Repeating a refuted item doesn’t make it true

November 6, 2008

mistakeWith the election just over,  I am reminded that politicians will repeat things that have been refuted. What reminded me was an article in Clinical Laboratory News about risk management. Here is such a quote from the article:

“According to Yost, many labs have implemented EQC successfully with no other quality problems identified during inspections. She also commented that with unit-use devices, doing QC on one cartridge doesn’t really tell you much about the performance of the rest of the cartridges, anyway. “You’re just using up a cartridge, whereas internal  controls can tell you a whole lot more about the test system,” she said.”

 

This quote actually has two problems. I have commented before that because no problems have been identified during inspections does not justify that EQC is a good thing to do. I said that if you take the airbags out of a car, and there are no crashes, the safety results of the car will be fine. This does not justify the removal of airbags.

 

But I am more interested in the second problem which implies that QC is useless for unit use devices, because a unit use device is used up during the test. This is also in the latest draft of EP23, which I won’t quote, since it is a draft.

 

The problem is this statement implies that all problems are random occurrences. This is simply not true. When problems are random by the way, the same problem with QC applies to non unit use devices. That is, the failed sample does not inform about other samples.

 

But many problems are not random but occur systematically for each sample until either the problem goes away or is identified and the system is stopped. This is shown in the following figure for both unit use and non unit use devices.

 

unitusefailure

 

In this figure in the unit use case, the factory is making thousands of unit use devices, but there is a problem (shown in red). All of the unit use devices are bad. Of course, the factory has its own QC procedures to guarantee that bad units are not released but this QC procedure also fails leading to thousands of bad units being shipped.

 

That QC can fail at the factory is a point that most people have problems with. Well, it happens.

 

The clinical laboratory performs QC on the system. QC detects this systematic problem and no bad  results are reported until the problem is solved (e.g., another lot of unit use devices is used).

 

In the non unit use case, a similar situation occurs. A bad reagent is shipped to a clinical laboratory, which detects this by running QC.

 


Historical vs. New Mitigations in a FMEA

November 1, 2008

One of the things that I often encounter when looking at FMEAs is failure to distinguish between what’s already in place and what is new because of the FMEA activity (1).

Consider an example. In a clinical laboratory it is standard practice to check each sample for hemolysis (because hemolyzed samples give misleading results for some analytes such as potassium).

A FMEA table might list a cause for an error, “hemolysis in the sample”, and the mitigation as “check the adequacy of the sample.” There is nothing wrong with this FMEA but one should be aware that nothing new has been added. This check in the process is already in place. Thus, it would be helpful in the FMEA to distinguish between items that are already in place and new items.

If every mitigation in the FMEA is already in place, then there is cause for alarm regarding the FMEA. To continue with the example, one could postulate a new error; that the check for adequacy of the sample fails meaning that the sample is hemolyzed but the check to detect it failed – causing the same error of misleading results. Perhaps, a cause for this error is postulated as “non cognitive error” meaning that for any task, a “slip” may occur. The mitigation suggested is a double check. There is another possibility – that after considering the severity and probability and perhaps other factors – the mitigation column for this error might read – “no mitigation required.”

This FMEA now thus considered new items – not just what is already in place. Note that all new, proposed mitigations will not be implemented as in , “no mitigation required” but they have been considered.

In summary, an important step in conducting a FMEA is the historical listing of mitigations. If the FMEA stops at this stage, it is not useful.

References

  1. Ten tips to improve risk assessment. Krouwer JS. IVD Technology 2008; 14;16-23. See http://www.devicelink.com/ivdt/archive/08/03/010.html