Interpreting the limits of agreement: do I have good or bad agreement?

I am often asked how do I tell from the limits of agreement whether the two methods of measurement agree. What would be good agreement and what would be bad?

This question can be answered only when you know what the measurement is and how closely you need the two methods of measurement to agree.

The question is usually: can we replace an old method of measurement by a new one and use the measurements that we get in the same way? In other words, can we just record the measurement without recording the method used to get it?

For example, if you have a new method of measuring blood pressure, you might say that this will be interchangeable with the method you now use if the systolic and diastolic measurements made at the same time are within 10 mm Hg. You would estimate the limits of agreement and then if the limits were closer than 10 mm Hg, you could say that the two methods are interchangeable.

How close the measurements have to be is a clinical decision, not a statistical one. The 10 mm Hg in my example would be decided by the researcher or clinician who wants to know, not by me. The limits of agreement do not provide a general coefficient which enables us to say whether method of measurements agree. They measure how close the agreement is and we then have to say whether this is close enough for what we want.

Frequently asked questions on the design and analysis of measurement studies.

Measurement studies menu.

Martin Bland's home page.

This page maintained by Martin Bland.
Last updated: 20 March, 2009.

Back to top