Discussion

We present our first-year experience with a novel antimicrobial withdrawal intervention designed to reach high-hanging fruit in hospital antimicrobial stewardship—including cases of empiric antimicrobial use without evidence of underlying infection, prolonged antimicrobial therapy for confirmed, clinically resolved infections, and prolonged postsurgical antimicrobial prophylaxis. We demonstrate our policy to be an infrequent but useful adjunct to address cases historically challenging to approach via traditional PAF.

Piperacillin-tazobactam followed by vancomycin were the most frequently withdrawn agents, together comprising half of all interventions. Both agents are unrestricted on our hospital formulary and are the most prescribed antibiotics for suspected healthcare-associated infections at our institution. Broad-spectrum antibiotic use has been previously described as a risk factor for nonadherence to PAF—in one cohort ^{Reference Horton, Rawlins, Manning and Ingram10} , piperacillin-tazobactam use was associated with 55% lower odds of adherence to PAF. Prolonged duration of therapy for confirmed, resolved infections, and prolonged postsurgical prophylaxis (ie, cefazolin prophylaxis for presence of surgical drain) were the most common reasons for withdrawal intervention. Prolonged duration of therapy has been previously used as a target for ASP-driven antibiotic discontinuation, resulting in significant decreases in mean total antibiotic days. ^{Reference Bolten, Bradford, White, Heath, Sizemore and White9}

The majority (∼76%) of withdrawal interventions occurred on surgical services. The disproportionate number of interventions is consistent with prior stewardship literature on suboptimal antibiotic prescribing among surgical services. A qualitative study ^{Reference Charani, Tarrant, Moorthy, Sevdalis, Brennan and Holmes3} of inpatient surgical teams identified antibiotic decision-making to be an uncoordinated, lower-priority task and described a general ambiguity in management of infections. In the same study, some surgeons described themselves as interventionists who acknowledged defensively prescribing antibiotics for fear of adverse postsurgical outcomes. An intervention ^{Reference Broom, Tee, Broom, Kelly, Scott and Grieve5} designed to reduce inappropriately prolonged antibiotic use in surgical patients with source control also found the hierarchical nature of clinical decision-making within surgical teams to be a barrier to optimization of antibiotic therapy, with junior team members feeling powerless to influence more senior team members’ prescribing practices.

Our withdrawal policy was designed with the goals of enhancing prescriber communication and accountability over antibiotic prescribing. We found a so-called “nudge effect” following a quarter of interventions, where the prescribing team self-discontinued the antimicrobial in question and interpreted this as a surrogate for prescriber concordance with ASP withdrawal decision; in fact, in a few instances, frontline prescribers were grateful for ASP intervention in providing leverage against more senior team members to discontinue unnecessary therapy. Our policy allowed the prescribing team a 24-hour window to appeal the withdrawal decision to the ASP physician. The appeal process was intended to create a flipped dynamic, where the prescribing team needed to justify continuation of antibiotic therapy to ASP (in contrast to ASP justifying discontinuation of therapy to prescribing team); we believe this process enhanced intra-team communication (ie, between junior frontline prescribers and more senior team members) and improved prescriber answerability and transparency.

In response to the minority of cases of reordering of the withdrawn agent (or ordering of an alternative antimicrobial), a clause was added to the withdrawal policy after the end of the study period to allow the ASP to immediately discontinue inappropriately re/ordered agents within 48 hours of intervention. We found our withdrawal intervention reduced burden on PAF by not revisiting cases where recommendations were previously not accepted, consistent with a previously published ASP-driven antibiotic discontinuation intervention. ^{Reference Bolten, Bradford, White, Heath, Sizemore and White9} However, we found this reduction was partially offset by increased burden on the ASP physician needing to address appeals and re/ordering of antimicrobials. We found limited burden on the busy ID consult service from appeals to withdrawal and reassuringly high concordance between withdrawal decision and ID consultant recommendations.

We found reassuring safety of our withdrawal intervention with respect to incident infections, readmission, and in-hospital mortality within 30 days of withdrawal. In one case, nitrofurantoin, while inappropriately prescribed, may have been suppressing growth of Klebsiella in a patient who developed bacteremia following its withdrawal; while nitrofurantoin susceptibility testing was not performed on this blood isolate, nearly half of all Klebsiella cultured from urine are resistant to this agent at our institution. No readmissions or deaths were attributed to preceding withdrawal intervention. An important caveat, however, is that we could not verify if patients were readmitted to another hospital or died following hospital discharge. A study ^{Reference Loo, Liew, Lee, Lee, Chlebicki and Kwa11} of 794 ASP interventions over a five-year period assessing the safety of ASP discontinuation of empiric antibiotics within 24 hours of prescribing in patients with no evidence of bacterial infection found significantly decreased DOT and length of stay and no difference in infection-related readmission or all-cause 30-day mortality.

Our intervention had some strengths: flexibility to intervene at any point during the course of prescribed therapy; discretion by ASP physician to intervene for multiple categories of inappropriate antimicrobial use; broad range with applicability to both formulary-restricted and unrestricted antimicrobials; wide reach across all inpatient services; enforceability via hospital policy and support from hospital leadership; appeal process to improve prescriber accountability over antimicrobial prescribing.

Our intervention was not without limitations. As this was a descriptive case series of an intervention without a comparison group, causal inferences regarding withdrawal outcomes cannot be drawn. Our experience may not be applicable to other centers. Our policy was introduced years after the ASP developed rapport and credibility with prescribers and leadership across the hospital. We emphasize this policy should be used as an infrequent adjunct and not as a substitute to traditional ASP activities including PAF, formulary restriction with prior authorization, and antimicrobial prescribing guidance development (eg, for postsurgical antibiotic treatment or prophylaxis). Judicious use is especially important when weighing respect for prescriber autonomy with concern for safety and quality of antimicrobial prescribing as an overly restrictive approach can lead to loss of trust between prescribers and stewards; ^{Reference Davey, Marwick and Scott12} this balancing act is not new for stewardship programs.

We present our early experience with an innovative antimicrobial withdrawal intervention that effectively targeted inappropriate inpatient antimicrobial use and improved prescriber accountability over antimicrobial prescribing. Our policy was found to be a safe and useful addition to traditional ASP activities. Continued follow up will be needed to gauge sustainability and long-term impact.