Using Bland-Altman for clustered data

[Kamini Reddy]

Can someone help me understand if I can use a Bland-Altman plot to visually assess the validity of comparing two different methods when I have a dataset containing IOP measurements from both eyes of the same patient, which means there's clustering at the patient level?

Thanks

[Leonardo Guizzetti]

The concept of the B-A plot (aka Tukey mean-difference plot) doesn't really extrapolate well to a clustered setting, because you need to account for intracluster correlation. However, I agree that a plot is useful. You could consider a scatter plot and a B-A plot where you show pairs of data (one measure from each method for each eye). However, the typical limits of agreement for a B-A plot are misleading, as would be the mean line in most cases.

You might mind the following paper useful:
Parker, R.A., Scott, C., Inácio, V. et al. Using multiple agreement methods for continuous repeated measures data: a tutorial for practitioners. BMC Med Res Methodol 20, 154 (2020). https://doi.org/10.1186/s12874-020-01022-x