A quasi-experimental analysis comparing antimicrobial usage on COVID-19 and non-COVID-19 wards.

OBJECTIVE

To describe antimicrobial usage (AMU) trends before and during the coronavirus disease 2019 (COVID-19) pandemic, between COVID-19 and non-COVID-19 wards, and if there was any association with a COVID-19 order set.

DESIGN

Quasi-experimental retrospective interrupted time series analysis of AMU rates with a contemporaneous comparison of COVID-19 versus non-COVID-19 control wards. Analysis using incidence rate ratios (IRR) was conducted using a Poisson regression generalized linear model.

SETTING

Five COVID-19 and 4 comparable non-COVID-19 wards and 6 intensive care units (ICUs) at 4 hospitals during pandemic waves 1-4.

PARTICIPANTS

All inpatients receiving systemic antimicrobials.

INTERVENTION

The COVID-19 checkbox antimicrobial order set was implemented in March 2020, to be used only if considered clinically indicated with modification in August 2021.

MAIN OUTCOMES AND MEASURES

The primary outcome was a change in AMU rates (defined daily dose per 100 patient days per month) comparing pre- versus peri-pandemic periods and COVID-19 versus control non-COVID-19 wards. Secondary outcomes included antifungal usage rate in ICUs and assessing AMUs following implementation and modification of a COVID-19 order set.

RESULTS

Significantly greater rates of AMU (IRR[95%CI]) were observed on COVID-19 wards versus non-COVID-19 wards during waves 1-4 for all systemic antimicrobials (1.76[1.71-1.81], 1.10[1.07-1.13], 1.48[1.43-1.53], and 1.06[1.03-1.09]); for azithromycin (11.76[9.80-14.23], 10.96[9.49-12.74], 12.41[10.73-14.45], and 4.88[4.31-5.55]); and for ceftriaxone (2.39[2.16-2.65], 3.64[3.29-4.03], 2.94[2.67-3.23], and 1.62[1.49-1.76]).

CONCLUSIONS

We observed significantly increased AMU rates of all systemic agents during the first 4 waves of the pandemic and on COVID-19 wards compared with control wards for azithromycin and ceftriaxone. These agents saw a twofold reduction following order-set removal, suggesting that the clinical decision-support tool order set, as utilized, had influenced prescribing behavior.