In the first study we showed that students’ diagnostic accuracy for the 16 “important” skin lesions was only 33% after completing their dermatology attachment and that it dropped to 24% a year later. Such absolute measures of diagnostic competence are obviously incomplete and potentially misleading because they are so dependent on the difficulty of the specific tests. However, in the absence of normative data elsewhere in the literature, they provide some sort of benchmark for future work, and other authors are welcome to use our test-sets.
A potential reason for what we take to be the students’ poor accuracy was identified in our second study, where we showed that students’ exposure to the 16 important skin lesions is highly variable but universally limited. Some limitation in clinical exposure is to be expected but, rightly or wrongly, we were shocked by how limited the extent of clinical exposure was. Although there is no systematic data on this issue we suspect that the clinical exposure our students gain is typical of, if not better than that at many other UK centres, and our course duration is actually greater than the UK average [12
]. Of note our course scores highly in terms of student feedback compared with other clinical attachments within the University of Edinburgh. In addition, due to the structure of our group demonstration clinics (which allows multiple students to see a single patient) the overall melanoma witnessing rate of 38% was predominately achieved because of a single case -if this one patient had not presented and agreed to be examined at the time of a demonstration clinic the overall melanoma observation rate would have dropped to 18%.
The students’ low exposure demonstrates the major difficulty encountered when teaching students to identify skin cancers and their mimics — a lack of reliable clinical examples. Unlike other specialties where pertinent clinical signs are often longstanding and can be demonstrated repeatedly to different groups of students, suspected skin malignancies are excised in dermatology outpatients as a matter of priority. Therefore if face-to-face patient teaching is to be relied on a constant throughput of new patients is required.
It might be argued that “skin cancer” education is not performed exclusively during students’ dermatology attachments but that additional teaching is happening during their allied clinical attachments. We are sceptical of this and note that our students’ scores, which were lower one year after their dermatology training, followed their 2-month attachment in primary care.
The finding that there appeared a mismatch between objective ability and self-confidence again should not be surprising. Although subjective measures are in widespread use for assessing teaching and learning, there is increasing acknowledgment that students’ (and other health professionals’) self-assessment of their own abilities are often erroneous [36
]. An additional reason for the students’ overconfidence could also be that the lesions encountered during their teaching attachment were not suitably representative of the variety of presentations that had been randomly selected in our test slide shows (and could be argued are encountered in real-life). Neither the lesions that the students witness in the group demonstration clinics nor those that they are exposed to in their additional study are unselected. In our experience both textbook images and the demonstration clinic lesions are often chosen precisely because they are “classical” examples. Irrespective of the cause, the mismatch in confidence and competence at the end of students’ only formal dermatology training has implications on the role of non-experts as gatekeepers to cutaneous cancer services.
Given the potential importance of undergraduate dermatology education it is surprising that there has been so little investigation of UK medical school dermatology teaching [12
]. Whilst additional studies have subjectively shown that both medical students and primary care physicians feel dermatology training could be improved [44
], there are only a few US studies that have objectively assessed students’ diagnostic acumen after undergraduate dermatology electives [47
]. These studies are, however, unlikely to be transferable to UK undergraduate students, and are not directly comparable to our study because the difficulty of the questions was not controlled for. Whilst the US studies demonstrated a higher level of improvement in diagnostic competence than we witnessed (71-82% accuracy), there are a number of other potential reasons for this. First, in these studies there could be an element of selection bias; the attachments were not compulsory and the students enrolled had volunteered to undertake dermatology electives, many presumably with the intention of pursuing a dermatology career. Second, the studies used the same images for both the initial and final assessments, which are likely to have artificially raised the final scores. More importantly the test contained not just lesions but also dermatoses and the images used for testing were not randomly selected, instead, the images chosen were often described as “classical” examples. Third, the answers were in multiple choice format rather than free text, which does not correspond to everyday practice. Finally these optional electives were of longer duration than in the UK, being full-time for 2–4
We believe the biggest difficulty in interpreting our work is the lack of a coherent and justified framework for the purpose, standards and remit of undergraduate dermatology education. Although the BAD have produced guidelines that have been integrated into UK medical schools’ curricula, there are no specific criteria for any of their recommendations. As a result the current statements, such as “graduates should be able to recognize melanomas”, are as lacking in precision as saying a mathematics graduate should be able to compute. If the guidelines are to be adopted successfully, these criteria need to be objectively defined. This requires that we have to tackle exactly how difficult particular cases are with reference to some common standard.
So how could students’ skin lesion identification be improved? Our own students clearly seek out online resources, and the number of images available online far exceeds that available in the clinic or in textbooks. Online content is, however, very variable in quality and many online images are of poor quality and not infrequently in our experience the diagnosis is questionable, if not wrong. Whilst we know of no experimental work directly comparing learning on real patients versus learning from image databases for skin cancer we strongly suspect that the latter is the way to proceed. Modern imaging techniques already allow for accurate 3D models of skin lesions to be captured simply and efficiently [50
]. These models have significant advantages over conventional 2D photography as they allow students to rotate and pan around the images as they would in real-life. A large database of such images could address the main obstacle to effective dermatology teaching by removing our reliance on a constant supply of new example lesions. Furthermore if the images were available online, widespread access could be achieved with relatively low production costs. This would allow standardized exposure across multiple institutions, limiting the intrinsic fluctuations of exposure that occur between students with present teaching arrangements.