Difficult Inspection Level/AQL - Verifying it is the correct part

[Doctor Pig]

I typically receive a lot of 500 pieces. These are finished medical devices and I'm only verifying the correct part. ( these were inspected at another facility prior to packaging) These are expensive titanium devices and have been sterilized. Currently and in past lots, no rejects have been noted. The previous inspector always verified 2 pieces. I need to set an inspection level and AQL to match the 2 samples. I came up with S-1 AQL 6.5 (ANSI-ASQ Z1.4-2003) If I found one reject, I'm SOL. I can't 100% inspect due to the sterlization process. Is there a way to up the reject criteria without increasing the amount of 2 samples?

[Coury Ferguson]

Anyone in the Medical device field have any suggestions?

[Tim Folkerts]

Could you clarify for me what you want?

It sounds like you have been inspecting 2 parts, and want to keep inspecting two parts. And you want to "up the rejection criteria", which makes it sound like you want to reject lots only if both are bad. Is that accurate?

If so, one possibility would be to try a multiple sampling process. I don't have a copy of the standard handy, but you might be able to find a plan like:

Draw 2 items

• 0 bad --> accept
• 1 bad --> inspect 2 more parts
  
• 2 bad --> reject

This way, when you find no defects (as you apparently always have), then you accept the lot. On the rare occasion when you find one bad piece, you check further.


I do wonder if I am mis-interpreting your needs. You say these are expensive parts. It would seem that you should expect to have only correct parts. Why would you want to accept a shipment of expensive parts where you know some are wrong?

Tim F

[Doctor Pig]

The parts are manufactured, assembled, inspected and sterilized outside the US by another division. I receive the sterilized finished product. I don't want or need to open more than 2 sterilized pieces. I'm trying to match an inspection level and an AQL that will allow me to sample only two pieces of product. If I use S-1 AQL 6.5 this will allow me to verify two parts. However, if I find one reject out of the two, it will force me to do a 100% inspection. 100% inspection could not occur with this product. If I opened all of the sterilized product, there would be no product. What I looking for is an inspection level and AQL that would allow 2 samples with one acceptable reject. I doubt if I were to ever see a reject. I'm mearly verifying that the two samples (from each part/lot) are the correct parts (not just by looking at the labeling) All of the parts look the same. Example: part number xxx has a 3mm thread length and part number yyy is a 4mm thread length. These would be different part/lot numbers. Does that make sense of what I'm looking for. Again this is only for verification of parts.

[Tim Folkerts]

The whole situation sounds backwards to me -- adjusting your plan to match current practice, rather than adjusting your current practice to to match a desired plan.

But then I come from an academic background and I tend to think from the fundamental statistics. In a regulated industry, i know it can be important to be able to state that you are following an accepted standard.

Originally Posted by Doctor Pig

...However, if I find one reject out of the two, it will force me to do a 100% inspection. 100% inspection could not occur with this product. ...
...I doubt if I were to ever see a reject. I'm merely verifying that the two samples (from each part/lot) are the correct parts (not just by looking at the labeling).

I think you need to take a step backwards and decide what sort of risks you are will to take vs what sort of cost you will incur.

For example, with a sample size of two (using the binomial distribution), a lot with 78% defective will still be show no defects 5% of the time. So even at AC=0, you will accept some pretty bad lots 5% of the time.

If you bump the criterion to AC=1, you could accept a lot with 97% defective 5% of the time! A lot that is 50% defective will be accepted 75% of the time! Are you will willing to accept a lot that is half defective most of the time?

If these risks seem excessive, then you might want to sample more items. Or at least keep AC=0. On the other hand, if your experience is that either a lot is good or it is bad (for example a whole lot of 4 mm threads was mis-labeled as 3 mm), then your plan should be fine. Of course, in this case you might as well only sample one item!

It comes down to economics

• how much does it cost to test an item?
• how much does it cost to pass on bad items to your users?
• what sort of lots are likely to be produced (mixed? all bad? a few occasional bad parts)

That is how I would approach it (but as I said, I am more of a theoretically-based thinker who doesn't need to satisfy regulators).


Tim

[Doctor Pig]

Tim:

Thanks for the info. I guess I should look at it as, if one is rejected because a size difference, I would in fact, reject the entire lot. If one is off, there would be a very high chance there are more than one mislabeled. I recieve an inspection sheet from my source. They inspect 13 of the 500. I'm not sure what level they are inspecting to. Possibly S-4, 1.0 I don't think I need to "up" my reject criteria. I just needed a head slap to see what I needed.