Sometimes we are in situations where we think that something causes problems, and we can’t do a trial randomising one group to get something which we think causes problems! How do we then go about finding out – how to we avoid the problems of ‘confounding’ – and what is that anyway? For example, think about necrotising enterocolitis. Which babies develop NEC? The tiny ones, the septic ones, the ones with terrible birth histories, those who are transfused more often, and those who have umbilical lines. Which of these factors are important in the cause of NEC? If we could work this out, we may be able to modify them and reduce the risk. But there’s a significant problem: these factors are all interrelated. Which babies get umbilical lines? The small, sick, and septic. Which babies get sick? Tiny ones with bad birth histories and infections. What we need to do is work out what factors are associated with our outcome, which factors are associated with each other, and then decide how these associations act. Are we dealing with confounders – something that explains the apparent relationship between two factors (e.g. marriage, which at a population level appears to cause babies) or effect modifiers (e.g. umbilical lines in septic babies which increase the chance of thrombosis). Teasing apart these aspects requires clever statistics but underneath that a fundamental clinical intelligence to suggest which factors should be examined, and when reading papers that purport to tell you how something causes something, make sure you use that first before believing the stats.