Public reporting of infection rates as quality indicators

To the Editor: Echoing Scott and colleagues’ recent call for caution regarding the use of hospital standardised mortality ratios for benchmarking and public reporting,1 we have concerns about the use of infection rates as hospital quality indicators.

The National Healthcare Safety Network (NHSN) at the United States Centers for Disease Control and Prevention defines central line-associated bloodstream infection (CLABSI) as occurring where the patient has “a recognized pathogen cultured from 1 or more blood cultures and organism cultured from blood is not related to an infection at another site”.2 This and other criteria (for organisms associated with skin contamination and in children) are also used as quality indicators by VICNISS (the Victorian hospital-acquired infection surveillance system) and the Australian Commission for Safety and Quality in Health Care.

The infection prevention unit at Alfred Health in Melbourne recently assessed three cases of hospital-acquired bacteraemia as having met the NHSN/VICNISS definition of CLABSI. This assessment was disputed by Alfred Health’s intensive care unit and we sought further opinions from other infectious diseases physicians and infection control practitioners to adjudicate, in the absence of established criteria to determine whether the bacteraemia was due to infection at a site other than the central line. We received conflicting advice. These three cases would have comprised a third of the CLABSI cases at our hospital in the 6 months to April 2011, and thus significantly impact on our reported quality assurance indicators.

Similar common situations may arise in considering cases of bacteraemia in patients who have central lines in the context of chemotherapy-related mucositis; in patients with extensive burns; where cultures are only taken from central lines (rather than from peripheral venepuncture); where there are single blood cultures of enterococci of doubtful clinical significance; and where there are unconfirmed clinical foci of deep infection.3 Previous published work suggests that interrater agreement remains poor, despite changes to the definition of CLABSI.4

Other infection rates are associated with other problems as outlined by Scott and colleagues.1 For example, the current NHSN risk stratification system for infection (http://www.vicniss.org.au/Resources/HCWType1Manual/Type1Manualv6_0708.pdf) related to coronary artery graft surgery is poorly discriminatory, as most patients fall into risk index categories 1 or 2, and statistical calibration of this system is poor.5 Interrater reliability of classification of superficial infections has also been shown to be poor.6 The relatively low rate of infections in patients undergoing coronary artery surgery may result in large stochastic variation, particularly in hospitals performing relatively few such procedures.

Public reporting of infection rates is probably inevitable. However, such data should be restricted to indicators that have been shown to validly reflect preventable infections, using definitions that are reliable, and with the appropriate caveats regarding their limitations.