Observer variation in reports on electrocardiograms.

It is now well known that the results of almost any examination or investigation may be reported differently by different workers, for example, the interpretation of a medical history (Cochrane et al., 1951), the detection of cyanosis (Comroe and Botelhoe, 1947) or clubbing (Pyke, 1954), the clinical diagnosis ofemphysema (Fletcher, 1952), the radiological detection of tuberculosis (Birkelow et al., 1947; Garland, 1950), and the assessment of pneumoconiosis (Fletcher and Oldham, 1949). Disagreement may be due to true errors of observation where certain abnormalities are wrongly identified or even missed altogether. But in the investigation reported here, as in many others there is a second factor, the observation may be correct, but there is a difference of opinion about its interpretation. The term observer variation used in this paper is defined as including disagreements due to both causes, for they occur together and it is sometimes impossible to tell which is responsible. Because of this, observer variation is a better expression of the problem than the term observer error. The practical significance of observer variation has often been disputed (Lancet, 1954) and only in radiology is its importance being accepted. With the single exception of a report by Sloan et al. (1952) on the detection of the third heart sound, the problem of observer variation in cardiology has not been studied. This is surprising, for an electrocardiogram might be expected to be as difficult to interpret as is a chest film. They are of equal importance in that one is as valuable in the diagnosis of heart disease as is the other in the diagnosis of pulmonary disease; and both are single investigations upon which diagnosis and treatment may largely depend. Yet the assumption appears to be that unlike the chest film, electrocardiography is a test free from observer variation. The purpose of this paper has been to test that assumption, and it has proved to be false. Not only has observer variation been found, but its extent was surprising and it has been possible to identify many of its causes. The practical significance of this variation will be discussed in the belief that it may often contribute to diagnostic errors.