I would be very concerned about using mobile control limits in this way.
1. You are using control limits to assess whether the new control point you
have obtained is in the same population as previous ones. The technique
(usually within +/- 2 SD) is statistically quite weak, and is unlikely to
detect small but analytically or clinically significant changes. You cannot
assume that the qc point is in the same population as previous ones, even
if it is within these qc limits.
2. Because of this, the use of mobile control limits leads to a potential
major analytical problem of undetected bias drift. I worked in a previous
hospital, where over a three year period, serum amylase values on a patient
population had doubled. This was not detected using conventional internal
QC because the bias changes each month appeared insignificant, and minor
changes to qc limits were made each month. All the qcs remained within
limits over the whole period of this assay drift.
Gordon Challand
----------
> From: Philippe MARQUIS <[log in to unmask]>
> To: [log in to unmask]
> Subject: QC programme
> Date: 15 September 1998 12:52
>
> Conversely, it is common statistical practice to get best
> estimates in a population using the largest available sample.
> When I validate a QC point, I assume it is in the same
> population as the previous ones. So why not re-calculate the control
> limits every day with an additional degree of freedom? It
> is a technique I call mobile control limits.
>
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|