Clinical Chemistry, Vol 36, 1629-1632, Copyright © 1990 by American Association for Clinical Chemistry

Precision requirements for cost-effective operation of analytical processes

JO Westgard and RW Burnett
Department of Pathology, University of Wisconsin, Madison 53792.

Current performance criteria for analytical methods are often based on recommendations developed many years ago. A common criterion for imprecision requires that two times the standard deviation (s) of the method be less than the allowable total error (TEa). Unfortunately, when this criterion is minimally satisfied, commonly used control procedures cannot achieve reliable detection of medically important errors. Studies of the power functions for statistical quality-control (QC) procedures show that the magnitude of medically important errors must be at least 3s to fall near the plateau of the power curves and be readily detected by current QC procedures. For methods that just meet the precision criterion 2s less than TEa, however, medically important errors will fall on the rising portion of the power curves and typically be detected less than half of the time. From a "reverse engineering" perspective, the 2s less than TEa criterion is inadequate because it does not allow for the known performance limitations (lack of sensitivity) of commonly used control procedures. A strong case can be made for using a criterion of at least '4s less than TEa, which calls for a twofold improvement in imprecision over, the current minimum requirements. This recommendation is consistent with current industrial guidelines for process capability and would lead to more reliable detection of medically important errors.

The following articles in journals at HighWire Press have cited this article:

				G. S. Cembrowski and R. N. Carey Adding Value to Proficiency Testing Programs Clin. Chem., January 1, 2000; 46(1): 7 - 8. [Full Text] [PDF]

				D. Chesher and L. Burnett Equivalence of Critical Error Calculations and Process Capability Index Cpk Clin. Chem., June 1, 1997; 43(6): 1100 - 1101. [Full Text] [PDF]