Use of patient data for quality control of clinical laboratory tests.

The average.of.normals quality-control method is examined mathematically. Two modifications are proposed. After modification, the essential unchanged feature of the method is the idea of censoring the laboratory test results so that only means of test values inside the normal range are plotted on the quality-control chart. Several limitations of the modified method are discussed, and I conclude that, although censoring does improve the probability of detecting drift, more statistical research and further modifications are needed before the average-of-normals method will be useful for quality control in the clinical laboratory.