Random effect modelling of patient-related risk factors in orthopaedic procedures: results from the Dutch nosocomial infection surveillance network 'PREZIES'.

In the Dutch surveillance for surgical site infections (SSIs), data from 70277 orthopaedic procedures with 1895 SSIs were collected between 1996 and 2003. The aims of this study were: (1) to analyse the trends in SSIs associated with Gram-positive and Gram-negative bacteria; (2) to estimate patient-related risk factors for deep and superficial SSIs after all orthopaedic procedures, with special attention to primary total hip arthroplasty (THA); and (3) to analyse inherent differences in infection risk between hospitals. A random effect model was used to estimate the odds ratios of patient-related risk factors for developing an SSI, and to describe the distribution of the most widespread bacterial species responsible for SSIs among hospitals. Gram-positive organisms, mainly staphylococci, were the main cause of both deep (84.0%) and superficial SSIs (69.1%) after orthopaedic procedures. The percentage of SSIs after THA caused by coagulase-negative staphylococci decreased over the surveillance period, while the contribution of Staphylococcus aureus increased. Temporary elevations in the incidence of the most widespread pathogen species were observed within hospitals. Patient-related factors such as the National Nosocomial Infections Surveillance System risk index or age had little effect on the predictive power of the random effect models. This study underlines the usefulness of a random effect model, which adjusts risk estimates for random variation between hospitals, in a multicentre study on risk factors for SSIs.

Clinical breakpoint changes and their impact on surveillance of antimicrobial resistance in Escherichia coli causing bacteraemia.

Dutch laboratories are currently changing their breakpoint criteria from mostly Clinical Laboratory and Standards Institute (CLSI) breakpoints to European Committee on Antimicrobial Susceptibility Testing (EUCAST) breakpoints. To evaluate the impact of these changes, we studied antimicrobial resistance trends of Escherichia coli in blood specimens from January 2008 to January 2012 using CLSI and EUCAST breakpoints and compared them with the antimicrobial susceptibility test (AST) interpretations reported by Dutch laboratories participating in the Infectious Disease Surveillance Information System for Antibiotic Resistance (ISIS-AR). ISIS-AR collects AST interpretations, including underlying minimal inhibitory concentrations (MICs) of routinely cultured bacterial species on a monthly basis from Dutch laboratories. MICs of Etests or automated systems were reinterpreted according to the CLSI 2009 and EUCAST 2010 guidelines. Trends in non-susceptibility (i.e. intermediate resistant and resistant) over time were analysed by the Cochran-Armitage test for trend. The effects of the change from CLSI to EUCAST breakpoints on non-susceptibility were small. There were no differences in non-susceptibility to amoxicillin, amoxicillin/clavulanic acid, cefuroxim, gentamicin and co-trimoxazol and only small differences (1-1.5%) for ciprofloxacin between AST interpretations by CLSI or EUCAST. However, for ceftazidime, and cefotaxime/ceftriaxone the proportion of non-susceptibility was substantially higher when EUCAST breakpoints were used (2-3%). The effects on time trends of the change in guidelines were limited, with only substantial differences for the oxymino-cephalosporins. Our study shows that the implementation of EUCAST breakpoints has a limited effect on the proportion of non-susceptible isolates and time trends in E. coli for most, but not all, antimicrobial agents.