The essential elements of a differential diagnosis study are, like most of critical appraisal, really simple and straightforward.

You need to start with a bunch of children/young people who turn up with the symptom, or symptom-complex, you’re interested in. Ideally, you need these folk to not already be known to have something, to attend a similar type of facility (e.g. office for general practitioners, or ED for ED types), and to be a consecutive group or random sample.

(If they are already known to have something, there can be a sort of “detection” bias – where if you know someone had Gilbert’s syndrome you may be more able to detect the jaundice than if you didn’t. If they are at a different level of referral, you may well see s different spectrum of illnesses, either aetiologic or of severity – for example lethargy and weight loss seen at a GP surgery and in an oncology outpatient clinic. The sampling aspect is about how the group may reflect the general population; non-random, non-consecutive samples can be ‘selected’ – so if you’re looking at joint tenderness, and there are no nighttime or weekend attendees, the diseases diagnosed may have a lower proportion of acute traumatic or infectious causes than if you’d sampled differently.)

You need to be able to have a convincing way of making a ‘final diagnosis’ for each patient in this group. Without this, you end up with a list of possible and maybe and no amount of fancy stats will get you out of this.

You also need, ideally, to be able to identify key differences between the symptoms clusters that lead to different groups of diagnoses. This is very much a supplementary, and has issues akin to other subgroup analyses if done post-hoc. (You can take chance findings and assume they are real and perpetuate this by then only looking at the (falsely) restricted differential diagnoses you have come up with).

You then need to list, with a rough idea of how likely/what proportion of people, end up with the diagnoses.

That’s it. Nowt to it. All you need to do then is interpret any differences from this ideal study design and decide how likely they are to bias the results, in what direction, and by what degree.

Critical appraisal in a nutshell. (Differential diagnosis: pistachio, peanut, walnut, hazelnut, almond … v rare cases of chestnut have been known.)