This is a creative proposal. When framing the analysis, it is suggested that the authors acknowledge that “outlier” status could be the result of random variation, unobserved differences in patient samples, and not necessarily because the sites are contributing “poor quality data”.