Hi,

 

I think a good monitoring of your analyzer's performance is looking at your analytes SD from month to month, and it should stay constant. If it changes that could be a sign there is a problem. My question is what is a criteria is used to say there was a change and needs investigation? 10%, 20%, 30% differenence? Or it doubled? What is a good guideline to watch for?

 

Thank you!

You should really be following Westguard Rules for monitoring trending.

 

Have a look at the Westguard Rules website

 

https://www.westgard.com/westgard-rules.htm

In general I would suppose that a particular analyte's CV% (SD*100/mean) on a particular analyzer should remain constant.  (Note that a SD will vary by QC level, but in general, CV% will only vary by analyte.)

 

So for example, if the CV for a RBC normally runs at 5%, and you start having more than usual random outliers, you might want to check the CV%.  If the QC mean is OK, then no Westgard-style accuracy drift is present.  But if the CV% is running high, the analyzer is probably to blame for an increase in random error.

 

Scott

You should really be following Westguard Rules for monitoring trending.

 

Have a look at the Westguard Rules website

 

https://www.westgard.com/westgard-rules.htm

 

[This is intended to be humorous] People from the UK need to add u's or e's to everything, even proper names!

[This is intended to be humorous] People from the UK need to add u's or e's to everything, even proper names!

 

OMG the mustached one will be unhappy I have added a U to his name.

 

And pst - we don't add them, they were always there. It's lazy ass Americans dropping the letters Mind you nothing is as bad as the Southern English teeage 'innit' for lazy talk. 

 

ETA - yes there is! 'Bae' grrrrr

Edited October 26, 20159 yr by Auntie-D

When we had the Advia it would produce a report I think once a week or every so many patient samples.. I think it was called Neut X and Neut Y.  I think coulter has something similar Xbar B or something like that.  Our sysmex has it too but its not configured.  It was handy on the Advia as it would give advanced warning of impending doom.  I look at the controls once a week at the L-J chart. Pretty easy to notice trends that way in my opinion although the sysmex has had very few problems with QC (knocking on wood)

When we had the Advia it would produce a report I think once a week or every so many patient samples.. I think it was called Neut X and Neut Y.  I think coulter has something similar Xbar B or something like that.  

 

It's the moving averages on the advia and does all basic parameters - usually it is set to compare the last 30 results. Any deviation from the mean is flagged - great fun when you put a rack of oncology samples on

 

Ours is set for

 

RBC

MCV

MCH

MCHC

WBC cor - the correction between the perox and baso cell counts

Hb cor -  the correction between the measured and the calculated Hb

MHC cor - the correction between the MCH and CHCM

 

The cut off is a values of below 0.95 or above 1.05 (5%) but people seem to accept anything between 0.8 and 1.2 *sighs*

 

Not sure why there is no platelet moving average though...

 

Neut x/y is the average point of your perox scatter plot - it gives you an idea of the shape of the scatter without having to look at the plot.

Edited November 2, 20159 yr by Auntie-D

Back in the dim times of hematology analysis, there was no such thing as control material for CBC counts.  Even today, lot numbers are only stable for about 6 weeks or so.  So analyzers were equipped with a ridiculously complex statistical analysis package that essentially involves tracking a moving average of parameters.  The idea is that for large batches of patients, certain CBC parameters should not vary much over time. 

 

The XB moving averages monitor RBC, MCV and Hgb only (leaving our distant Lab ancestors to worry over WBC and Plt counts I guess). However, the availability of assayed controls has made dependence on XB as a primary QC method unnecessary.  We still use it for troubleshooting sometime, though.

 

Scott