How should we interpret hospital infection statistics?

To the Editor: Cheng and colleagues provide a perspective on how to better report and interpret hospital infection data.1 They suggest that the National Health Performance Authority (NHPA) report on rates of Staphylococcus aureus bacteraemia (SAB) across comparable public hospitals2 would have benefited from the use of alternative statistical approaches, such as funnel plots.

To support fairer comparisons between hospitals, the NHPA assigned 352 public hospitals to one of eight peer groups based on hospital size, services provided and the percentage of at-risk patients. The report identified all hospitals with zero reported SAB cases, to assist the clinical community to determine whether these hospitals are exemplars or underreporting.

There are numerous reasons why funnel plots were not used. Funnel plots show the observed infection rate of a hospital in the context of the range of infection rates that would be expected. Underreporting by hospitals will affect the calculation of confidence intervals, the shape of the funnel and the subsequent identification of outlier hospitals. Funnel plots suggest that hospitals within the funnel have acceptable infection rates, therefore removing the incentive to decrease rates further and implying that action is only required at hospitals statistically significantly different than the average (ie, above the funnel). The unwarranted variation in rates identified by Cheng et al and the NHPA demonstrates the opportunity for most hospitals to reduce rates. Given that SAB is associated with a mortality rate of 20%–35%,2 it is clear we should be aiming for as few cases as possible.

As this letter goes to press, the NHPA is due to publish an updated report on rates of SAB in early March, which includes data from the 2012–13 financial year and which takes into account feedback from various sources, including that of Cheng and colleagues.