Analytical quality, performance indices and laboratory service.

Faulty data lead to suboptimal diagnostics and decision making unless the flaw is known and amenable to correction. However, pure noise (added analytical variance) has only minor effects on clinically appropriate indices of diagnostic performance. This fact is illustrated by an idealized screening programme, using the preventive benefit-to-cost ratio as the index of performance. Further, this article illustrates that the effect of an unnoticed source of noise is roughly just twice the effect that the same noise will have when its magnitude is known and clinical decision limits are adjusted accordingly. Owing to the small size of these effects, however, it may be profitable to spend resources on other aspects of good laboratory service, such as timeliness, documentation and interpretative support.