Limitations of ranking lists based on cardiac surgery mortality rates.

BACKGROUND Ranking lists are a common way of reporting performance in cardiac surgery; however, rankings have shown to be imprecise, yet the extent of this imprecision is unknown. We aimed to determine the precision of, and fluctuations in, ranking lists in the comparison of cardiac surgery mortality rates. METHODS AND RESULTS Information on all adult cardiac surgery patients in all 16 cardiothoracic centers in The Netherlands from January 1, 2007, until December 31, 2009, was extracted from the database of the Netherlands Association for Cardio-Thoracic Surgery (n=46883). Ranks were assessed using crude and adjusted mortality rates, using a random effects logistic regression model. Risk adjustment was performed using the logistic EuroSCORE. Statistical precision of ranks was assessed with 95% confidence intervals. Additional analyses were performed for patients with isolated coronary artery bypass grafting. The ranking lists, based on mortality rates in 3 consecutive years, showed considerable reshuffling. When all data were pooled, the mean width of the 95% confidence intervals was 10 ranks using crude and 8 ranks using adjusted mortality rates. The large overlap of the confidence intervals across hospitals indicates that rank statistics were not materially different. Results were similar in the isolated coronary artery bypass grafting subgroup. CONCLUSIONS Rankings are an imprecise statistical method to report cardiac surgery mortality rates and prone to (random) fluctuation. Hence, reshuffling of ranks can be expected solely due to chance. Therefore, we strongly discourage the use of ranking lists in the comparison of mortality rates.