ESPE Yearbook of Paediatric Endocrinology

Publisher: W&N (18 Mar. 2021). ISBN-10: 1474619967

The authors describe this book as ‘a short, practical, timely guide to the tools you need to understand the numbers we read in the news everyday - and how we often get them wrong’.

One of the societal consequences of the COVID-19 pandemic has been the rising public interest in numbers and statistics. ‘Armchair experts’ in statistics and epidemiology are no longer limited to academic settings, but few have a secure mathematical training. It is therefore concerning that, with this rising interest also comes the worrying explosion of confusing use of statistics and dubious interpretations of data on COVID infections, R numbers, excess deaths, vaccine effectiveness and vaccine side-effects. Public understanding of science needs clear and careful guidance to improve numerical literacy.

David Chivers is an assistant professor of economics and Tom Chivers is a freelance science writer. This combination is welcome because a person’s interpretations of a 1 in 100 risk, versus 1 in 10 000, and 1 in a million, is not simply quantitative. To some, these all seem remote chances, but to others they are all worth taking significant steps to avoid. The authors helpfully suggest to describe other more commonly understood risks for comparison, such as the risks involved in driving or other forms of transport.

Some sections delve into deeper statistical understanding, such as the Poisson distribution and even Bayes’ theorem. Most clinical researcher follow the ‘frequentist’ theory of statistical interpretation, in which probability is based on observed frequencies and proportions, and they look doubtfully on the alternative Bayesian approach, which also takes into account of (often broadly estimated) ‘prior probabilities’. However, when we operate in the real world, the Bayesian approach actually does seem the sensible option. Consider this question: what is the likelihood that you have a COVID-19 infection if your swab test is positive? Even if we know the true false positive rate, the answer varies enormously depending on factors that alter the prior probabilities, i.e. the reason for taking the test (routine surveillance or because you have symptoms) and also on the prevalence of COVID-19 infections in your neighbourhood and workplace.