Discussion

We demonstrated that implementation of an ATO was feasible and practical, but its impact on antibiotic use was minimal. As indicated by an e-mail survey, the ATO process was well accepted by both pharmacists, who felt empowered to engage in antibiotic discussions, and team members, who appreciated pharmacist input. We chose to use team-based pharmacists to perform ATOs because the same 1–2 pharmacists are typically assigned to a medical team, so they develop a rapport with team. Additionally, while physicians rotated monthly, team pharmacists provided a stable core group of personnel who were facile with EHR use and documentation. Finally, we chose to make the activity team-based and not to provide ASP team involvement to determine whether the addition of ATOs to our current activities provided benefit.

ATO compliance initially was high, but it was less well adopted in the UC group. Moreover, 6 pharmacists served the UC teams, and only 1 documented >1 ATO per day staffed, with 3 pharmacists documenting <0.4 ATOs per day. In comparison, only 1 of the initial 6 ATO pharmacists documented <1 ATO per day. The reason for this disparity is unclear because the education and support provided to pharmacists in both groups was identical. Individual pharmacist comfort and engagement in stewardship may have varied, as may have the team’s attending physician’s interest in addressing antibiotic issues.

The preferred timing of ATOs is unknown. Ideally, they occur whenever clinical or microbiologic data become available to allow adjustments in antibiotic use. Our data suggest that the first 72 hours or immediately after ICU transfer were the most fruitful times for antibiotic regimen review. A significant opportunity for intervention may have been missed by targeting the 48–72-hour window. Much clinical and microbiologic data return before this time, allowing regimen adjustment that may benefit from an ATO. Late ATOs exhibit lower yields because many patients had already been discharged or a definitive antibiotic plan had already been created. No recommendation for change was the second most common intervention overall and the most common in the late ATO category. This finding suggests that antimicrobial plans later in the course of care are not generally amenable to a time-out performed by a team member, although they may still benefit from outside intervention by stewardship experts.

We did not detect a difference in overall antibiotic use or in specific antibiotic use with the implementation of ATOs. Intravenous antibiotic DOT days decreased relative to oral DOT with ATO implementation in the UC group, driven primarily by decreased intravenous (IV) levofloxacin use. Pharmacist-directed IV to PO switch interventions have previously been shown to be effective, and early transition to PO was one of the ATO targets.^{Reference Dunn, O’Reilly, Silke, Rogers and Bergin8,Reference Kuti, Le, Nightingale, Nicolau and Quintiliani9}

There are a number of possible explanations for why the ATOs did not measurably alter antibiotic use. First, our measure of DOT may have been too blunt to detect a difference. Previous studies of stewardship interventions that focused on de-escalation have detected changes that were statistically significant but were generally performed <24 hours after the order.^{Reference Banerjee, Teng and Cunningham10,Reference Beganovic, Costello and Wieczorkiewicz11} Such a change would be difficult to detect using DOT. Measuring time to most narrow-spectrum appropriate therapy or time to first de-escalation may better reflect the impact of the intervention, but these metrics are more challenging to measure. The duration of our study may also have impacted the outcome. Possibly, over a longer period, antibiotic use changes may have been greater, but our study was of relatively short duration due to its cluster-randomized design and the need to avoid group contamination. Our hospital has an active ASP focusing on audit and feedback for many years, and clinicians are generally aggressive with antimicrobial de-escalation. This factor may have decreased the influence of the ATOs on antibiotic use as well.

Another reason for a lack of impact may have been the time points chosen (as noted previously). Our program does not utilize preauthorization to alter empiric choices, and the ATO may be have had more impact if it had been performed earlier or as part of a daily patient checklist. Daily checklists have been effective at decreasing antibiotic use in the ICU and at improving targeted antimicrobial use.^{Reference Weiss, Moazed and Mcevoy12} A subsequent study compared an electronic checklist to face-to-face prompting and found that face-to-face prompting significantly decreased antibiotic use compared to the paper format.^{Reference Weiss, Dibardino, Rho, Sung, Collander and Wunderink13} Interestingly, an ATO using a paper checklist on hospital days 3–5 showed no impact on antibiotic use.^{Reference Thom, Tamma and Harris5}

Scheduling and personnel issues may have impacted our study. The ATOs were not performed on weekends or holidays, reducing opportunities by ~30%. Such issues are common for ASPs because weekend staffing is uncommon. In a recent single-center study, the impact of not performing weekend antibiotic reviews on restricted antimicrobials resulted in an additional 3.8 days compared to those that were reviewed during the week.^{Reference Freeman, Eschenauer and Patel14} Pharmacist variability in expertise and experience may have influenced our results. To attempt to mitigate this factor, we provided identical training to all pharmacists regarding antibiotic use in common infections and met with them at study initiation and 2 weeks into both periods A and B. No direct observations of ATOs were performed, nor was active monitoring and feedback of ATO use or documentation performed. Addressing these issues may have improved ATO utilization, but we were seeking to determine whether an ATO process without direct stewardship intervention and active support could be successful. Our use of team-based pharmacists limits the applicability of our intervention to only those sites that use a similar team model. Pharmacists generally are ideally positioned to intervene to improve antibiotic use, but the role and effectiveness of non-ID pharmacists who are not supported by ID-trained experts is poorly defined.

Our finding that ATOs do not impact antibiotic use is not isolated. Multicenter implementation of a prescriber-driven ATO using a paper form filled out on hospital days 3–5 decreased “inappropriate” antibiotic prescribing from 45% to 31% but did not change any other antibiotic use measure.^{Reference Thom, Tamma and Harris5} Two studies utilized electronic notifications to prompt ATOs. Graber et al required users to enter a structured clinical note if vancomycin or piperacillin-tazobactam were continued for >3 days.^{Reference Graber, Jones and Glassman15} Vancomycin early discontinuation was more common after the ATO implementation, but piperacillin-tazobactam use was unchanged. Another center created an interruptive alert appearing after 72 hours for patients on empiric antibiotics with no defined duration.^{Reference Richardson, Neuner and Athans6} The alert fired nearly 33,000 times for 6,138 patients. Antibiotics were discontinued or de-escalated in 21% of patients in the 6 hours after the alert fired, although no clear link between the alert and de-escalation could be determined, nor were these researchers able to compare antibiotic use to the prealert period, and overall antibiotic remained unchanged.

Although ATOs make inherent sense, the current literature suggests that their impact is limited. Notably, all studies of ATOs cited (other than the ICU checklists) were conducted at centers with active ASPs, and adding ATOs to an active ASP may yield minimal benefit. An ATO may be beneficial when an active ASP is absent or the program lacks ID expertise, although ATOs have not been studied in these situations. ATOs will benefit from better tools and training of those performing the activity. Nonetheless, asking clinicians with variable ID experience and expertise to self-police antibiotic use through an ATO may not be successful. Greater benefit may come from expanding access to stewardship and ID experts through the use of audit and feedback and “handshake stewardship,” which have been shown to provide meaningful improvements in antibiotic use.^{Reference Anderson, Watson and Moehring16,Reference Hurst, Child, Pearce, Palmer, Todd and Parker17} The downside to these interventions is that they require time and the availability of experts.

In summary, implementation of a pharmacist-led ATO was feasible and well accepted but did not change overall antibiotic use or impact patient outcomes. ATOs may promote increased use of PO antibiotics, but more effective strategies for self-stewardship are needed.