Sample Size Determination - Medical Devices

Ok... another newbee question

I recently started working for a new company which product is a medical device, for the time-being, limited to software. We are preparing for ISO 13485 audit.

My question:
The manufacturing process for the software has not been decided completely yet (outsourced or not). Manufacturing process, for us, "simply" imply burning cd, label & pack... I am doing a procedure for purchased products verification but I am having trouble defining stats variables. I do understand (or think I am) the AQL but I cannot understand how you can determine the sample size regardless of the population size (or size of the lot). Doesn't the sample size should vary with the size of the lot and the quality assured by the manufacturer?

Thanks for your help...

Re: Sample Size Determination

Karey13,

requirement for ISO/TS 16949 may vary from the standard you are auditing to but here is my thoughts. First, AQL sampling is a good start. However you need to take into consideration other factors when creating a sampling plan. Some considerations are:

Past performance from the supplier

The importance of the part being inspected (a blank jewel case may need little inspection but the printed jacket may require mor inspection)

What types of inspections are being done a the supplier. (sometimes the review of there data may suffice. We have many suppliers taht are dock to stock and onoly send the process results(SPC data)).

I would suggest you base your initial sampling plan on the importance of the component coming in and of the past performacne of the supplier. This should also be acceptable to the requiremetn in the standard you are using.

Re: Sample Size Determination

unless you have very small lots and categorical data (pass fail, counting data, also known as 'attributes' data) the sample size is nto dependent on the lot size. The sample size is dependent on the amount of actual variation the lot has - or that you are looking to detect - and the amount of 'confidence' you want in your inspection.


Are you working with categorical or continuous (measurable or 'variables') data?

I can attach a spreadsheet and a brief description of the formulas for categorical data if you'd like....

Re: Sample Size Determination

Hum... stats theory is a bit far-away I realise...

We'll produce medical devices (for now, only software) in relatively small quantity since the targeted user is not general public but particular classes of diagnosticians. It might be it: I don't see how I can have a sample size of 60 (for example, CDs or labels) if I produce 15 units per year...

For the current product, purchased products would either pass or fail (nothing to measure). Lot size will be max 100 (guessing) and we would not have that many lot per year (1 or 2?) hence wondering if it is even possible to have a complete sampling plan including different levels of inspection procedure like normal, tightened....

If you can provide me with guidelines, Bev D, I would really appreciate it!

Re: Sample Size Determination

To simplify things for everybody, is this approach suitable:

Using standard MIL-STD-105E (and eventually ASQ z1.4 when I receive it):

- Use an AQL of 1.0 which seems ok in the medical field
- Use single sampling (is this the easiest approach in terms of implementation and paper work? we do not have that many resources for now...)
- Determine the sample size from table I - Sample size code letters
- Perform tests
- Use table II-A - Single sampling plans for normal inspection to determine if lot is accepted or not
(if number of defect units <= acceptance nb = lot accepted;
if number of defect units > acceptance nb = lot rejected)

Since we are a small company, that we won't have regular shipping of the same products shipped (for now), I would be tempted not to implement switching of level of inspections since it seems to me that it will require more time and paperwork than the benefit we would gain from it... Can we do that? If so, would it cause a problem if a lot is rejected and manufacturer sends another one?