Optimizing Empiric Antibiotic Selection in Sepsis: Turning Probabilities Into Practice

To the Editor—We read with interest the retrospective cohort study by Guillamet et al [1]. In this article the authors developed decision trees, based on epidemiologic and microbiologic predictors, to categorize a patient’s risk for piperacillin-tazobactam, cefepime, or meropenem resistance in the context of septic shock with gram-negative bacteremia. By predicting an infecting isolate’s probability of resistance to common broad-spectrum β-lactam agents, this tool could be used to improve empiric antibiotic decision making. This report adds to a growing literature in which epidemiologic and microbial factors are used to predict resistance in order to guide therapy at various empiric windows [2, 3]. There remain 2 important barriers to using the results of this study to guide empiric therapeutic decisions. First, the decision tools produced in this approach rely on species identification to stratify risk of resistance. Even if a rapid diagnostic test (RDT) were available that could provide species identification instantaneously, when practical factors such as laboratory transit time and specimen collection are added, this may exceed the acceptable time frame for initiation of empiric antimicrobials [4, 5]. While RDTs may be increasingly rapid and discriminatory [6], until the testing is truly available at point of care, predictive tools that use only patient characteristics available at the time of empiric therapeutic decision making are needed. Second, the model outputs in this study do not generate the type of dichotomized results provided by phenotypic or some molecular and genotypic susceptibility tests. Rather they must rely on clinicians’ “thresholds” of appropriate adequate coverage for interpretation and operationalization (prescribe or do not prescribe). For example, all clusters in the cefepime algorithm produce risks of resistance that some clinicians might reasonably consider “too high” to justify empiric use of this agent in the setting of septic shock. To apply the proposed approaches, and identify relevant “thresholds,” we need to understand at least one of the following: (1) the clinical impact of different thresholds for adequate coverage or, in the absence of this, (2) current thresholds that are implicitly (unconsciously or indirectly) used by physicians when initiating empiric therapy in sepsis syndromes. Specific thresholds from the literature, based on the above approaches, are absent. If clinicians are provided with a probability of antibiotic resistance for a pathogen, we need them to be informed as to how they should react to this probability, or else there could be unintended consequences. These include not only potential unnecessary increases in broad-spectrum antibiotic use, but may also include paradoxical decreases in adequate therapy. To add to the complexity, clinicians will not only be receiving probabilistic data from epidemiologic sources, but also integrating them with imperfect data from increasingly utilized molecular and genomic tests. To develop and apply any decision-making aid to time-sensitive empiric therapeutic choices, we must better understand the required standards of speed and thresholds of adequate coverage for serious bacterial infections. Potential conflicts of interest. All authors: No reported conflicts of interest. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.