Antimicrobial stewardship has flourished in the United States and worldwide over the past decade, with resultant reductions in antimicrobial use being realized in many disparate settings. As the science of antimicrobial stewardship implementation advances and antimicrobial stewards receive increasingly granular and comparative data, stewards have an enhanced ability to target specific organizational goals.^{Reference Graber, Jones and Goetz1} The goals most typically pursued are (1) reductions in antimicrobial resistance and (2) reductions in incidence of Clostridioides difficile infection. These goals are typically complementary, but they occasionally conflict: decisions regarding where to promote use of piperacillin-tazobactam versus ceftriaxone (assuming lower risk of resistance development but higher putative C. difficile risk with ceftriaxone) are particular examples.^{Reference Graber2}

We are also hampered somewhat in our ability to use antimicrobial stewardship measures to reduce C. difficile infection by incomplete knowledge of just how much certain agents (and classes of agents) predispose a patient to C. difficile relative to others. While multiple studies indicate that fluoroquinolones and extended-spectrum cephalosporins (ie, second-generation and greater) predispose to C. difficile more than other antimicrobial agents or classes, the exact magnitude of increased risk is somewhat unclear.^{Reference Brown, Langford, Schwartz, Diong, Garber and Daneman3–Reference Slimings and Riley5} However, given their common use, particularly in inpatient settings, antimicrobial stewardship programs are currently devising efforts to limit the use of fluoroquinolones and extended-spectrum cephalosporins with a primary goal of C. difficile reduction.

With these efforts in mind, Szymczak et al^{Reference Szymczak, Muller and Shakamuri6} sought to determine behavioral barriers to limiting the use of fluoroquinolones and extended-spectrum cephalosporins. Through interviews with 64 frontline clinicians and antimicrobial stewardship stakeholders at 4 hospitals in 2 university-based healthcare systems, these researchers were able to generate rich thematic content on the perceived benefits and risks of these agents, particularly when used for treatment of urinary tract infection and pneumonia. Their findings are of tremendous use to the stewardship community at large. Primary themes of perceived benefit focused on low cognitive barriers in prescribing (eg, easy to dose, easy to switch from intravenous to oral, confidence in breadth of spectrum) that allowed for expending more cognitive energy on activities deemed more important and/or interesting. The patient’s ability to adhere to daily fluoroquinolone dosing regimens as outpatients was also frequently cited. Providers also described inertia and uncertainty in de-escalating from these agents, particularly when they were “inherited” by providers who saw the patient at their sickest. Fear of an adverse outcome also prompted an inability to de-escalate or discontinue therapy, even if providers intellectually knew that it was indicated.

Most directly related to the antimicrobial stewardship programs’ goal to reduce C. difficile, providers also seemed to have a poor understanding of the rationale of specifically restricting fluoroquinolones and extended-spectrum cephalosporins, frequently wondering about the quantification of their excess C. difficile risk. Some were openly skeptical of a strategy to limit certain classes of antibiotics; they argued (as we have promoted as stewards) that careful consideration of all antibiotic use should be the primary goal of antimicrobial stewardship. Even when providers were conscious of the increased C. difficile risk with fluoroquinolones and extended-spectrum cephalosporins, they often did not perceive the risk to outweigh the benefits of these agents, noting that C. difficile infection is often self-limited and that multiple treatment options are available. In the mind of the provider, fear of “sepsis uncovered” outweighs C. difficile risk most of the time.

Applying the dual process model of cognition proposed by Daniel Kahneman and Amos Tversky,^{Reference Kahneman7} these findings suggest that providers typically want to remain in system 1 (heuristic, rule-based, often emotion-based “fast” thinking) instead of system 2 (deliberative, contemplative, logical “slow” thinking) when it comes to prescribing antimicrobial agents. Two approaches can be taken to combat this tendency: (1) make information about the risks of adverse consequences of certain agents more available to system 1 or (2) force providers to move into system 2 through a timeout or other means. The first approach seems very achievable for fluoroquinolones because their side effects beyond promotion of C. difficile (eg, tendinopathy, dysglycemia, arrhythmia, neuropathy, drug interactions) have garnered much recent attention. For either approach to work for extended-spectrum cephalosporins, better quantification of the increase in C. difficile risk and risk for promoting resistance in general (eg, extended-spectrum β-lactamases, ampC-mediated cephalosporinases) needs to be researched, published, and advertised to the medical community at large. Finally, the potentially devastating consequences of C. difficile, an infection for which nearly 223,900 persons required hospital care and of which 12,800 persons died in the United States in 2017,⁸ should continue to be emphasized at every opportunity.

Antimicrobial stewardship, at a fundamental level, is about using data to influence behavior. Strategies for behavioral influence range from providing information and monitoring behavior, using social or financial incentives, altering choice architecture, and restricting or eliminating choice altogether.^{Reference Ubel and Rosenthal9} The optimal strategy for any stewardship intervention depends on the skill sets of stewards, resource availability, and organizational climate. However, for any strategy to work, data need to be robust regarding the reason that the stewardship intervention is indicated, and messaging needs to be clear on how the intervention will improve patient outcomes. The work by Szymczak et al^{Reference Szymczak, Muller and Shakamuri6} suggests that our work is not finsihed in either regard.