Validation of an Empiric Candidemia Treatment Algorithm

BACKGROUND: Judicious use of echinocandins may limit the development of resistance in Candida species. Guidelines endorse the use of echinocandins as initial therapy in candidemia, with fluconazole as an alternate choice in select patients. We compared the ability of providers to predict the need for echinocandin therapy in Candida bloodstream infections to that of a proposed institutional treatment algorithm designed to optimize empiric antifungal use. METHODS: In this retrospective study (10/2015–10/2016), patients were included with Candida isolated in ≥1 blood culture, without candidemia in the prior 14 days. Empiric treatment (the first antifungal prescribed for ≥24 hours after index blood culture draw) was considered “overly broad” if an echinocandin was administered to a fluconazole-susceptible isolate and “inappropriate” if fluconazole was administered to a fluconazole-non-susceptible isolate. An institutional algorithm was created recommending empiric echinocandin use based on the presence of ≥1 risk factors (Table 1). Provider choice and the recommended agent according to the algorithm were compared with the final fluconazole susceptibility of the organism. RESULTS: Among 65 episodes of candidemia, the majority of isolates were C. glabrata (Figure 1). Ninety-one percent of patients received non-azole therapy, primarily micafungin. Fluconazole was recommended by the algorithm in 25% of cases but initially prescribed in only 9% (Figure 2). Providers prescribed both overly broad and inappropriate treatment at a higher rate than algorithm recommendations (Figure 3). CONCLUSION: An algorithm using risk factors for fluconazole-non-susceptible Candida was able to predict appropriate empiric antifungal therapy better than provider decision making in cases of candidemia. Implementation of this algorithm into local treatment guidelines may improve empiric antifungal prescribing. DISCLOSURES: All authors: No reported disclosures.