Inter-observer variation on registration of signals of cancer

Previous studies have dealt with the role of seven warning signals of cancer in public health education and cancer diagnostics in general practice. Registrations of warning signals from medical records have some inconveniences common to most retrospective record studies. In the present inter-observer variation study it is found that agreement is good for warning signals in cancer patients when the warning signal has a probable connection with the disease. In control patients, and for more incidentally occurring warning signals in cancer patients, agreement is not so good. It seems that reproducible registrations from records are possible when there is a precise definition of what to look for and the data are relevant to the patient's disease. This study stresses the importance of good record keeping in medical practice.