How much quality control is enough? A cost-effectiveness model for clinical laboratory quality control procedures (illustrated by its application to a ligand-assay-based screening program).

Quality assurance testing represents a substantial proportion of the clinical laboratory budget, but current guidelines are based on criteria that pertain to analytic error rather than to optimization of the cost-effectiveness of patient care. A general Bayesian mathematical model for the cost-effectiveness of assay quality control has been developed, and is demonstrated using previously published data. The cost-effectiveness of quality assurance as defined here depends upon the prevalence of disease, the shapes of the distributions of test results observed in the non-diseased and diseased populations, the decision limit selected for labeling results positive or negative, the costs and benefits associated with each of the possible therapeutic outcomes, the magnitude of random and systematic analytical errors, the statistical power of the quality control test in use, the costs associated with delays due to re-assay, and the proportion of total test cost attributable to quality control procedures. Given current clinical laboratory practice, much of this information will not be routinely available. The model combines these factors into a simple equation with three terms: one for the cost of the original and any required repeat laboratory analyses, one for the cost of delay entailed by the rejection of an assay batch, and one for the change in total costs consequent to rejection of erroneous assay results.