I did not to find anywhere in Australia a single instance of truly integrated teaching of statistics and research methods through PBL. I have to conclude from this that there are considerable difficulties. I did, however, find an instance of unintegrated PBL teaching, at the University of Western Australia, where anatomy, biochemistry and physiology have continued to be taught in a more traditional manner.
What are these difficulties?
• If tutors are drawn mainly from the laboratory disciplines, they may be unsympathetic to the population and clinical foci of EBM and its core subjects. We need strong advocates to persuade them of the importance of these topics.
• If tutors are drawn from the clinical staff, they may be unsympathetic to the idea of EBM as a core activity. Converting the existing clinical teachers to the new paradigm of EBM may take a generation, but this will be a problem whatever teaching method we use.
• Tutors may be ignorant of the principles and details of statistics. This may be true. My own experience as a PBL tutor has been that lack of knowledge has seldom been a problem. It is not the function of a PBL tutor either to impart knowledge or to explain concepts. Persuading tutors of this should be part of their general PBL training.
• The changing patterns of cases as the course develops is a particular problem for statistics. This is undoubtedly true.
• The nature of the subject does not lend itself to PBL. It would certainly be difficult to deduce the calculations required for a t test from a problem based on a patient with asthma. However, we must consider what we actually want to teach. I do not think there is much point in teaching undergraduate medical students to analyse data, even using computers. What we need to teach them is how to understand research publications, the evidence on which we hope their future evidence-based practice will be based. We can certainly do this using triggers such as published papers, as the University of Western Australia has demonstrated.
• The patient case is not suited to teaching these subjects. This is true, though I do not think it is beyond us to devise cases which lead to research and evidence questions. However, we must persuade our colleagues that the patient vignette is not the only type of problem which we can use.
We should not see only the difficulties, however, but also the opportunities. If we can integrate statistics and research methods into PBL, there may be great advantages. One consequence of integration would be that the subject would not be marginalised or seen as separate. It would be just one aspect of medicine. Indeed, one of the features of a PBL course is that the distinctions between the different subjects should blur.
A second advantage would be that students would be learning statistics and research methods in the contexts in which we hope that they will apply them: the interpretation of clinical data and the assessment of research evidence. The relevance of the subject should be very clear and during their professional careers they would be more likely to be able to recall and use this material when needed.
So how can we do it? There must be an advocate for statistics and research methods at the start of the preparations for PBL, who is able to argue convincingly for the inclusion of these subjects. They must form part of the matrix of topics and problems. This is not an easy task and I have been told of great difficulty experienced at this level, as other teachers sought to exclude these subjects so that their own discipline could have more time. This is a natural human reaction, of course. I think that statistics and research methods are central to medical education, and would argue for more of them, at the expense of some anatomy if necessary.
We must get away from the idea that the problem must be a patient vignette. There are some statistical topics which can be covered quite conveniently in this way, such as measurement error, coefficients of variation, reference intervals, sensitivity and specificity, etc. After all, when my GP looks at my serum cholesterol on his practice computer, it has by the side of it a 95% reference interval. However, problems could equally be a published paper. We use these routinely in the teaching of statistics, research methods, and critical appraisal.
We could use a paper as a trigger for non-research methods topics. For example, a paper on an asthma trial could trigger questions about asthma as well as about randomisation. This may lead to too many questions being raised by the trigger. Another possibility would be to have such a trigger immediately following a case vignette problem on the disease in question. The patient vignette would raise the questions on anatomy, biochemistry, pharmacology, etc., associated with the disease. The following problem using a research paper would then raise only the research methodological questions. The fixed resource sessions in this week would then be devoted to these.
We may also, for variety, link research publications to the patient vignette by devices such as newspaper articles which the patient presents to the physician (e.g. reporting a trial), or say that in a patient problem the clinician has already found a Cochrane review.
Fixed resource sessions could include lectures, but I am very reluctant to suggest formal lectures in statistics and research methods for medical students. I would prefer to offer open question and answer sessions, where students can ask the statistician lecturer to explain anything they are unsure of. For example, if the PBL trigger is a randomised controlled clinical trial with results presented in terms of P values, we might be expecting this trigger to lead to questions such as 'why randomize?', 'what does P < 0.05 mean?', and 'should patients be told they are in a trial?'. Students should have made some attempt to answer these before the fixed resource session.