This was a retrospective case–control study of patients with active infections caused by CPE and non-MDR bacteria between January 2010 and October 2022. The research was carried out at the Parc de Salut Mar, in Barcelona (Spain).

During this time, every positive culture from a clinical sample showing growth of CPE isolates was examined; only those patients with criteria for infection were included as cases. Patients with a positive culture but no signs of infection were considered as colonised and excluded. Inclusion criteria were age 18 years or older, isolation of CPE in a clinical sample, and signs and symptoms of infection in the same focus as the culture. Outpatients were only included if they required hospitalisation due to the infection. Only the first episode of infection per patient was considered. Exclusion criteria were: (i) polymicrobial infections, (ii) colonisation without active infection, and (iii) end-of-life patients (palliative care patients) who did not receive active therapy for the infection.

The control group was made up of patients with active non-MDR bacterial infections. The matching criteria were based on controls and cases sharing a susceptible strain of the same bacterial species, focus of infection, and same sex and age ± 5 years.

As the initial control groups did not reach the sample size required for statistical analysis when all four matching criteria were combined (see further), the requirement for control inclusion was relaxed to both the same focus of infection and susceptible strain of the same microorganism. Matching by age and/or sex was applied whenever possible.

Sample size determination was based on the results of a previous study [18] to detect a 10% difference in 30-day mortality between CEP and non-MDR bacteria; statistical power was set at 80% and alpha error at 0.05. The sample size ratio for matching was 1:2 and 82 cases and 164 controls were obtained.

Patients were followed for 90 days from the onset of the infection. The follow-up was carried out through the electronical clinical records, which integrates hospital, primary care, homecare, long-term facilities, and social services records in the healthcare region of the hospital setting.

The primary outcome variable was all-cause mortality at 30 days, considering day 1 as the first day of positive culture. Secondary outcomes were clinical failure at day 7 (early clinical failure) and end of treatment, crude mortality at day 90, hospitalisation costs and resource consumption.

Information on the study variables was obtained from an electronic medical record and economic data through the hospital’s economic database.

The variables considered for the clinical impact study were age, gender, underlying diseases according to the Charlson comorbidity index [19], antibiotic treatment in the previous 3 months, hospitalisation, invasive devices (urinary catheter, vascular access) or dialysis, microorganism isolated, source of infection, empirical and definitive treatment, duration, and antibiotic-related side effects.

An assessment of clinical severity was made for septic shock and intensive care unit (ICU) admission in the first 72 h, as well as a Sequential Organ Failure Assessment (SOFA) [20]. Septic shock was defined according to the Sepsis-3 definition [21]. In case of bacteraemia, the Pitt score [22] was applied.

Clinical failure was defined as persistence or worsening of signs and/or symptoms of the infection and/or death. Empirical antibiotic therapy was considered appropriate when at least one antibiotic active in vitro active antibiotic was given against the microorganism was administered.

For the consumption of healthcare resources, several variables were considered: length of stay, need for outpatient parenteral antibiotic treatment and long-term care facilities, and readmissions at 90 days of follow-up. Finally, the overall cost of hospitalisation, pharmacy and diagnostic tests for each patient were also assessed.

MDR was defined as resistance to at least one agent in three or more antimicrobial categories for each organism, according to current standard definitions [23]. Non-MDR was considered when resistance to fewer than three active antimicrobial categories was observed. According to this classification, the antimicrobial categories used to define MDR in Enterobacteriales were: aminoglycosides, anti-MRSA cephalosporins, antipseudomonal penicillins + β-lactamase inhibitors, carbapenems, 1st and 2nd generation cephalosporins, 3rd and 4th generation cephalosporins, cephamycins, fluoroquionolones, folate pathway inhibitors (trimethoprim-sulphamethoxazole), glycylcyclines (tigecycline), monobactams, penicillins + β-lactamase inhibitors, phenicols (chloramphenicol), phosphonic acids, polymyxins and tetracyclines [19].

Routine identification and susceptibility testing of causative microorganisms was performed using automated systems (the Vitek-2® [BioMérieux] for blood cultures, and the MicroScan® WalkAway [Beckman-Coulter] for other sample types). Results were interpreted according to the European Committee on Antimicrobial Susceptibility Testing (EUCAST) standards in force at the time of culture. Carbapenemases were identified by multiplex PCR assay, using the LightMix® Modular Carbapenemase panel (Roche Diagnostics).

Qualitative variables were shown as numbers and percentages, and quantitative variables as median and interquartile range (IQR). Quantitative variables were analysed using the Student’s t-test or Mann–Whitney U test, and qualitative variables using the Chi-Square (χ2) or Fisher’s exact test, as appropriate.

A logistic regression model was used to examine variables associated with clinical failure at day 7 and at end of treatment, expressing the results as odds ratio (OR) and 95% confidence interval (CI).

Kaplan–Meier survival curves and the log-rank test were used to detect differences in 30- and 90-day mortality.

The Cox proportional hazards model was used to perform multivariate survival analyses for 30- and 90-day mortality. Results were expressed as hazard ratio (HR) and 95% CI. The proportional hazards assumption was tested.

To control for confounding, variables in the crude analysis that were associated with exposure (p ≤ 0.20) or outcome (OR or HR < 0.65 or > 1.5), and those that were not considered to be intermediate variables between exposure and outcome were candidates for multivariate analysis. Variables leading to substantial confounding if removed (10% or more change in the coefficient estimate) were retained in the model, along with exposure (CPE group). Manually selected, backward stepwise regression was applied. Variables with > 25% missing values were not considered in multivariate analysis.

For the economic analysis, median regression (to deal with non normal dependent variables) was used. Results were expressed as the difference in medians (DM). Interpretation of the coefficients was based on the difference of means, as it was for the interpretation of coefficients in multiple linear regression.

All p-values were 2-tailed and statistical significance was set at ≤ 0.05. Statistical analyses were performed using STATA 15.1. STROBE guidelines were used for the reporting of the study (Supplementary Table S1).

The Clinical Research Ethical Committee of the Parc Salut Mar approved this study (registration no. 2022/1046). Written informed consent was waived due to its observational and retrospective nature.