagnosis, understanding the psychological and social impacts of the illness and negotiating an effective treatment plan. The first of these, the arrival at an accurate diagnosis, if we are to undertake this using high-quality evidence, can be more difficult than we would like to admit.

There are a number of different schools of diagnosis. Although this might at first seem like an odd statement, a step back and it makes sense. There is the classical Bayesian approach: you start with inkling about something, do a test (such as asking a question) and incorporate the answer of this into your modified probability of disease, undertake a further test, and so on. Other ways of looking at diagnosis are highly visual: the categorical approach of a haematologist examining an abnormal blood smear or the gestalt approach of a neurologist to classifying a movement disorder. In some settings, such as rheumatic fever or Kawasaki disease, we seek a criterion-based diagnosis. Different diagnosticians seek to create further knowledge about the problem—for example, the neurologist asking, “What level is the spinal cord lesion at?”1

When you try to break things down to pretend that we are undertaking scientific, probability-modifying diagnostic testing, you soon see that it is a simplification. Using estimates of accuracy for a single test to quantify the way that probability of disease changes underestimates how it is actually used. Many tests provide far more information about the patient and their condition than simple presence or absence of disease (eg, location of a tumour and its risk of complications), and diagnostic tests are often pieced together in a chain of information to arrive at the underlying problem.

We have also known that the arithmetic …