Using Centers for Disease Control National Nosocomial Infections Surveillance surgical site infection risk-adjustment for a group of related orthopaedic procedures

An important component of hospital infection control is surveillance to detect diminished levels of care and unforeseen problems. This involves morbidity and mortality audit and sequential data analysis using control charts and time series methods. In addition, regular public reporting of among-institution aggregated data is necessary for transparency and accountability. Analysis of hospital adverse event data may require risk-adjustment (RA) to ensure that changes are not due to differing patient populations. We examine the use of National Nosocomial Infections Surveillance surgical site infection (SSI) RA using data on 12 838 orthopaedic procedures. We evaluate the effectiveness of RA for these data using observed and expected tabulations and assessing discrimination by calculating the area under the Receiver Operating Characteristic Curve (AUC). RA may be of greater use with complex (deep and organ space) rather than all SSIs (superficial plus complex). We therefore suggest that, when reference data are available, the value of RA should be tested empirically. When there is no practically important difference between observed and expected reference data SSI rates, or when the AUC value is low, for example below 0.6, RA may be unnecessary.