ABSTRACT
<p><b>INTRODUCTION</b>Institutional febrile neutropenia (FN) management protocols were changed following the finding of a high prevalence of ceftazidime-resistant Gram-negative bacteraemia (CR-GNB) among haematology patients with FN. Piperacillin/tazobactam replaced ceftazidime as the initial empirical antibiotic of choice, whereas carbapenems were prescribed empirically for patients with recent extended-spectrum beta-lactamase (ESBL)-producing Enterobacteriaceae colonisation/infection. An audit was conducted to determine the impact of these changes.</p><p><b>METHODS</b>Data from all FN episodes between October 2008 and December 2010 were collected prospectively, with mid-November 2009 demarking the transition between pre-intervention and intervention periods. Outcomes measured included 30-day mortality post-development of FN and the presence of CR-GNB.</p><p><b>RESULTS</b>There were 427 FN episodes (200 in the pre-intervention period) from 225 patients. The prevalence of CRGNB was 10.3%, while the 30-day mortality was 4.7%, with no difference between pre-intervention and intervention periods. Independent risk factors for 30-day mortality included the presence of active haematological disease, vancomycin prescription and older age. Independent factors associated with initial CR-GNB were profound neutropenia, the presence of severe sepsis and active haematological disease. Recent ESBL-producing Enterobacteriaceae colonisation/infection was not predictive of subsequent CR-GNB (positive predictive value 17.3%), whereas a model based on independent risk factors had better negative predictive value (95.4%) but similarly poor positive predictive value (21.4%), despite higher sensitivity.</p><p><b>CONCLUSION</b>A change in the FN protocol did not result in improved outcomes. Nonetheless, the audit highlighted that empirical carbapenem prescription may be unnecessary in FN episodes without evidence of severe sepsis or septic shock, regardless of previous microbiology results.</p>