ISO 17025:2017 Clause 7.1.3 - Decision Rule

[Bran]

I recently went through a lab assessment and had a deficiency against clause 7.1.3 surrounding communication of a decision rule to customers. The clause states:

When a customer requests a statement of conformity to a specification or standard for the test or calibration (e.g. pass/fail, in-tolerance/out-of-tolerance), the specification or standard and the decision rule shall be clearly defined. Unless inherent in the requested specification or standard, the decision rule selected shall be communicated to, and agreed with, the customer.

The lab is testing to methods that have decision rules "inherent" in them, and I make a statement about this in our lab manual. This was something that was confirmed during the assessment.

Separately, the lab also has a set of guard band limits to which it will typically pass/fail a test result. In all cases, the guard band limits are set within the limits identified by the method or standard. Because we are pass/failing results based on the guard band limits, the assessor gave us the deficiency under the reasoning that we are no longer using the limits (decision rule) that is inherent in the method, and therefore we must define those limits to, and gain agreement from the customer per clause 7.1.3

Wanted to get some different thoughts surrounding if this deficiency holds any weight? I'm having a difficult time with the interpretation that we need to disclose the use of guard band limits to customers, when all that we are claiming to comply with is the method.

 

[Benjamin Weber]

Maybe you can give us more information on the "inherent decision rules" of your test methods? Do those methods specify measurement uncertainties and/or tolerances?

Here is our situation that might brinng some clarity: We perform safety tests for electrical devices. The standards specify the test methods and they have an "inherent" decision rule. Usually a measures value is compared to specific limit value (minimum or maximum allowd value). If the measured value is above/below the limit the test is passed/failed. There are no tolerances specified in the standard and no maximum allowed measurement uncertainty is specified.

But ISO 17025 requires us to define our own decision rule, i.e. how do we take measurement uncertainty (MU) into account? There are several possibilities, and we have to inform and agree with the client on our own decision rule.

Example: The maximum allowed surface temperature is 40,0°C. We measure a maximum value of 39,0°C. Is the test passed or failed? The answer is: It depends on our decision rule and how we take MU into account.

Let's assume the MU is +/-1,5°C (actually it is the expanded MU with >95% confidence). The limit value is now within a band of 38,5°C to 41,5°C were the pass/fail decison is not any more within the confidence interval of >95%. We could report "undetermined" (often seen in calibration reports). But our clients usually expect a "pass" or "fail". So we decided to apply the following (also known as "shared risk"):

• Measured value is out of the undetermined band of "limit value +/- EMU" -> Clear pass/fail statement ("P" or "F")
• Measured value is within the undetermined band -> Limited pass/fail statement ("#P" or "#F")

In the above example we would report "#P", measning that the measure value is nominally below the maximum allowed value, but there is certain chance of also being above (due to our EMU).

Maybe this what the assessor requires?

 

[dwperron]

I recently went through a lab assessment and had a deficiency against clause 7.1.3 surrounding communication of a decision rule to customers. The clause states:


The lab is testing to methods that have decision rules "inherent" in them, and I make a statement about this in our lab manual. This was something that was confirmed during the assessment.

Separately, the lab also has a set of guard band limits to which it will typically pass/fail a test result. In all cases, the guard band limits are set within the limits identified by the method or standard. Because we are pass/failing results based on the guard band limits, the assessor gave us the deficiency under the reasoning that we are no longer using the limits (decision rule) that is inherent in the method, and therefore we must define those limits to, and gain agreement from the customer per clause 7.1.3

Wanted to get some different thoughts surrounding if this deficiency holds any weight? I'm having a difficult time with the interpretation that we need to disclose the use of guard band limits to customers, when all that we are claiming to comply with is the method.

This is a fairly typical deficiency for a 17025 audit. When you claim to be testing to a procedure, and in this case also claiming that the decision rule is built into that procedure, then you have to follow that procedure without deviation. You state that you added your own guardbanding policy to the existing test method, so you are not in compliance with the test methodology as written.

The finding that you will need to notify the customer that you are using guardbanding in your decision rule is valid, since that is not inherent in your test method.