‡ In the figures showing cumulative distributions, plotting the 17 hospitals that performed the greatest number of procedures in each category along the horizontal axis and numbering them 1–17 is equivalent to plotting a percentage of the number of hospitals. Calculation of percentages simply divides all values by a fixed number. Regardless of whether that fixed number is chosen to be 119, 62, or 17, the graph is unaffected. Only the labels on the axis would change.§ Barnard’s test compares proportions from multiple groups (i.e. , different hospitals) with one control group (i.e. , hospital G). The P value adjustment was used because of the multiple comparisons made with one hospital. Previous simulation studies 17showed that the resulting P values are conservative estimates.|| A one-sided test was used because the question was whether hospital G performs a disproportionately large percentage of rare procedures versus other hospitals. A disproportionately small percentage would be of no relevance.**P value by Fisher’s exact test. Confidence interval is for the difference of the two proportions (StatXact-5).