TY - JOUR T1 - Prediction models for neonatal outcomes: are they useful? Are they feasible? JF - Archives of Disease in Childhood - Fetal and Neonatal Edition JO - Arch Dis Child Fetal Neonatal Ed DO - 10.1136/archdischild-2013-305700 SP - fetalneonatal-2013-305700 AU - Lisa Barker AU - David Field Y1 - 2014/03/25 UR - http://fn.bmj.com/content/early/2014/03/25/archdischild-2013-305700.abstract N2 - Randolph and colleagues1 report their attempts to develop a prediction model for death and/or neurodevelopmental impairment (NDI) in a group of extremely low birthweight infants. They focused on the ability of low pH and or acidosis to enhance the precision with which prediction of this combined outcome could be achieved. Ultimately they were unsuccessful with other parameters being more important in the final model than measures of acidosis.They conclude that NDI is not just the product of the early neonatal course, and as a result prediction based on condition at birth, or in the period immediately after birth, is perhaps not feasible. This may well be true but it is perhaps also worth reflecting on our rationale for seeking predictive models and whether there would be benefit from adopting a somewhat different approach.In general these types of studies have had one or both of the following broad goals: To provide a tool that permits risk adjustment between services that have different case mixes. The use of early data allows babies’ outcomes to be compared and adjusted based on their risk at the time they entered the service. To provide clinicians and parents with a means of looking at the likely outcome of a child early in their course so that care can be orientated appropriately, perhaps from an intensive to a more palliative approach.These goals are linked as the first (risk adjustment) looks at the risk for a population whereas the second (likely outcome) looks at … ER -