Posted by Sten Westgard, MS
A recent abstract from the 2009 IFCC/EFCC (Euromedlab) caught my eye:
Quality Indicators for Laboratory Process; assessment in the Trento Department of Laboratory Medicine.
I Caola, C Pellegrini, N Bergamo, E Saurini, P Caciagli.
CCLM 2009
Examining the quality records of five laboratories, they applied a set of Quality Indicators from the IFCC. Errors were tracked and tabulated. The results are quite interesting.
Here are the results of the indicators they were able to track:
% Error
Quality Indicator | # Errors | # opportunities | % Error |
QI 8: Samples transported btw. labs and not recvd | 15 | 140,000 | 0.01% |
QI 9: Inappropriate container | 10 | 1,038,002 | 0.001% |
QI 10: Samples hemolyzed (chemistry) | 15,561 | 1,454,541 | 3.5% |
QI 11: Samples clotted | 232 | 1,038,002 | 0.022% |
QI 12: Insufficient sample volume | 128 | 1,038,002 | 0.012% |
QI 13: Inadequate sample- anticoagulant ratio | 215 | 1,038,022 | 0.021% |
QI 14: Damage in transport | 39 | 1,038,002 | 0.004% |
QI 15: Improper labelling | 50 | 1,038,002 | 0.005% |
QI 16: Improper storage | 69 | 1,038,002 | 0.007% |
QI 17: Unacceptable EQA performance | 29 | 2,519 | 1.15% |
Now, let's translate those error rates into Sigma metrics. Here's the graphic we get (click on it for a close-up):
With the exception of the hemolyzed chemistry samples rate, all the other pre-analytical quality indicators are at 5 sigma and higher. That's good news. The bad news is that the EQA performance is 3.8, almost as bad as the problem of the hemolyzed samples (3.4 Sigma).
It's almost a given that everyone considers analytical error rates to be much lower than pre-analytical rates. This study says that might not be true. If you were to add up all the other pre-analytical quality indicators and sum their defects (admittedly a crude way to measure), you'd have a 3.7 Sigma for pre-analytical performance. That's nearly equivalent to the EQA error rate.
[We note that the EQA performance rate is an imperfect measure of analytical error. The study authors note that they were unable to measure an actual analytical quality indicator, QI 18, which would have measured how many times the CV was higher than the selected target for the test. That's another problem: how do we know the analytical error rate when most labs are unable to define a quality target and track deviations from it?]
One of the big differences between pre-analytical errors and analytical errors is visibility. When a sample goes missing, no matter how rare the occurrence, it's a very obvious error. In contrast, analytical error can be insidious; it can accumulate over time. You might not recognize the error for a while, even as the variation or bias impacts a huge number of results.
We assume that because analytical errors are hard to detect, they must not be occurring. That's a big assumption. Pre-analytical errors are easier to detect, and might be easier to correct. But when the error rates show otherwise, we should pay attention to the analytical issues at the core of our profession.
Comments