The COVID-19 pandemic, caused by the SARS-CoV-2, exposed the vulnerability of global healthcare systems, compelling governments and institutions to coordinate efforts to respond efficiently and collaboratively to an unprecedented challenge.

The pandemic has been associated with the indiscriminate use of antibiotics, which, according to some authors, may mark the beginning of the “post-antibiotic era” (Kwon and Powderly, 2021; Owoicho et al., 2021). COVID-19 management guidelines recommend initiating antibiotic treatment only when there is a well-founded clinical suspicion of additional bacterial coinfection or superinfection (Sturza et al., 2023). Despite the recommendations, 74.6% of patients admitted to the hospital received antibiotics (Langford et al., 2021). In contrast, a meta-analysis determined that the incidence of bacterial coinfection and superinfection was 3.5% (95% CI, 0.4–6.7) and 14.3% (95% CI, 9.6–18.9) of patients, respectively (Langford et al., 2020). The similarity in symptomatology, as well as in laboratory results and radiographic findings, to community-acquired bacterial pneumonia, the uncertainty surrounding the pandemic, and the absence of treatments with proven efficacy are factors that may have contributed to the excessive prescription of antibiotics in COVID-19 patients. The unjustified use of antimicrobials introduces an avoidable risk of potential adverse outcomes due to the high toxicity potential and long-term negative effects associated with antibiotics (Spernovasilis and Kofteridis, 2021).

Antimicrobial stewardship programs (ASPs) based on non-taxation audits and feedback are designed to provide specific guidance regarding prescribed antibiotics that are considered amenable to optimization (Dellit et al., 2007; Rodríguez-Baño et al., 2012). During the pandemic, many hospital resources were diverted to the care of COVID-19 patients, affecting the usual activities, including those related to ASP (Ashiru-Oredope et al., 2021; Wimmer et al., 2022; Leung et al., 2022). Since the onset of the pandemic, there has been a call to implement high-value ASP to contribute to the identification, selection, and treatment of patients affected by COVID-19 (Huttner et al., 2020; Mazdeyasna et al., 2020; Stevens et al., 2020). Some experiences have been published as the COVASP clinical trial that demonstrates that prospective audit and feedback are effective and safe in optimizing and reducing antibiotic use in adults admitted to hospital with COVID-19 (Chen et al., 2023). We present the results of implementing a specific three-step antimicrobial stewardship program for hospitalized patients with a confirmed diagnosis of COVID-19 (COVID-ASP), based on a pharmacist pre-evaluation followed by prospective audit and feedback.

A quasi-experimental before-and-after study has been designed in which a COVID-ASP was implemented for hospitalized patients with a confirmed diagnosis of infection caused by the SARS-CoV-2 virus. During the ASP period, all prescriptions were audited by ASP members. Due to the high acceptance rates of the recommendations, nearly all patients received optimal antibiotic therapy. Overall, prospective audit and feedback demonstrated a reduction in DOT/1,000 patient-days (105.71 vs. 162.96; p < 0.001) and in the rate of inappropriate prescriptions (15.3% pre-ASP vs. 8.4% ASP; p < 0.001) without impacting the readmission rate or mortality of the intervened patients. The absence of effect on antibiotic use for superinfections likely reflects that most were nosocomial infections routinely managed through infectious diseases consultation within our long-standing ASP program. As these cases were already optimized before the COVID-ASP, the margin for additional improvement was limited. In contrast, the intervention primarily targeted empirical prescribing in suspected coinfections, where diagnostic uncertainty is greater.

At the start date of this study, January 2021, there was no published experience of audit and feedback conducted in COVID-19 patients. Recently, a pragmatic, cluster-randomized, non-inferiority trial was published, which determined that prospective audit and feedback are effective in optimizing and reducing antibiotic use in adults admitted to hospital with COVID-19. Despite the intervention in the COVASP study, DOT/1,000 patient-days remained higher in the intervention cohort than the non-intervention cohort (384.2 DOT vs. 364.9 DOT per 1,000 patient-days), and 53% of the patients received antibiotics during their hospital stay (Chen et al., 2023). In our study, the DOT/1,000 patient-days rate was substantially lower than that reported in the COVASP study. A published before-and-after study that included 298 patients identified an antibiotic use rate of 61.8% in the pre-ASP period, which decreased to 44.4% following the implementation of the ASP intervention (Anderson et al., 2023). Moreover, the rate of antibiotic use in our study was 29.1% in the pre-ASP cohort, suggesting a narrower scope for improvement compared to findings reported in other studies.

In the ASP cohort of our study, patients were older, there were more cases of hematologic malignancy, and a higher proportion of patients were classified as phenotype B and C according to the FEN-COVID scale. Patients in the intervention cohort had lower baseline oxygen saturation, lower lymphocyte count, and higher CRP levels. These data confirm a poorer prognosis for the ASP cohort, which may contribute to a higher prescription rate of antibiotics despite the presence of an ongoing ASP intervention (Berenguer et al., 2020; Izcovich et al., 2020; Zhao et al., 2020; Gutiérrez-Gutiérrez et al., 2021; Martínez-Lacalzada et al., 2021; Khoury et al., 2022; Abdullah et al., 2023). Suranadi et al. (2022), in a retrospective study, determined that antibiotic therapy was associated with higher mortality. In this sense, the use of an antibiotic may represent a surrogate indicator of severity presentation in patients with COVID-19. Patients in the intervention cohort showed a worse prognosis, as assessed by the FEN-COVID-19 scale, a factor that could contribute to an expected increase in antimicrobial therapy prescriptions compared to the ASP cohort. The worse prognosis of patients in the intervention cohort is likely due to changes in hospitalization criteria during the pandemic. Initially, milder cases were admitted due to uncertainty, but as knowledge about the disease improved, only more severe cases were hospitalized (Lee et al., 2022). COVID-19 vaccination data were not considered in the results as the proportion of vaccinated patients was negligible (0.4% pre-ASP vs. 9.7% ASP). In the intervention cohort in our study, remdesivir, tocilizumab, and corticosteroids were used more extensively due to newly available evidence and improved accessibility to these treatments.

The local guidelines of HUVM recommend the empirical use of ceftriaxone in typical pneumonia with admission criteria. In pneumonic coinfection, most prescriptions in both cohorts consisted of ceftriaxone, followed by amoxicillin/clavulanic, with the prescription of piperacillin/tazobactam being infrequent. Finally, the use of combination therapy and azithromycin was noted as anecdotal. Similarly, in bacterial coinfections from all causes, ceftriaxone was the most commonly prescribed antibiotic. In terms of antibiotic prescriptions for pneumonic superinfection, ceftriaxone stands out as the most prescribed antibiotic in both cohorts, followed by piperacillin/tazobactam. The local guidelines of HUVM also recommend ceftriaxone for empirical treatment of early-onset nosocomial pneumonia. Additionally, they suggest empirical treatment with piperacillin/tazobactam for late-onset pneumonia or severe early-onset pneumonia. Similarly, the most commonly prescribed antibiotic for secondary infection from all causes in the pre-ASP cohort was ceftriaxone, followed by piperacillin/tazobactam. In contrast, in the ASP cohort, piperacillin/tazobactam was the most prescribed, followed by ceftriaxone. The higher use of antipseudomonal coverage in the intervention cohort could be attributed to the greater severity of patients, as assessed by the FEN-COVID scale, the higher number of patients with hematologic malignancies, and the resumption of many hospital activities, such as elective surgeries, during the third and fourth waves of the pandemic.

The intervention reduced the rate of antibiotic prescription for suspected coinfection for all causes, as well as the rate of antibiotic use for suspected pneumonic superinfection. The incidence of bacterial infections associated with COVID-19 showed no significant differences between cohorts (coinfection: 1.8% vs. 2.3%; superinfection: 4.2% vs. 4.0%). Despite these reductions, the overall rate of antibiotic use showed no statistically significant differences between the two cohorts (29.1% vs. 26%; p = 0.216), which may be attributed to the relatively low rate of antibiotic prescription compared to data from published cohorts, leaving little room for improvement. The differences observed in our study in terms of coinfections, superinfections, and the rate of antibiotic use were lower than those reported in other studies. In the meta-analysis by Langford BJ et al., 71.9% of patients received antibiotics during hospitalization, with coinfection and superinfection rates of 3.5% and 14.3%, respectively. Similarly, in the COVASP trial, antibiotic use (53%) was much higher than the rate of bacterial infections associated with COVID-19 (4%) (Langford et al., 2021; Anderson et al., 2023; Chen et al., 2023). Considering that 32.3% of the recommendations in the ASP cohort advised discontinuing antibiotics due to being unnecessary, and given the high acceptance rate, it is likely that the antibiotic use rate would have been lower if measured after the intervention. We have no record of published studies identifying factors associated with the inappropriate use of antibiotics in COVID-19, which could be valuable for guiding ASP activities.

Similarly to findings in other published studies, no significant differences were observed in the incidence of CDI (Anderson et al., 2023; Chen et al., 2023). Studies with a larger number of patients and longer follow-up periods are necessary to determine its effect on this variable. ASPs have been proven to be safe in hospitalized patients, including critically ill adults and pediatric patients (Davey et al., 2017; Lindsay et al., 2019; Chorafa et al., 2023). Similar to the COVASP trial, our study demonstrates that prospective audit and feedback in COVID-19 patients did not impact 30-day mortality or readmission rates.

Compared to the COVASP study, similar recommendations were made. Notably, in our study, 32.3% of the recommendations suggested discontinuing antibiotic therapy due to its being unnecessary, compared to 57.0% in the COVASP study. The acceptance rate of recommendations was 84% in the COVASP study, while in our study, it was as higher at 97%. The high acceptance rate likely reflects our hospital’s long-standing ASP culture and established ID–Pharmacy collaboration. Nevertheless, the intervention relies on simple audit-and-feedback procedures that are easily scalable and do not require substantial resources, supporting its applicability in hospitals with less mature ASP structures.

In addition to comparing process indicators of the ASP, the study also included the evaluation of outcome indicators, such as the incidence of CDI, mortality, and readmission rates. In the intervention cohort, an audit-and-feedback process was conducted for all antibiotic prescriptions. The cohorts represent two distinct time periods, ensuring the absence of contamination between the groups. The high acceptance rate of the recommendations (>95%) indicates that nearly all patients in the intervention cohort received treatment in accordance with the ASP.

Our study presents some limitations. This is a single-center quasi-experimental study comparing two patient cohorts at different times during the pandemic, with the inherent limitations of such studies. In addition to the direct effect of the intervention, other unexplored factors may have influenced prescribing behavior. Although multivariate adjustment was performed using multiple indicators of baseline severity, the quasi-experimental before-and-after design cannot eliminate the risk of residual confounding. Given that the ASP cohort exhibited a higher clinical severity at baseline, it is possible that unmeasured differences between periods could partly influence the observed associations. However, the results across adjusted models support the robustness of the intervention effect. Although most multivariable models met the recommended events-per-variable ratio (Vittinghoff and McCulloch, 2007), the model assessing inappropriate antibiotic use in superinfection included a limited number of events. Therefore, these results should be interpreted cautiously, and residual overfitting cannot be completely excluded. In addition, the results of the sensitivity analysis, conducted to address potential residual confounding arising from differences in baseline severity, showed consistent findings after stratification by age and FEN-COVID phenotype. The results remained consistent across strata, except for underrepresented subgroups (phenotype A and antibiotic prescribing for coinfection in patients younger than 65 years). Additionally, the Hawthorne effect may have altered prescription patterns. Antibiotic treatment audits included a subjective component, conducted retrospectively in the pre-intervention cohort and prospectively in the intervention cohort. Given the absence of a gold standard for initiating antibiotic therapy in COVID-19 patients, prescriptions were assessed using clinical, analytical, microbiological, and radiological criteria, alongside local hospital guidelines. The pre-intervention cohort was evaluated retrospectively, whereas the ASP cohort was assessed prospectively. Although both periods were reviewed by the same ASP team using standardized criteria and a paired-review approach, some variability related to the different modes of assessment cannot be fully excluded. The economic impact of these interventions remains unassessed.

As the pandemic progressed and data on rates of coinfection and superinfection associated with COVID-19 became available, the proportion of patients receiving empirical antibacterial therapy decreased (Sili et al., 2024). The evolving phases of the pandemic likely influenced antibiotic prescribing by altering diagnostic capacity, clinical experience, and treatment protocols. Because the pre-ASP period already encompassed both early and intermediate phases, part of the natural decline in empirical antibiotic use was captured in the baseline data. Despite the higher baseline severity in the ASP cohort, inappropriate prescribing still decreased significantly after the intervention, suggesting that the observed effect is not solely explained by temporal trends. Nevertheless, some residual confounding related to unmeasured pandemic-phase factors cannot be fully excluded.

In conclusion, despite the challenges posed by the COVID-19 pandemic, our hospital, building on its long-standing antimicrobial stewardship tradition, implemented a targeted ASP for hospitalized COVID-19 patients during the third and fourth waves. This three-step, non-restrictive strategy, based on pharmacist-led pre-evaluation and follow-up combined with audit and feedback from infectious disease specialists, was associated with a reduction in antibiotic consumption (DOT/1,000 patient-days), overall antibiotic use, inappropriate prescribing, and treatment duration, while maintaining patient safety. These findings suggest that the ASP likely contributed to optimizing antibiotic therapy in this setting. Moreover, identifying factors associated with antibiotic prescription and inappropriate use may help inform the development of targeted ASP strategies in similar clinical contexts.