It’s been the misfortune of every living doctor, and those no longer with us, to pick up the result of a test they’ve ordered and wish they’d never signed the request in the first place. The test result is there, you understand what it means (technically), but also struggle with what it means (practically) to the patient before you. There are psychic echoes of radiology consultations, “and how with this x ray result change your management?”, and you wish you’d internally digested what was being said.

In diagnostic testing, we hope to reach a state where we are adequately convinced of the presence or absence of a condition, so that we may either treat it, test for it, or no longer worry about it. This instinctive process can be thought of as creating two lines in the shifting sands of diagnosis and working out where we now stand. Above one line, you’re certain enough of the diagnosis to treat. Below the other line, you’re convinced of the condition’s absence, and so need not consider it any further. Between the two there is an uncertainty that requires some form of “test”.

The upper threshold is created based on the risk versus benefits of a treatment, and so how certain you need to be about a diagnosis before you begin therapy. To take two extreme examples: one needs to be absolutely certain before embarking on a course of chemotherapy and/or radiotherapy that the lump you’re dealing with is a sarcoma. On the other hand, you don’t have to be that convinced that a child is suffering from a viral coryzal illness before you prescribe paracetamol.

The lower threshold is drawn when you are convinced that the chance of disease, and consequence of not investigating further at this point, are so low as to require no more action. The child with acute shotty cervical nodes and tonsillitis is extremely unlikely to have lymphoma, and wouldn’t require biopsy. (It’s not that the diagnosis is impossible—just that the balance of risk:harm weighs against excision biopsy for every child with tonsillitis and cervical lymphadenopathy.)

In between these thresholds is where you would test. But your test (or series of tests) needs to be good enough to either push you into a convincing denial of the disease, or enough evidence to commence therapy. (It shouldn’t be forgotten that in many instances we use “waiting” as our chosen test.)

Where the evidence comes into this mix is in refining the balance of benefits versus harms (for more on this, see Phillips¹) and demonstrating how good a test is in its ability to move you up towards treatment, or down towards exclusion (for more on likelihood ratios, see Phillips²). What it can’t do is rescue you from the test whose accuracy you don’t know about in a patient whose chance of disease you hadn’t thought about.

This month’s Archimedes helps us along the path of diagnosis by exploring the clinical examination (for ankle fractures), technical aids to examination (for jaundice), and the risk:benefit of testing (in isolated hemi-hypertrophy). Where we drawn our lines is based on clinical expertise, experience, and the attitude of the family we’re dealing with.