Intraclass Correlation Coefficient

What is it?

Intraclass correlation coefficient (ICC) is a measure of the reliability between two measurements. It gives an estimation of the extent to which measurements can be repeated.

When should you use it?

The ICC is used when estimating the reliability of a numeric outcome measured several times. For discrete outcomes, other methods such as Cohen's kappa or Fleiss' kappa are more appropriate.

In biomedical research there are three common cases where ICC is used:

1. Interrater reliability: variation between observers who measure the same group of patients, blood tests, ...
2. Intrarater reliability: variation of a single observer measuring several times the same outcome
3. Test-retest reliability: variation of an instrument measuring several times the same thing

How to use it on EasyMedStat?

Your data should be organized with one record (line) for each subject and one column for each observer (interrater reliability) or for each measurement (intrarater reliability).

For example, if you want to assess the interobserver agreement between 2 raters, your data should look like:

1. Head to "Test variables"
2. Pick the first rater variable
3. Pick the second rater variable
4. Open the panel "Assess the interobserver or intraobserver agreement ..."

How to present your results?

• As usual, texts will be generated to provide you with an exemple of how you should present your results.
• Provide the ICC and its 95% confidence intervals in the results section of your article
• Provide the type of ICC used in the methods section of your article

Which ICC should be used?

Ten forms of ICC have been described for biomedical research. Six of them are calculated when you perform an analysis with EasyMedStat. By default, we make some assumptions that are the most commonly used in biomedical research for interobserver agreement: two-way random effects model, single rater, absolute agreement.

If you have any doubt regarding which ICC you should use, we advise you to refer to these articles:

• https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4913118/
• https://www.sciencedirect.com/science/article/pii/S2211568420302175