Derivation versus validation

Derivation versus validationAssessing the probable clinical course of children presenting to their care is one of the day to day dilemmas facing the practising paediatrician.With accurate prognostication, invasive or expensive treatment may be targeted towards those most likely to benefit.There is little point in expending vital resources on, or administering invasive treatments to, children who are likely to recover without such intervention.Children with certain characteristics may tend to do better or worse than others.For example, younger children or those with specific clinical signs might be expected to deteriorate more rapidly.Past experience often acts as a guide for the experienced paediatrician.Algorithms (or prognostic models) can be developed to provide a means of transferring expert knowledge to the novice.For example, Apgar scores are routinely used to assess the health of newborn babies, and APACHE scores can be used as a measure of prognostication among admissions to paediatric intensive care. 1 In this issue, Brogan and RaZes 2 present an algorithm for identifying children presenting to A&E with fever and petechiae who are at increased risk of significant bacterial sepsis.It is interesting to note that of the many prognostic models that are published each year, 3 4 relatively few are suYciently validated and fewer still find their way into clinical practice.5 Determining which patients will benefit from treatment is only one use of prognostic algorithms.They also provide a means of informing parents of the likely outcome and can be used in research to give baseline measures of severity in diVerent groups.