> Subject: QC rules and modern analysers
> Date: 19 June 1998 10:42
>
> I have had my arm thoroughly twisted to follow up my last contribution to
> the debate on QC rules, so here goes.
>
> Existing rule-based QC criteria such as 'action limits' or Westgard rules
> are designed to detect unacceptable changes in bias which can occur on
> calibration or reagent changes. They are appropriate for 'batch assays'
> particularly those requiring calibration for each batch.
> The two problems with them are:
> 1. If action limits are tightened, or more complex rule-based systems
such
> as Westgard's are used, there is an increased probability of detecting a
> bias change, but also an increased possibility of registering a
significant
> change in bias when this has not occurred (ie more false positives and an
> increased rate of unnecessary batch repeats).
> 2. The criteria which are used are based almost entirely on technical
> achievability, not clinical utility.
>
> Modern analysers do not require frequent calibration or reagent changes,
> and usually do not operate in batch mode. The QC problems are not
primarily
> about sudden stepwise changes in bias. They are about very slow changes
in
> bias with time (drift), about periods of increased imprecision associated
> with slightly less than optimal analyser performance, and about
occasional
> fliers caused by electronic or other glitches. It is comparatively easy
to
> show that application of conventional QC rules to these problems results
in
> a very considerable amount of unnecessary re-runs, and actually worsens
the
> quality of the results from patient samples.
>
> More appropriate QC approaches to modern analysers are
> 1. Trend detection applied to drift in bias. Action is required when bias
> increases to a level set by clinical utility requirements, not analytical
> achievability.
> 2. Continuous monitoring of assay imprecision to detect sub-standard
> performance. I have up my sleeve a method for doing this which requires
> neither the use of replicate QC sera nor duplicate patient samples (of
> course, these can be used, but at the cost of carrying out analyses which
> are otherwise un-necessary).
> 3. Continuous checking for fliers. Apart from the clinically ridiculous
> results, I believe this can only be done by comparing results with
previous
> results on that patient (so called delta checks, which aren't always
useful
> particularly in rapidly changing patients such as those on renal
dialysis);
> or by comparing results with clinical information on that patient - this
is
> an onerous job for a duty biochemist which requires great skill and is
far
> from perfect, particularly when clinical information is poor or
potentially
> misleading.
>
> I rest my case!
> Gordon Challand
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|