The use of the simplest version of Bland-Altman analysis requires some assumptions that are often not met in clinical practice. Furthermore, a complete Bland Altman analysis requires the use of repeated measurements with the same method, which is not possible with the simplest approach. To allow an appropriate analysis of repeated measurements, and to address unmet assumptions, the authors suggest a series of simple corrections. However, these are commonly not used in existing literature, nor are they available in commonly used software, such as R packages. When these corrections are not used, the estimate of the standard deviation of differences between methods is wrong, leading to a wrong estimate of the limits of agreement. This error can lead to the acceptance of inadequate methods (when the limits of agreement are too narrow) or the rejection of adequate methods (when the limits of agreement are too wide). The aim of this thesis is to estimate how the reliability of Bland Altman results is affected by the omission of the proposed corrections.
|
