RQL for Validation
- This topic is empty.
- August 23, 2018 at 5:56 am #56078
I’m pretty sure this is an old hat for you, but nevertheless I’m stuck and need help on this:
For validation in medical devices it is not recommended to use AQL based sampling plans. I understand this, as it does not look at the crucial thing – that is: the consumer risk. So I shall go for RQL. Fine with me. However, I have to justify my choice of RQLs and confidence levels. And my simple question is: how do I get this justification?
In my company we have a list for defect criticality vs. AQL (minor defects: AQL 4.0 and so on) but how do I get an information what RQL would be okay?
I guess this should be also depend on criticality of the defect, but if I’m referring to our risk acceptability matrix (severity vs. likelihood of occurrence) we start at less than 1:15.000 devices for the negligible things. That would mean I’ll never accept something higher than 0,01% even for the most minor defects. That would lead to ridiculously low RQLs! Or is it wrong to try to translate our risk acceptability to RQLs?
What kind of RQL values do you use and how do you justify?
Grateful for any help :-(
Sitoo0August 27, 2018 at 10:25 am #202986
Since the RQL is the minimum defect rate you want to detect XX% of the time (if it is present) the RQL would be whatever YOU (your organization decides is not tolerable). You know your Customer base, you know the severity of the effect of the defect(s) etc. that is your justification.
Of course the issue here is that these rates are often very low although higher than an AQL and so sample sizes get quite large. This is the inherent difficulty with categorical data. The real solution is to move to continuous data and directed testing under worst case conditions to test for margin.0August 30, 2018 at 6:33 am #202996
thanks for your advice.
Unluckily, I actually was referring to continuous data and sampling plans as for example in ISO 3951-1: Sampling inspection by variables or ANSI/ASQ Z1.9 Sampling procedures and tables for inspection by variables.
But still I’m not sure how I can derive, which RQL would be adequate.
Sorry, if I’m too slow…. but again, I would appreciate some help.0
You must be logged in to reply to this topic.