Materials were electronic data produced from the format of the Japanese medical insurance system or the Diagnostic Procedure and Combination since March 1, 2019 through February 28, 2022. We admitted the first COVID-19 patient March 3, 2020 and subsequent patients in two dedicated wards for the care using clinical pathways that included antivirals. Non-COVID-19 patients were admitted to remaining eight wards.
Methods of stewardship included intervention and weekly audit of antimicrobials by the infection control team, in accordance with the therapeutic guidance for COVID-19 patients by certified infectious diseases specialists. In line with the guideline , perioperative prophylactic antimicrobials were mostly cefazolin. On the other hand, patients diagnosed as sepsis-3 underwent bacteriological tests and the administration of broad-spectrum antimicrobials in compliance with the surviving sepsis campaign guideline [5, 6]. They were, however, corrected by bacteriological results. Should extended spectrum beta-lactamase (ESBL)-producing microbes be isolated and susceptible for cefmetazole, for example, we recommended susceptible antimicrobials sparing carbapenems.
As an index of the stewardship, antimicrobial use density (AUD) was defined as:
(total dose) / (Defined Daily Dose) / (patient-day) x100
where the Defined Daily Dose was described by the World Health Organization . We collected parenteral monthly AUDs as semi-automated data by the Japan Surveillance for Infection Prevention and Healthcare Epidemiology System (J-SIPHE). To use the data of our institute alone, we obtained written consent from the headquarter.
These AUD data underwent the Kolmogorov-Smirnov tests for the normal distribution, which led into parametric or non-parametric tests for multiple comparisons using analysis of variance with the Bonferroni correction or Kruskal-Wallis tests, respectively. For computation, we used a software SPSS Statistics Version 27 (IBM Japan Inc., Tokyo, Japan).
Outcomes were measured twofold. (1) Clinical and bacteriological events were documented. For COVID-19, we recorded antivirals, fatality, and antimicrobials with their preceding culture tests. (2) AUDs were compared over the years and between the COVID-19 wards and non-COVID-19 wards.
As to Outcome (1), a total of 20,013 patients (7,534, 6,146, and 6,333 for PreY, Pan1Y, and Pan2Y, respectively) were subjected; operations were performed for 6,816, 5,577, and 5,772 cases, sepsis-3 was diagnosed in 152, 132, and 283 patients, ESBL-producing microbes were isolated in 125, 100, and 102 patients and Clostridioides difficile toxin tests were positive in 17, 5, and 7 patients, respectively.
In patients with COVID-19 (N=622) at a median age of 58 (range, 1-99), 11 (1.8%) died. No patients were positive for C. difficile toxin. Parenteral antimicrobials were given in 59 patients (9.5%), which were preceded by microbial tests in 48 (81.4%) or computed tomography to confirm additional aspiration pneumonia. A total of 619 patients (99.5%) underwent clinical pathways that included antivirals, such as favipiravir (N=15, 2.3%), remdesivir (N=205, 33.0%), casirivimab/imdevimab (N=102, 16.4%) , and sotrovimab (N=105, 16.9%).
As for Outcome (2), the categories and detail of antimicrobials are listed in Table 1. To compare AUDs over the years, we used the parametric analyses because 14 out of the 18 hospital-wide AUDs (77.8%) met the normal distribution. The means of the first generation cephalosporins decreased from PreY to Pan2Y (P=0.001) as well as from Pan1Y to Pan2Y (P=0.015). The means of carbapenems’ AUD decreased from PreY to Pan1Y (P=0.001) but increased from Pan1Y to Pan2Y (P=0.001) (Table 2).
Comparing PreY and pandemic years (Pan1Y+Pan2Y), the means of total AUD for all the wards decreased (P=0.050) as did those of the first generation cephalosporins (P=0.011) (Table 3).
In comparison between COVID-19 wards and others, we used non-parametric analyses because all AUDs were out of the normal distribution. The median AUDs in the COVID-19 wards were significantly (P<0.050) less in 22 out of 34 antimicrobials (64.7%).
Our study revealed that the means of AUD for the first generation cephalosporins decreased from the pre-pandemic year to the pandemic years as did the numbers of operative cases. This was most likely derived from the decrease in perioperative prophylactic antimicrobials due to the citizens’ reluctance to visit hospitals during the pandemic. A similar observational study during the pandemic was reported by Gu and others .
Despite the emergence of COVID-19, however, our antimicrobial stewardship worked to reduce carbapenems’ AUD from PreY to Pan1Y. Their subsequent increase from Pan1Y to Pan2Y, on the contrary, may reflect the increased number of sepsis-3 patients necessitating broad-spectrum antimicrobials based upon the Early Goal Directed Therapy . During Pan1Y and Pan2Y, we observed decreased numbers kept for ESBL-producing microbes and for the patients with C. difficile infection. These may have resulted from reducing AUDs of overall antimicrobials as described by Granata and others during the pandemic .
On the mortality of COVID-19 as of February 28, 2022 or the end of our study period, the domestic surveillance  reported the cumulative numbers of deaths and “case-patients requiring inpatient care” were 23,625 and 703,137, respectively. The nationwide mortality thus being 3.4%, ours at 1.8% was low despite lowered AUDs in COVID-19 wards. Likewise, Chan et al  described that, during the pandemic, reduction in broad-spectrum antibiotic use in intensive care unit did not increase mortality.
Nori et al  reported “outpatient pandemic stewardship” for outpatient administration of neutralizing antibody agents. The domestic ministry of health, however, had regulated that these agents be given in-hospital setting. Our in-hospital stewardship prioritized the early administration of antivirals for COVID-19, per se, viral infection, using clinical pathways. Likewise, Bartlett and others  reiterated that clinical pathways for COVID-19 patients would augment antimicrobial stewardship
At the emergence of COVID-19, the main difficulty in the pandemic stewardship resided in previously unknown pathophysiology in the care of patients. No guidelines having been available initially, the fear of co-infection with bacterial pneumonia may have led into overuse of antimicrobials. Prior to the pandemic, however, the diagnostic stewardship did maintain the antimicrobial principle based on the bacteriological diagnosis. After the onset of COVID-19 as well, Rubin and others  stressed to add imaging diagnosis to differentiate from superimposed infection.
The limitation of our study includes being a retrospective study in a single institute. A multi-institutional study before and after the emergence of COVID-19 may be possible using, for example, the big data from J-SIPHE. Secondly, frequently revised guidelines on the use of steroids may have created historical biases on the administration of prophylactic antimicrobials. Thirdly, the time course of AUDs may fluctuate seasonally or by the climate change. Theoretically, therefore, time series analyses may provide better insights into perennial trend of AUDs.