Evaluation of Analytical Performance Specifications for Clinical Laboratory Tests Based on Test Misclassification

Analytical performance specifications for laboratory tests are essential components of quality assurance for clinical laboratories. A widely adopted performance criterion is total allowable error (TEa), which includes contributions from both bias and imprecision, which may contribute to test inaccuracies. However, the relationship of TEa to test result misclassification, a clinically relevant quality measure, is unclear.

Hypothetical clinical laboratory test results for 4 test models were generated and subjected to proportional bias and imprecision. Test misclassification (TM) as a function of bias and imprecision was determined using pre-defined cutpoint(s). Simulation analyses were then performed on 14 chemistry analytes using 3 data sets from patient results reported at the National Institutes of Health (NIH).

We observed a complex and nonlinear relationship between bias and imprecision, and their impact on TM was not additive as may have been expected with TEa. TM scores were influenced by population distribution, the location of cutpoints, and the fraction of abnormal test values at baseline. Stringent TEa requirements did not correspond to low TM scores. On the contrary, TM scores for electrolytes were among the highest. Instead, TM scores closely correlated with the ratio of TEa to the width of the population distribution.

TM has the advantage of correctly accounting for the differential effects of bias and imprecision. Compared to TEa, it provides a more accessible metric for evaluating performance and the clinical impact of errors.