This study is the continuation of a first study that took place at Lausanne University Hospital, a 1500-bed tertiary university hospital in Lausanne, Switzerland. The study setting has been previously described in details elsewhere [2].

We conducted a retrospective observational single-center study. All consecutive adult patients with P. aeruginosa isolated from a clinical sample between 01.08.2019 and 31.07.2021 were identified. Those who received an antibiotic for a P. aeruginosa infection and that could be treated either by ceftazidime, cefepime and/or piperacillin-tazobactam based on susceptibility testing results available in the Electronic Medical Record (EMR) were included in the study. We excluded patients with a P. aeruginosa isolate resistant to meropenem, and those with a P. aeruginosa infection that could not be treated by ceftazidime, cefepime or piperacillin-tazobactam due to allergy. We also excluded patients with a polymicrobial infection requiring a treatment with a carbapenem, including ESBL-producing Enterobacterales co-infections.

Three periods were defined: the first from 01.08.2019 to 26.01.2020-patients treated “before EUCAST update” (period 1), the second from 27.01.2020 to 20.12.2020—patients treated “after EUCAST update without selective reporting” (period 2) and the third from 21.12.2020 to 31.07.2021—patients treated “after EUCAST update with selective reporting” (period 3).

Epidemiological, clinical and microbiological data were extracted from the EMR. Data collection for patients meeting the same criteria from 01.08.2019 to 31.07.2020 had been done in the previous study [2]. Epidemiological data included age, sex, and relevant comorbidities. We also collected data on microbiology results, antimicrobial therapy, stay in intensive care, infectious diseases specialist (IDs) consultations, and other clinical aspects: site and severity of infection, community versus healthcare-associated infection—including vascular catheter-associated infections, catheter-associated urinary tract infections, ventilator-associated pneumonia, surgical site infections and infections occurring more than 48 h after admission to hospital. We entered all the data in an electronic clinical report form (eCRF) using the Redcap® platform (Research Electronic Data Capture v10.3.3, Vanderbilt University, Tennessee, USA).

In case of multiple episodes of P. aeruginosa infection, patients included a first time before the implementation of selective reporting could be included once again after selective reporting.

The primary outcome was meropenem prescription as targeted treatment (i.e. after P. aeruginosa susceptibility testing release). We took into consideration the antipseudomonal antibiotic initiated after the susceptibility testing results have been made available. For patients receiving empiric antipseudomonal antibiotic therapy initiated before susceptibility testing results, we took into consideration the ongoing antipseudomonal antibiotic 24 h after the susceptibility testing results have been made available.

Secondary outcomes were the use of increased dosage for non-meropenem anti-pseudomonal drugs, and IDs consultation rates after susceptibility testing results have been made available.

For the descriptive analysis, we summarized categorical variables as numbers (percentages), normally distributed continuous variables as mean ± standard deviation (SD), and continuous variables with a skewed distribution as median [interquartile range (IQR)]. Between-group comparisons were performed using chi-square or Fisher’s exact test for qualitative variables, and Student’s t-test, analysis of variance or Kruskal–Wallis test for quantitative variables.

Analyses for our primary outcome were performed through uni- and multivariable logistic regression models. Models were built manually, adding demographic, and clinical characteristics. We then added variables with a P value below < 0.2 from our univariable analysis. Models were built based on Akaike Information and Bayesian Information Criteria. We calculated Odds ratio (OR) with 95% confidence interval (95% CI) to determine the weight of risks factors for meropenem prescription. P values < 0.05 were considered as statistically significant. Due to the important number of dependent events and the small number of meropenem prescriptions, especially for period 1, these models did not allow us to perform a robust analysis. Therefore, we report only descriptive results. We used Stata SE 17.0 (StataCorp, College Station, TX) for all analyses.