Impact of Antibiotic Stewardship Rounds in the Intensive Care Setting: A Prospective Cluster-Randomized Crossover Study

Abstract

Background

Few groups have formally studied the effect of dedicated antibiotic stewardship rounds (ASRs) on antibiotic use (AU) in intensive care units (ICUs).

Methods

We implemented weekly ASRs using a 2-arm, cluster-randomized, crossover study in 5 ICUs at Duke University Hospital from November 2017 to June 2018. We excluded patients without an active antibiotic order, or if they had a marker of high complexity including an existing infectious disease consult, transplantation, ventricular assist device, or extracorporeal membrane oxygenation. AU during and following ICU stay for patients with ASRs was compared to the controls. We recorded the number of reviews, recommendations delivered, and responses. We evaluated change in ICU-specific AU during and after the study.

Results

Our analysis included 4683 patients: 2330 intervention and 2353 controls. Teams performed 761 reviews during ASRs, which excluded 1569 patients: 60% of patients off antibiotics, and 8% complex patients. Exclusions affected 88% of cardiothoracic ICU (CTICU) patients. The AU rate ratio (RR) was 0.97 (95% confidence interval [CI], .91–1.04). When CTICU was removed, the RR was 0.93 (95% CI, .89–.98). AU in the poststudy period decreased by 16% (95% CI, 11%–24%) compared to AU in the baseline period. Change in AU was differential among units: largest in the neurology ICU (–28%) and smallest in the CTICU (–2%).

Conclusions

Weekly multidisciplinary ASRs was a high-resource intervention associated with a small AU reduction. The noticeable ICU AU decline over time is possibly due to indirect effects of ASRs. Effects differed among specialty ICUs, emphasizing the importance of customizing ASRs to match unit-specific population, workflow, and culture.

antibiotic stewardship, handshake stewardship, intensive care unit, ICU

Intensive care units (ICUs) provide care for patients with significant clinical complexity, many of whom require antibiotic treatment. Moreover, patients in ICUs are at high risk for developing infections with antibiotic-resistant pathogens and adverse events due to the intensity of antibiotic treatment, use of invasive devices, and significant contact with healthcare personnel. These characteristics make the ICU a high-yield target area for antibiotic stewardship programs (ASPs) to help optimize antibiotic use (AU).

Descriptive information and quasi-experimental data available regarding antibiotic stewardship (AS) in ICU settings have shown reductions in broad-spectrum antimicrobial use, incidence of infections and colonization with multidrug-resistant bacteria, antimicrobial-related adverse events, and healthcare-associated costs, all without increase in mortality [1–7]. However, while many groups may practice AS in ICUs, design and frequency of AS interventions vary widely. Few controlled studies exist to assist AS programs at estimating the impact of such strategies relative to their resource investment.

“Handshake stewardship” rounds have been recommended as an effective and sustainable rounds-based service for AS programs to implement [8]. The purpose of this study was to evaluate the impact of weekly AS rounds (ASRs) on AU in ICUs in an academic medical center, including information on unit-level effects among specialized ICUs.

METHODS

We performed a 2-arm, cluster-randomized, crossover trial to study the intervention of ASRs in ICUs from October 2017 through June 2018 compared to usual care. The study was performed in 5 adult ICUs at Duke University Hospital, a 952-bed academic medical center: a medical ICU, a surgical ICU, a cardiothoracic ICU (CTICU), a cardiac ICU, and a neurologic ICU. Each ICU was composed of 2 sides, or “half-units.” Standard practice in the ICUs assigned a distinct and independent rounding team to each half-unit (Figure 1). For example, in an ICU with 16 beds, beds 1–8 were referred to as the “low side” half-unit as they have the low numbers, and the “high side” of the unit referred to beds 9–16. In our study, the half-unit was the unit of randomization. We used the half-unit as the unit of randomization because (1) each ICU cares for a unique patient population and we wanted a control group with similar case mix, and (2) existing rounding structures by half-unit avoided overlap in primary team personnel between arms. In each ICU, one half-unit was randomly assigned to the intervention and the other half-unit was assigned to usual care for the first 4 months of the study. After 4 months, control half-units switched to the intervention and the intervention half-units switched to control.

Figure 1.

Schematic of crossover trial design. A half-unit refers to either the low side (eg, beds 1–8 in a 16-bed unit) or the high side (eg, beds 9–16 in a 16-bed unit). After 4 months in the trial, the half-units in each intensive care unit crossed over to either receive the intervention (if the half-unit received usual care first) or usual care (if the half-unit received the intervention first). Abbreviations: ASP, antibiotic stewardship program; ICU, intensive care unit.

The intervention involved weekly AS handshake rounds in addition to routine pharmacist-led postprescription reviews. The AS team consisted of infectious disease (ID) physicians and pharmacists. The AS team first reviewed ICU census lists to apply inclusion and exclusion criteria for rounds discussions. Patients were excluded from review on ASRs if they were <18 years old, had no active antibiotic order, or if they had a marker of high complexity including receipt of an existing ID consult, transplantation, ventricular assist device, or extracorporeal membrane oxygenation therapy. High-complexity patients were excluded to maintain ASR efficiency and to avoid duplication of efforts from the AS team to the ID consult team. If included, the team then performed a detailed chart review and developed questions and recommendations to be discussed on rounds. Team members recorded reviewed data in a standardized REDCap form.

For ASRs, physicians and pharmacists from both AS and ICU teams met face-to-face on the unit once a week at predesignated times. On rounds the teams discussed antibiotic optimization for the reviewed patients. Patients could be reviewed multiple times from week to week if they remained in the ICU and continued to meet inclusion criteria. The following week, AS team members re-reviewed patient charts to see if the ICU teams implemented the changes recommended by the AS team the week prior. Team members recorded these data in another standardized REDCap form.

Measures

The primary comparison was between patients in half-units randomized to the ASR intervention compared to patients in half-units without ASRs. The primary outcome measure was AU in days of therapy (DOT) per 1000 days present during their initial ICU stay and for the remainder of the hospitalization. All demographic and antibiotic exposure data were extracted from a preestablished AU surveillance database. We obtained information on adverse effects from antibiotics from our hospital’s prospective, voluntary safety reporting system.

Analysis

We used multivariable negative binomial regression to produce rate ratios (RRs) of antibiotic DOT per 1000 patient-days comparing patients in the control vs intervention time periods. We defined an “order” variable to indicate whether the half-unit received the intervention or control during the first 4 months. The model was adjusted for the order of the intervention and the ICU half-unit. To assess for interaction between variables, we used a test of homogeneity. Specifically, we assessed if there was a lingering effect when the half-unit received the intervention first. We also reviewed AU before and after the study period to assess overall and ICU-level trends. AU data for each unit and overall were collected per month. We calculated the median AU data for the 10 months prior to the study and defined this as the baseline AU. We calculated the median AU for the 6 months after the study ended and defined this as the postintervention AU. Using these medians, we calculated the absolute change and percentage change in AU for each ICU separately and among all 5 ICUs. Analyses were performed in SAS version 9.4 software (SAS Institute, Cary, North Carolina). The study was reviewed by the Duke University Institutional Review Board and deemed an exempt quality improvement activity.

RESULTS

Our analysis included 4683 ICU-exposed patients among 5 ICUs; 2330 patients were randomized to time periods and half-units with active ASRs and the remaining 2353 patients served as controls. The mean age in the study population was 61.2 years (standard deviation [SD], 15.9 years) and 42% were female. The most common primary admission diagnosis was ischemic heart disease, followed by neoplasm. Descriptive characteristics were similar between the control and intervention groups (Table 1).

Table 1.

Demographic Data of 2330 Intervention Patients and 2353 Control Patients

	Control	Intervention
Variable	(n = 2353)	(n = 2330)	P Value
Age, y, mean (SD)	61.0 (15.9)	61.3 (16.0)	.52a
Female sex	968 (41.1)	1000 (42.8)	.37b
Race			.52c
American Indian/Alaska Native	23 (1.0)	21 (0.9)
Asian	33 (1.4)	31 (1.3)
African American/Black	623 (26.5)	671 (28.8)
White	1556 (66.1)	1490 (64.0)
Other	118 (5.0)	117 (5.0)
Primary diagnosis			.29c
Ischemic heart disease	264 (11.2)	244 (10.5)
Neoplasm	222 (9.4)	239 (10.3)
Sepsis	183 (7.8)	164 (7.0)
Heart valve disorder	135 (5.7)	153 (6.6)
Intracranial hemorrhage	119 (5.1)	112 (4.8)
Hypertension complication	101 (4.3)	115 (4.9)
Cerebrovascular disease	94 (4.0)	92 (3.9)
Other cardiac diseased	261 (11.1)	243 (10.4)
Other respiratory diseasee	117 (5.0)	143 (6.1)
Other	857 (36.4)	825 (35.4)
Intensive care unit			.09c
Medical ICU	327 (14.0)	359 (15.3)
Surgical ICU	511 (21.7)	481 (20.6)
Cardiothoracic ICU	549 (23.3)	488 (20.9)
Neurologic ICU	464 (19.7)	457 (19.6)
Cardiac ICU	502 (21.3)	545 (23.4)
Charlson score, mean (SD)	6.77 (5.05)	6.87 (5.20)	.50a

Data are presented as No. (%) unless otherwise indicated.

Abbreviations: ICU, intensive care unit; SD, standard deviation.

^aPaired t test.

^bFisher exact test.

^cχ ² test.

^dOther cardiac disease: heart failure, cardiac device complication, arrhythmia, aortic dissection, cardiomyopathy, endocarditis, cardiac arrest, and congenital circulatory defect.

^eOther respiratory disease: respiratory failure, interstitial pulmonary disease, chronic obstructive pulmonary disease, pulmonary embolism, and pneumonitis/pneumonia.

Among patients randomized to intervention and screened by the AS team, 1569 (68%) patients were excluded from discussion on ASRs (Table 2). The most common reason for exclusion was no active antibiotic order (60.6%). The unit with the highest number of excluded patients was the CTICU with 88% of patients excluded from review, mostly due to high-complexity features including transplantation, ventricular assist device, and an existing ID consult.

Table 2.

Reasons for Exclusion From Antibiotic Stewardship Round Discussions

Reason for Exclusion	Total (N = 2330)	Medical ICU (n = 359)	Surgical ICU (n = 481)	Cardiothoracic ICU (n = 488)	Neurologic ICU (n = 457)	Cardiac ICU (n = 545)
Total patients excluded from ASP rounds	1569 (68.2)	183 (51.0)	322 (66.9)	430 (88.1)	280 (61.3)	354 (65.0)
Reasons for exclusion
No antibiotic	951 (60.6)	100 (54.6)	186 (57.8)	127 (29.6)	240 (85.7)	289 (81.7)
Transplant recipient	216 (13.8)	31 (16.9)	46 (14.3)	142 (33.0)	0 (0.0)	16 (4.5)
ECMO	16 (1.0)	6 (3.4)	0 (0.0)	24 (5.6)	0 (0.0)	0 (0.0)
VAD	40 (2.5)	0 (0.0)	0 (0.0)	44 (10.2)	0 (0.0)	5 (1.4)
Existing ID consult	346 (22.1)	46 (25.1)	90 (27.9)	93 (21.6)	40 (14.3)	44 (12.4)

Data are presented as No. (%).

Abbreviations: ASP, antibiotic stewardship program; ECMO, extracorporeal membrane oxygenation; ICU, intensive care unit; ID, infectious disease; VAD, ventricular assist device.

Among 2300 intervention patients, the most common suspected/documented infection was pneumonia (42.1%) (Table 3). In most cases the infectious diagnosis was suspected or presumed without confirmatory microbiology data (29.1%). A total of 761 (33%) patients assessed in the intervention group were actively discussed during weekly multidisciplinary ASRs.

Table 3.

Intervention Data

Variable	Intervention Group (n = 2300)
Infection suspected/documented
Pneumonia	968 (42.1)
Intra-abdominal infection	226 (9.8)
Unknown	220 (9.6)
Urinary tract infection	197 (8.6)
Bloodstream infection	161 (7.0)
Skin and soft tissue infection	133 (5.8)
Clostridioides difficile infection	49 (2.1)
Endocarditis	10 (0.4)
Meningitis	10 (0.4)
Othera	326 (14.2)
Presence or absence of microbiologic data on day of intervention
Clinical diagnosis without microbiology data	615 (26.7)
Diagnosis confirmed with microbiology data	670 (29.1)
Clinical diagnosis workup pending	1015 (44.2)
No. of patients discussed on rounds	761 (33.1)
Patients discussed who also received recommendations	746/761 (98.0)
Total No. of recommendations made on rounds	1051
Median recommendations per patient (range)	1 (0–3)
Stewardship recommendations (multiple response per patient)b
Define end date for antibiotics	267 (25.4)
Discontinue antibiotics on day of review	156 (14.8)
Noninfectious diagnosis	121 (11.5)
Discontinue antibiotics if certain criteria met	109 (10.4)
De-escalation of antibiotics if certain criteria met	78 (7.4)
De-escalation of antibiotics on day of review	67 (6.4)
Obtain additional microbiology data	45 (4.3)
Antibiotic dose optimization	39 (3.7)
Diagnostic imaging	34 (3.2)
Catheter removal	32 (3.0)
Other procedural intervention	29 (2.8)
Infectious disease consult	27 (2.6)
Intravenous to oral antibiotic change	17 (1.6)
Different antibiotics (horizontal change)	16 (1.5)
Broaden antibiotic coverage	14 (1.3)
Recommendation followed by day 7	815/746 (77.5)
Antibiotic recommendations	603/763 (79.0)
Procedural or additional image/test intervention	107/140 (76.4)
Noninfectious diagnosis	86/121 (71.1)
Infectious disease consultation	19/27 (70.4)

Data are presented as No. (%) unless otherwise indicated.

^aExamples of “other” infections included, but were not limited to, sinusitis, intracranial shunt infection, brain abscess, and osteomyelitis.

^bFrequency calculated from 1051 recommendations.

Of the patients who were reviewed, at least 1 recommendation was made by the ASP team in 98% of discussed cases. The most common recommendation made by the team was defining an end date for antibiotics (25.4%). At day 7, 77.5% of the recommendations made by the ASP team were followed by the ICU team. Antibiotic recommendations were most likely to be accepted (79.0%) at day 7; an ID consultation was least likely to be accepted (70.4%).

Patient-level outcomes in both the control and intervention groups were similar (Table 4). The median duration of antibiotic therapy throughout hospitalization was 5 days for both groups. Similarly, ICU DOT, total length of stay (LOS), and ICU LOS were not statistically significantly different. Patients in the control group had a median LOS of 10 days (SD, 18.3 days) compared to a median of 9 days (SD, 19.8 days) in the intervention group (P = .11). We observed no significant adverse drug events during the study period as reported through the hospital safety reporting system. There was no difference in the incidence of Clostridioides difficile infections between the control and intervention groups. Overall 30-day mortality was 12.0% (549 patients): 260 in the control population and 289 in the intervention population (P = .11).

Table 4.

Outcomes in Intervention and Control Groups

Outcome	Control (n = 2353)	Intervention (n = 2300)	P Value
ICU DOT, median (SD)	3 (14.1)	3 (14.2)	.45a
Duration of therapy, d, median (SD)	5 (51.7)	5 (45.0)	.25a
Total LOS (ICU + post-ICU), d, median (SD)	10 (18.3)	9 (19.8)	.11a
ICU LOS, d, median (SD)	4 (10.0)	4 (10.1)	.23a
ICU readmission, No. (%)	62 (2.6)	60 (2.6)	.93b
Post-ICU Clostridioides difficile infection, No. (%)	24 (1.0)	23 (1.0)	.99b
30-d mortality, No. (%)	260 (11.1)	289 (12.4)	.11b
Medical ICU	72 (3.1)	77 (3.3)
Surgical ICU	41 (1.7)	53 (2.3)
Cardiac ICU	78 (3.3)	83 (3.6)
Neurology ICU	36 (1.5)	41 (1.8)
Cardiothoracic ICU	33 (1.4)	35 (1.5)

Abbreviations: DOT, days of therapy; ICU, intensive care unit; LOS, length of stay; SD, standard deviation.

^aWilcoxon rank-sum test.

^bFisher exact test.

The RR of AU in DOT per 1000 days present during ICU stay and following transfer out of the ICU in the intervention vs control groups was 0.97 (95% confidence interval [CI], .91–1.04). We saw large variation in effect size when analyses were stratified by unit (Table 5), with the largest reductions observed in the surgical ICU and the smallest reductions seen in the neurologic and cardiothoracic ICUs. When we excluded the unit with the highest percentage of patients excluded from ASR discussions (CTICU), the RR was significant (0.93 [95% CI, .89–.98]). We evaluated whether order of randomization impacted the effects and found that the interaction term was not significant (P = .46).

Table 5.

Comparing Antibiotic Use in Days of Therapy per 1000 Days Present During Intensive Care Unit Stay for the Remainder of Hospitalization With Rate Ratios Using Multivariate Negative Binomial Regression

ICU Type	Unique Admissions, No.	Rate Ratio (95% CI)	Patients Reviewed on AS Rounds From Intervention Group, no./No. (% Total)
Surgical	992	0.87 (.81–.94)	159/481 (33.1)
Cardiac	1047	0.91 (.86–.97)	191/545 (35.0)
Medical	686	0.94 (.92–.96)	176/359 (49.0)
Neurologic	921	1.05 (.93–1.18)	177/457 (38.7)
Cardiothoracic	1037	1.11 (1.04–1.19)	58/488 (11.9)
All units	4683	0.97 (.91–1.04)	761/2330 (32.7)
All units except cardiothoracic	3646	0.93 (.89–.98)	703/1842 (38.2)

Abbreviations: AS, antibiotic stewardship; CI, confidence interval; ICU, intensive care unit.

AU decreased in all 5 ICUs overall when examined over time (Figure 2). Specifically, the median AU in the poststudy period decreased by 16% (95% CI, 11%–24%) compared to the mean AU in the baseline period. However, the change in AU was differential among units with the largest decrease in the neurologic ICU (28%) and the smallest in the CTICU (2%).

Figure 2.

Antibiotic use measured in days of therapy (DOT) per 1000 days present prior to, during, and after the study period in the 5 study intensive care units.

DISCUSSION

In our study of 4683 ICU patients, weekly ASRs in the ICUs decreased AU. Overall, AU decreased by 16% over time comparing baseline rates to poststudy rates. However, when comparing AU rates among patients randomized to intervention periods compared to concurrent controls, the effect size was small and nonsignificant overall. In subgroup analyses by ICU, however, the effect of ICU ASRs was differential among specific types of specialized adult ICUs. The smallest effects were seen among the highest-complexity CTICU patients, which we believe was largely related to limited interactions with the AS team.

Antibiotic stewardship “handshake” rounds are an increasingly common intervention promoted as successful due to the benefits of in-person discussions and developing long-term relationships with front-line providers [9, 10]. However, formal study of the process and outcomes is rarely described in the medical literature using a randomized design in the ICU setting [11]. To the best of our knowledge, no prior study has looked at AS in the ICU using a cluster-randomized crossover design. Prior quasi-experimental and pre–post design studies have found that routine rounds with ICU teams are associated with declines in antibiotic use [2, 5, 6]. For example, Morris et al conducted a multisite cohort study among 4 Canadian ICUs where ASRs were conducted 3–5 times per week [5]. AU decreased significantly from 120.90 to 110.50 defined daily doses/100 patient-days, a decline of about 8% from baseline. Elligsen et al conducted ICU audit and feedback on the third or 10th day of receiving broad-spectrum antibiotic administration and reduced the mean monthly broad-spectrum AU from 644 to 503 DOT/100 patient-days, a decrease of approximately 22% [2]. Rimawi et al conducted daily ASRs Monday through Friday in 1 medical-surgical ICU [6]; the pre/post analysis reported a decline in overall AU of 7.4%. Taggart et al also reviewed patients daily Monday through Friday in 1 surgical and 1 medical ICU. Mean monthly AU decreased in the surgical ICU by 28%, but interestingly increased in the medical ICU by 14% [7]. Last, Onorato et al conducted ASRs in 2 ICUs 3 times a week. These authors found that AS decreased antibiotic consumption by 324.8 defined daily doses per 100 patient-days (P = .04) and particularly in the use of fluoroquinolones [12]. Thus, prior literature has included shown varied effects of ASRs in both direction and size of the effect.

Our study has several important differences compared to prior investigations. First, the randomized crossover design of our study makes our estimate of effect less susceptible to temporal biases. A before-vs-after assessment of the difference in AU across units would have shown an estimate closer to 20% rather than 10%. Second, our study ASRs occurred less frequently compared to prior studies. We performed weekly ASRs as opposed to ASRs 3–5 times a week or daily in order to spread our team’s efforts across our large institution and 5 different ICUs. During initial planning with ICU leadership, agreement to have a weekly prescheduled time dedicated to ASRs was preferred to disruption during ICU rounds multiple times during the day or throughout the week. Third, our analysis included 5 specialty adult ICUs, which is the largest number of units to date, and demonstrated that the intervention had differential effects among different types of ICUs. Last, our analysis addressed the whole ICU population as opposed to only the ICU patients that were intervened upon, which may have led to a bias toward the null effect.

We designed our analysis specifically to include the entire ICU population and patients who were excluded from ASR review. We aimed to estimate the population-level effects of the intervention for 2 reasons: (1) this is a high-resource intervention that demands significant personnel time to screen and review patients; and (2) we desired to understand both direct and potential indirect effects on ICU populations. Centralized AS teams performing postprescription review can rarely review every patient every day unless they work in a highly resourced or small-practice setting with low patient volumes. Thus, AS leaders must use their limited time, expertise, and opportunities to further antibiotic principles, teach critical clinical decision-making methods, and then trust that these interventions may have future indirect effects on front-line clinician learning for more than just the individual reviewed patient. We believe the “coaching” during rounds and increased awareness of the AS team’s presence on the unit may have led to such indirect effects as well as the declining global trends on the units.

Overall, the unit-level rates of antibiotic use decreased comparing before to after the study period, although the RR comparing intervention-period patients to concurrent controls was not significant. We hypothesize that the intervention had an indirect effect on AU as discussed above. Furthermore, when we removed the ICU with the largest percentage of excluded patients from rounds reviews, we noted not only a 10% decrease in AU, but also statistical significance. AU decreased the most in the neurologic ICU and decreased the least in the CTICU where the lowest proportion of patients was reviewed on ASRs (11.9%). Taggart et al also reported differential effects in their specialized ICUs, with antibiotic use actually going the opposite direction as intended in their medical ICU after the intervention [7]. Rounding with each ICU team led us to appreciate the cultural differences within each ICU, which we believe also led to differential effects. Recognizing the unique needs, communication strategies, and priorities of each ICU helped us tailor the ASR process to each over the duration of the study. In addition, the continuity of the ASP team for each ICU helped build trust between the primary and AS teams that rounded each week. Furthermore, effects on AU were influenced by the population of patients in the unit and the receptiveness or “buy-in” of the primary team to AS recommendations. In units with complex, critically ill patients, we found both reviews and discussions to be long and potentially obtrusive at times. In these cases, providing antibiotic recommendations based on chart review alone were difficult. Thus, for high-complexity populations such as transplant recipients or those similar to our CTICU population, it may be more efficient to embed a steward within the ICU team to work at the bedside or to encourage full ID consultation.

Our study intervention and design were subject to several limitations, including bias, generalizability, and sustainability. Specifically, the study was subject to performance bias. We were unable to blind the ASP and ICU teams to which patients were part of the intervention, which was unavoidable. Moreover, we tried to interact with only one half-unit in each ICU during each study period. However, we were unable to avoid some contamination of the ICU’s 2 half-units. Specifically, if the AS team was asked to comment on a patient in the control group, we did not withhold AS input. In addition, our outcome included AU after a patient was discharged from the ICU. After discharge from the ICU, the patient was no longer subjected to this ASP intervention. Both of these biases could have pushed our results toward the null. Our results may lack generalizability to hospitals that are not tertiary academic medical centers with specialized ICUs. The impact of ASRs may be greater in settings with less patient complexity, such as community hospital settings, and warrants further study. Sustainability is a significant barrier for maintaining this intervention. ASRs are a resource-intensive process for both the ASP and ICU teams. Coordinating rounds for each ICU required planning, communication, and flexibility if emergent issues arose. Efficiency for the AS team was another challenge. Specifically, in several cases the AS team would spend a considerable amount of time reviewing a patient’s chart and be unable to make significant different recommendations to the primary team. Thus, future study is needed to develop selection tools that assist in identifying clinical scenarios that are most likely to result in meaningful stewardship interventions to reduce the personnel time required for reviews.

Based on this experience, we gained a considerable amount of qualitative knowledge about our ICUs, which has been used to adjust our stewardship strategy to be most efficient. First, we increased the frequency of rounds in ICUs where the housestaff do not routinely get significant antibiotic input from attendings and where the culture is more receptive to AS input. We changed the roles for our ASP pharmacists and physicians. Physicians currently focus chart review for diagnostic opportunities and rationale for antibiotic use in the ICU with support from pharmacists for dosing and duration. The stewardship pharmacists focus chart review outside of the ICU with support from physicians for diagnostic challenges.

In summary, we performed a randomized crossover study of AS handshake rounds in the ICU setting. Our effect size was measurable, but small. The effect varied across units and highlighted the importance of customizing AS strategies to match the patient population, workflow, and culture in each unit. Antibiotic stewardship ICU rounding is a high-resource intervention, but effective in reducing AU and building qualitative knowledge and relationships between ICU and AS teams.

Notes

Acknowledgments. The authors acknowledge the critical care pharmacists at Duke University Hospital who were key stakeholders in this effort and significantly impacted the continuity of care of the intensive care unit patients.

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.