Reducing the use of empiric antibiotic therapy in patients on admission to the hospital with COVID-19

Background: Empiric antibiotics for community acquired bacterial pneumonia (CABP) are often prescribed to patients with COVID-19, despite a low reported incidence of co-infections. Stewardship interventions targeted at facilitating appropriate antibiotic prescribing for CABP among COVID-19 patients are needed. We developed a guideline for antibiotic initiation and discontinuation for CABP in COVID-19 patients. The purpose of this study was to assess the impact of this intervention on the duration of empiric CABP antibiotic therapy among patients with COVID-19. Methods: This was a single-center, retrospective, quasi-experimental study of adult patients admitted between 3/1/2020 to 4/25/2020 with COVID-19 pneumonia, who were initiated on empiric CABP antibiotics. Patients were excluded if they were initiated on antibiotics >48hours following admission or if another infection was identied. The primary outcome was the duration of antibiotic therapy (DOT) prior to the guideline (March 1 to March27, 2020) and after guideline implementation (March 28 to April 25, 2020). We also evaluated the clinical outcomes (mortality, readmissions, length of stay) among those initiated on empiric CABP antibiotics. Results: A total of 506 patients with COVID-19 were evaluated, 102 pre-intervention and 404 post-intervention. Prior to the intervention, 74.5% (n=76) of patients with COVID-19 received empiric antibiotics compared to only 42% of patients post-intervention (n=170), p<0.001. The median DOT in the post-intervention group was 1.3 days shorter (p<0.001) than the pre-intervention group, and atypical antibiotic DOT was reduced by 2.8 days (p<0.001). More patients in the post-intervention group were initiated on antibiotics based on criteria consistent with our guideline (68% versus 87%, p=0.001). There were no differences between groups in terms of clinical outcomes. Conclusion: Following the implementation of a guideline outlining recommendations for initiating and discontinuing antibiotics for CABP among COVID-19 inpatients, we observed a reduction in antibiotic prescribing and DOT. The guideline also resulted in a signicant increase in the rate of guideline-congruent empiric antibiotic initiation.


Introduction
Patients admitted to the hospital with coronavirus disease 2019 (COVID-19) are often prescribed empiric antibiotic therapy for possible community-acquired bacterial pneumonia (CABP), as presenting symptoms are di cult to distinguish between viral or bacterial etiologies. However, the widespread prescribing of empiric antibiotics for possible bacterial pneumonia is not well supported by available literature regarding co-infections in the setting of COVID-19. [1][2][3][4][5] A recent review identi ed that despite a low incidence (8%) of reported co-infections among patients with COVID-19, 72% of patients receive antimicrobial therapy. 1 Though initiating empiric antibiotics for CABP may be reasonable, antibiotic therapy should be re-evaluated once COVID-19 pneumonia is con rmed. Prescribing empiric antibiotics when the clinical presentation is inconsistent with bacterial pneumonia or continuing antibiotics longer than necessary should be avoided in order to minimize the potential for adverse consequences.
It is well established that antibiotic use results in increased rates of antimicrobial resistance and increased risk of antibiotic-associated complications such as Clostridioides di cile infection and antibiotic related toxicities. 6 Antimicrobial stewardship interventions may facilitate avoidance of unnecessary antibiotic prescribing among patients with COVID-19 and help front-line clinicians make decisions regarding appropriate initiation and de-escalation of antibiotics for CABP based on laboratory data and chest imaging. In an effort to reduce unnecessary prescribing of empiric antibiotics among COVID-19 inpatients at The University of Chicago, the Antimicrobial Stewardship Program (ASP) and Infectious Diseases COVID-19 Consult Service developed guidance for antibiotic initiation and discontinuation. The purpose of this study was to determine the impact of this intervention on the prescribing of antibiotics for CABP among COVID-19 patients.

Methods
This single-center, quasi-experimental, retrospective cohort study was conducted at an 811-bed academic medical center in Chicago, IL, USA. All adult patients admitted with COVID-19, con rmed by SARS-CoV-2 testing (nasopharyngeal swab), between March 1, 2020 and April 25, 2020 who received at least one dose of empiric antibiotics for CABP initiated within 48 hours of admission were included. Patients were excluded if another source of infection was identi ed that was not pneumonia for which antibiotics were indicated and initiated. This project received a formal Determination of Quality Improvement status according to University of Chicago Medicine institutional policy. As such, this initiative was not considered human subjects research and was therefore not reviewed by the Institutional Review Board.
On March 27, 2020, the Antimicrobial Stewardship Program (ASP) in conjunction with ID providers outlined recommendations regarding when to initiate antibiotics for possible bacterial pneumonia and when to discontinue empiric antibiotics among patients with COVID-19 ( Figure 1). These recommendations were incorporated into the institution's inpatient COVID-19 management guideline.
Indications to initiate empiric antibiotics for CABP included the presence of leukocytosis, fever, or chest imaging suggestive of a bacterial process. The guideline also included recommendations for ordering a respiratory bacterial and viral panel (RBVP; Bio re Diagnostics FilmArray® respiratory Panel, Biomerieux, Salt Lake City, UT), a Legionella urinary antigen, and a Streptococcus pneumoniae urinary antigen. Discontinuation of atypical coverage (e.g. azithromycin, doxycycline, or levo oxacin) was recommended in patients with a negative Legionella urinary antigen and a RBVP negative for atypical bacterial pathogens. Additionally, discontinuation of antibiotics prescribed for CABP (e.g. ceftriaxone or cefdinir) was recommended in patients with negative RBVPs for non-atypical bacterial pathogens and a negative Streptococcus pneumoniae urinary antigen.
Throughout the study period, recommendations and education regarding antibiotic use among COVID-19 inpatients were given to COVID unit providers during daily virtual rounds with the ID COVID-19 Consult Service. 7 Education was also provided to emergency department (ED) staff. All admitted patients with con rmed COVID-19 received an automatic ID consult for evaluation of antibiotic therapy in addition to COVID-speci c management. Each ID consult team included an ID/ASP pharmacist who, along with the ID providers, evaluated each patient case. After updating the institution's guideline to include recommendations for CABP, this evaluation also included a standardized and targeted stewardship intervention to recommend obtaining an RVBP, Streptococcus pneumoniae urinary antigen and/or Legionella urinary antigen (if not performed on admission), along with recommendations to discontinue or de-escalate antibiotics for CABP, in accordance with the institutional guideline.
The primary endpoint was the median duration of antibiotic therapy for CABP between two time periods during the COVID-19 pandemic, March 1 to March 27 (pre-intervention) and March 28 to April 25 (postintervention). Secondary endpoints included hospital length of stay (LOS), 30-day readmissions (for suspected bacterial pneumonia, based on documentation of antibiotic indication), inpatient mortality (allcause), re-initiation of antibiotics following discontinuation during the same admission (for any indication or speci cally for suspected pneumonia based on documentation of antibiotic indication), and rates of Clostridioides di cile infections. C. di cile infection was de ned as a positive C. di cile test in conjunction with symptoms of diarrhea requiring treatment.
Categorical data were analyzed with a Fisher's exact test or a Chi-Square test. Continuous data were analyzed by the Shapiro-Wilk test to determine if the data were normally distributed. Continuous data were analyzed with Student's t-test for parametric data or a Mann-Whitney U Test for non-parametric data. The signi cance level for all tests were set at alpha = 0.05. All statistical analyses were performed with STATA®, version 16, College Station, TX.

Results
A total of 506 inpatients with COVID-19 were screened for inclusion (102 patients in the pre-intervention and 404 patients in the post-intervention). One hundred and fty-ve patients were excluded because they did not receive any antibiotics during the admission, 80 patients were excluded because a source of infection other than pneumonia was identi ed during the admission, and 24 patients were excluded because antibiotics were initiated greater than 48 hours following admission. A total of 246 patients received empiric antibiotics for CABP and were included in the antibiotic duration and clinical outcomes analysis, with 76 patients in the pre-intervention group and 170 in the post-intervention group. Baseline characteristics are shown in Table 1. More patients in the post-intervention group had a fever (55% vs. 25%, p=0.001) and leukocytosis (24% vs. 7%, p=0.002) at the time of antibiotic initiation, while more patients in the pre-intervention group required mechanical ventilation within the rst 24 hours of admission (11% vs. 24%, p=0.01) and at any point (17% vs. 36%, p=0.02) during the hospital course. Overall, there were a total of 11 (4.5%) non-SARS-CoV-2 respiratory pathogens identi ed by either RBVP, respiratory culture, or Streptococcus pneumoniae urinary antigen tests.
The median antibiotic duration of therapy in the post-intervention group was 1.3 days shorter (2.3 versus 1 day, p<0.001) than the pre-intervention group, and the duration of atypical antibiotic coverage (azithromycin, doxycycline, levo oxacin) was reduced by 2.8 days, (3.8 versus 1 day, p<0.001). (Table 2) There was no difference between groups in terms of Clostridioides di cile infections, the need for antibiotic re-initiation, all-cause readmission rate, mortality rate, or length of stay (Table 2). One patient in the pre-intervention group (1.3%) and 3 patients (1.8%) in the post-intervention group were readmitted for suspected bacterial pneumonia (p>0.99). Six (8%) patients in the pre-intervention group and 24 (14%) in the post-intervention group were reinitiated on antibiotics (p=0.24). The reason for reinitiating antibiotics was documented as hospital acquired pneumonia in 2 (2.6%) patients pre-intervention and 15 (8.8%) patients in the post-intervention group (p=0.1). No patients in the pre-intervention group and one (0.6%) in the post-intervention were reinitiated on antibiotics for the indication of CABP (p=>0.99).

Discussion
Following the implementation of a guideline outlining antibiotic use for bacterial pneumonia among COVID-19 inpatients, we observed a 32.5% absolute reduction in antibiotic prescribing and a 1.3-day shorter duration of therapy. The reduction in duration of therapy was most pronounced with antibiotics targeted at atypical pathogens (e.g. azithromycin, doxycycline, levo oxacin) likely due to guideline recommendations to utilize the RBVP panel and Legionella urinary antigen results to aid in de-escalation decisions. Of note, at no point was the use of azithromycin recommended as part of the institution's COVID-19 treatment guideline (in the absence of possible bacterial pneumonia). As every patient received an ID consultation, we were able to monitor this practice directly. Similar to previous reports 1-4 , we observed a high rate of antibiotic prescribing (49%, 246/505) among patients admitted with COVID-19 despite available data suggesting that bacterial co-infection is uncommon among patients with the disease. Our data similarly re ects low rates (4.5%) of co-infection with bacterial pathogens which further supports a need for stewardship interventions to reduce antimicrobial prescribing in this patient population. We found that guideline implementation reinforced by ID COVID-19 Consultation Service recommendations was able to ll this need and increase appropriate antibiotic initiation and deescalation for CABP in this population.
We observed a higher percentage (non-statistically signi cant) of patients being re-initiated on antibiotics in the post-intervention group for the indications of hospital-acquired or ventilator-associated pneumonia. Whether the continuation of empiric antibiotics initiated within 48hours of admission for CABP would have prevented the need to reinitiate antibiotics in these patients is unclear, though unlikely as antibiotics initiated for CABP would have likely been narrower in spectrum than what would be necessary to treat nosocomial pathogens.
Other targets for antimicrobial stewardship interventions include duration of therapy, guideline concordant selection of antibiotics, and intravenous to oral conversion of antibiotics. [8][9][10][11][12][13][14][15][16][17][18][19] Stewardship interventions targeting these aspects of antibiotics for the indication of CABP have been found to be associated with reduced length of stay, 12,18 improved concordance with guideline recommended management (antibiotic selection and duration), [13][14][15][16]19 reduced duration of IV antibiotics, 10,17,18 and fewer adverse drug reactions. 12 While the bene ts of stewardship interventions for patients with CABP in general has been shown in these previous studies, this is the rst study to evaluate the impact of ASP on antibiotic use for CABP among patients with COVID-19.
To our knowledge, this is the rst report of an antimicrobial stewardship intervention to reduce the prescribing of empiric antibiotics for CABP in COVID-19 patients. Reductions in antibiotic use have important implications and can potentially reduce antimicrobial resistance and antibiotic-related toxicities. 2 Several previous studies have evaluated the impact of stewardship interventions on CABP therapy among the general population. Similar to our ndings, most of these studies found no difference in clinical outcomes, suggesting a lack of harm with reduced antibiotic use. [8][9][10] Furthermore, two previous studies have identi ed a mortality bene t with antibiotic de-escalation in the setting of CABP (15.1% vs. 25%, p=0.04 and 1.8% vs. 5.5%, p=0.04), and one found a signi cantly reduced length of stay (5 vs. 9 days, p<0.001). [10][11] Additional ndings that support the safety of reduced antibiotic prescribing in our study include similar rates of antibiotic re-initiation and readmission between groups.
There are a few pertinent limitations to outline. Given the quasi-experimental study design, there are several confounders that may have contributed to the study results. First, the higher rate of mechanical ventilation in the pre-intervention group suggests the disease severity at baseline may have differed between groups. However, this difference may be attributed to changes in critical care practice in utilization of non-invasive ventilatory interventions such as proning, high ow nasal cannula, and helmet ventilation. 20 Additionally, more patients in the post-intervention group had fever and leukocytosis, which also speaks to impact of the intervention in terms of facilitating appropriate initiation of empiric antibiotics based on the presence of fever and/or leukocytosis. Second, after several weeks of managing COVID-19 inpatients, there was likely improved clinician comfort with COVID-19 management as well as more data suggesting low concern for bacterial co-infection during the post-intervention period. Third, changes in SARS-CoV-2 testing may have resulted in a reduced turnaround time in the post-intervention period. Although the timeliness of the SARS-CoV-2 test result may not have had a direct impact on prescribing empiric antibiotics, this may have contributed to a longer duration of antibiotics in the preintervention period. Lastly, this study was underpowered and not designed to investigate clinical outcomes such as adverse drug effects, mortality, and length of stay.
In conclusion, a targeted clinical guideline implemented by an ASP/ID COVID-19 consult service was an effective tool to reduce inappropriate prescribing of antibiotics for CABP in patients with COVID-19 pneumonia. Additional studies are needed to further explore the potential clinical impact of stewardship interventions targeting prescribing of antibiotics for CABP among patients with COVID-19.

Declarations
Ethics approval by the University of Chicago Medical center and consent to participate has been obtained from all authors. This project received a formal Determination of Quality Improvement status according to University of Chicago Medicine institutional policy. As such, this initiative was not considered human subjects research and was therefore not reviewed by the Institutional Review Board.

Consent for publication: Obtained from all authors
Availability of data and materials: All de-identi ed data can be made available upon reasonable request Competing interests: None Funding: None Authors' contributions: All authors contributed to protocol design, data review, and manuscript preparations. A. Lew, C. Nguyen, N. Pettit, P. Bhagat, A. Nelson also completed data collection. N. Pettit completed the statistical analysis.