Figure 1.

Power to detect correlations in samples of 100, 160 and 1000 healthcare organisations. As the strength of the correlation to be detected increases, the power at all sample sizes increases, but 80% power is only possible at a higher true correlations when sample sizes are smaller. Illustrations are provided for sample sizes of 100 organisations (the approximate number of hospitals in the Netherlands), 160 organisations (applicable to the number of acute NHS hospitals in England) and 1000 (a possible sample size in studies of subsamples of general practices in England or hospitals in the United States) is calculated for a range of true correlations.