Chapter 9: Medical diagnostics – Measurement, uncertainty and distributions

Blood tests and other clinical diagnostic tests are important for health practitioners to be able to make informed evaluations of patients. Collection and interpretation of this diagnostic data is central to evidence-based medicine. A diagnosis can be thought of as a prediction made by a health practitioner that can be supported by the presence of findings (i.e., certain diagnostic indicators) that occur, by definition, in patients with that diagnosis and not in others. If the diagnosis is to have some value to the patient, then the treatment, advice or intervention associated with the diagnosis should provide a benefit. On the other hand, for any diagnosis which is excluded, the patient should not lose any benefit by not being offered any treatment, advice or intervention.

For a diagnosis to be evidence based, there needs to be some sense of its reliability or certainty. That is, we must ask the question, what proportion of times is the diagnosis correct? To answer this, we need to understand types of data used for clinical diagnostics, how variability in data can be measured and the probability of correct diagnosis.