Observer variations in the diagnosis of stroke. WHO collaborative study on the control of stroke in the community.

Validity and consistency of the diagnosis of stroke and its types were tested with 60 case reports including 15 diguised duplicates, drawn at random from community-based stroke registers. Seven European, 5 Japanese, and 5 other centers participated the test. The diagnosis of stroke as such (regardless of its type) seems to have been established accurately in all the 3 groups of centers; this leads to the conclusion that the incidence rates of stroke as registered in the participating centers were comparable. The diagnosis of the types of stroke was less reliable, since intra-and inter-observer bias was found in the diagnosis of the identical sample of test cases. In some European centers, the type of stroke was rarely determined, unless objective and definitive evidences were available. The Japanese centers appeared to have diagnosed the type of stroke in a relatively more uniform way between centers, however, the consistency of the diagnosis at separate times was lower. Subarachnoid hemorrhage, when diagnosed, was generally based on firmer gounds. These observations were confirmed in a small number of autopsy-verified cases. These varying diagnostic attitudes introduce false differences. Such "softness" of type-diagnosis must be borne in mind when comparison of type of stroke at different times or between populations is attempted.