A little less conversation

Einstein once said, ‘You do not really understand something unless you can explain it to your grandmother’. With that in mind, I set out to design a two-hour tutorial for Part 1 Anaesthesia trainees (equivalent to Primary FRCA) who, according to the syllabus, must be proficient enough in statistics to enable understanding of the medical literature. A tall order for a two-hour tutorial. One thing I get asked most often is, when I have conducted my trial and have my data, how do I know which statistical test to use? This ostensibly simple question cannot, in fact, be answered simply. In order to know which statistical tests can be applied to the data, you have to have a sound understanding of the type of data you have, and the type of distribution this data follows in the general population. It is difficult for most registrars to understand and remember a lot of the statistical concepts because they lack hands-on experience in handling different types of data. This gap in knowledge is best filled with some activities which can be easily understood and applied to situations encountered in the clinic. The first activity I prepared for the class was a demonstration of when and how to apply the Chi-square ‘goodness of fit’ test. The question to be answered is: ‘are the number of red M&M’s the same between packets of different flavours?’ It doesn’t take much explanation to get all those present to agree that colour is a type of categorical data, unless you are a physicist and want to define colour by the wavelength of peak absorption. The residents were divided into two groups, and opening a packet of M&M’s into a plastic bag, proceeded to count the number of red sweets, and the total number of sweets. The data was entered into a previously prepared Microsoft Excel file where I demonstrated how simple it is to answer the question of the distribution of coloured sweets, even without specialist software. A brainstorm of the type of categorical data which we might encounter in the clinic included: type of surgical procedure; anaesthetic technique; types of intravenous fluid (crystalloid, colloid, and blood products); and a whole host of dichotomous (yes/no) questions often found in clinical questionnaires. The Chi-square test is often applied to questions such as: are the surgical procedures performed on my control group significantly different than the procedures performed on my drug intervention group? Or, are there significantly more females who asked for rescue analgesics compared with males [1]? Another often used statistical test is the t-test, in all its different forms. The aim of the next activity was to demonstrate when to correctly use the paired version of the t-test. Each registrar was given five minutes to complete a multiple choice paper consisting of ten vocabulary questions (Appendix 1). The words ranged from the weird and wonderful (Supercalifragilisticexpialidocious) to weird, but which an anaesthetist might be expected to know (sphenopalatine ganglioneuralgia). The multiple choice questions (MCQ) were graded, resulting in a pre-revision test score, then the correct answers were shown to everyone on PowerPoint slides. After a gap (during which the third activity was conducted), the regsitrars were asked to sit the same vocabulary MCQ test again. This score was the post-revision test score. The paired t-test can look at the mean difference within each individual’s preand post-revision test score and give a result which reflects whether people did, on average, improve after they saw the answers. It was then demonstrated that had the unpaired t-test been applied to this pre and post set of data (i.e. if the preand post-scores had been treated as though they were scores from two separate populations and we simply wanted to see whether the two populations scored the same), then we will not be able to discern improvements in individuals (if some individuals