ABSTRACT
In all the previous chapters we have been concerned with how statistical concepts, principles and analysis may be applied to a great variety of problems in clinical medicine, both in practice and research. It is, however, still true to say that in clinical practice the majority of the inferences and decisions made are not processed through any prescriptive statistical system but are in fact the intuitive or ‘reasoned’ judgements of clinicians, radiologists, steroid chemists, laboratory technicians, etc. In cases where there are prescriptive statistical methods such as in diagnosis, where there may be eventually a true assessment for a case, it is of interest to ask to what extent the clinician’s judgement diverges from the normative assessment. This is an extension of the idea of observer error studied in Chapter 6. There we were concerned with the quality of the observation; here we are concerned with the much more complex problem of comparing inferences or decisions. Such studies are currently popular in clinical medicine, and we shall examine a number of studies of different types in diagnosis, prognosis, in calibration and assay and in treatment allocation. The general structure of such performance analysis is discussed in Section 11.2. We have found that such analyses have a considerable educational impact and comparisons of different groups of subjects can be very illuminating.
