[Preventive screening in two factories. II. Exactitude of laboratory tests and questionnaire (author's transl)]

The exactitude of some widely used laboratory tests (triglycerids, cholesterol, glucose, uric acid, creatinine) is examined in a situation relevant for practising physicians. Different statistical methods for reporting errors are compared. Accuracy and precision are useful measures for the quality of analytic procedures. They are not sufficient for medical judgement for a single patient, since sampling procedures, type of sample or transport and storage of sample are not considered in accuracy and precision. Such sources of error can largely devaluate the quality of accuracy and precision and of the analytic procedures. Error coefficients are proposed as a new method of reporting laboratory errors. The mathematical model starts with an analysis of variance and total error, technical error and transport error are defined. The study shows that the error introduced by transporting serum via postal service compared to that conveyed in a cooled transport box by a special car is considerable. It is nearly impossible for a physican to use single laboratory value for a rational decision "normal-not normal" if the values are based on samples which are sent by postal service and if the values lie in the broad borderline between the normal and pathological area. This is especially true for creatinine (transport error 47.0 per cent) and uric acid (transport error 38.7 per cent), but not so much for cholesterol (22.9 per cent), triglicerids (14.3 per cent) and glucose (13.3 per cent). Variables with high transport error like creatinine should not be used in screening programs, as long as the transport is made by postal service and the method is not improved. The agreements between the answers of a questionnaire and of an interview are analysed with the same patients (n=235). These agreements lie between 98.7 per cent and 61.7 per cent depending on the single question. The exactitude of medical questionnaires is in the same size order as the exactitude of laboratory tests-at least when the sample is sent by postal service.