Challenges and Opportunities in Leveraging Spectrum Scores to Assess Antibiotic De-escalation Practices and Outcomes in Sepsis

This is the authors' abstract. We don't add key points for this paper.

To the Editor—We thank Cunningham and Ilges [1] for their interest in our study [2] and appreciate their many insightful points. In particular, we agree that antibiotic spectrum scores, such as the Antibiotic Spectrum Index (ASI) [3], offer more granularity than the de-escalation definitions we used in our study. We defined de-escalation as cessation of anti–methicillin-resistant Staphylococcus aureus (MRSA) and anti-pseudomonal beta-lactam agents in patients with suspected sepsis when no MRSA or ceftriaxone-resistant gram-negative organisms were isolated. However, as Cunningham and Ilges pointed out, clinicians can narrow treatment to a variety of different agents that differ in their relative coverage spectra. Antibiotic spectrum scores provide a way to quantify the degree of de-escalation that may, in turn, be correlated with patient outcomes. We did initially consider using spectrum scores in our study but ultimately decided against doing so because of several unresolved issues. There are at least 4 published antibiotic spectrum scoring systems that all differ in important ways [4]. For example, the ASI [3] referred to by Cunningham and Ilges ranges from 1 to 13 and encompasses 50 agents and 14 organism categories, while the Madaras-Kelly spectrum score (MKSS) [5] ranges from 4 to 60 and covers 27 groups of agents and 21 organism categories. Some scoring systems, such as MKSS, also assign weights to conversion of intravenous to oral antibiotics while others do not. Spectrum scores further differ in how they handle patients receiving multiple agents with overlapping coverage. Moreover, all spectrum scores include some subjective and arbitrary decisions and do not account for the appropriateness of antimicrobial spectrum chosen for specific infections. Overall, we felt that these issues complicate the use and interpretation of spectrum scores, particularly when comparing de-escalation rates in patients with different infections caused by different organisms across diverse healthcare settings. Instead, we focused on de-escalation of anti-MRSA and anti-pseudomonal beta-lactam antibiotics because it is simple, clinically intuitive, and pragmatic. Epidemiologic studies demonstrate that many patients with suspected sepsis are administered empiric anti-MRSA and/or anti-pseudomonal beta-lactam therapy and this proportion is increasing over time, even though a small and decreasing proportion of those patients are infected with MRSA or ceftriaxone-resistant gram-negative organisms [6]. Furthermore, it is best practice for front-line prescribers to assess whether there is ongoing need for anti-MRSA and extended-spectrum gram-negative therapy when no MRSA or resistant gram-negative organisms are isolated after 48–72 hours of empiric therapy. In the absence of microbiological studies confirming infection with 1 of these organisms, it is generally appropriate to narrow or stop therapy. Although these practical considerations informed our approach, we recognize that there are potential advantages to spectrum scores. Like Cunningham and Ilges, we also hope our study will motivate future research that applies and compares different antibiotic spectrum scores to better understand de-escalation practices and their impact on patient outcomes. This could ultimately inform a more robust, evidence-based definition of antibiotic de-escalation that reliably reflects outcomes and that can be used to facilitate benchmarking within and between institutions across time. Financial support. This work was funded by the Centers for Disease Control and Prevention (grant number U54CK000611) and the Agency for Healthcare Research and Quality.