This was a retrospective, single-centre, observational study conducted at the IRCCS Policlinico San Matteo in Pavia, Northern Italy, from January 2019 to October 2024.

The study included patients with rectal colonisation by multi-drug-resistant organisms (MDROs). However, for the final analysis, only Klebsiella pneumoniae carbapenemase (KPC) and New Delhi metallo-β-lactamase (NDM)-producing Klebsiella pneumoniae, vancomycin-resistant Enterococcus faecium (VRE) and carbapenem-resistant Acinetobacter baumannii (CRAB) were considered.

Other MDROs were excluded owing to low colonisation rates or low incidence of BSIs in colonised patients.

Some patients experienced multiple hospitalisations and were colonised by more than one MDRO. Since comorbidities and colonisation status changed across hospitalisations, analyses were performed at the hospitalisation level to allow for a more accurate reflection of the burden of disease.

The IRCCS Policlinico San Matteo is a tertiary-care teaching hospital with a wide range of wards, including intensive care units (ICUs), haematology, surgical and medical wards.

During the study period, the patients admitted to ICU and haematology wards were screened for MDRO carriage on admission and weekly by rectal swab. Conversely, in other medical and surgical wards, screening has been performed only in high-risk patients, such as those transferred from long-term care facilities or other hospitals.

Clinical and microbiological data were retrieved from electronic health records. The following variables were recorded: (i) demographics (age and biological sex), (ii) comorbidities such as cardiovascular diseases, hypertension, diabetes, obstructive pulmonary diseases, chronic kidney diseases, chronic liver diseases, rheumatic diseases, neoplasia, solid organ transplant (SOT), neutropenia, haematological malignancies and bone marrow transplant (BMT), (iii) microbiological data, namely MDRO colonisation status, infection and antimicrobial susceptibility testing (AST) and (iv) clinical outcomes, such as development of BSI, length of stay (LOS) and intra-hospital mortality.

Data were retrieved from the hospital’s data platform, integrating multiple sources: (i) admission/discharge/transfer (ADT) system: demographic data, hospital admission/discharge details and ward of stay as well as (ii) laboratory information system (LIS): microbiological data, including rectal swabs and blood culture (BC) results. A relational database (Oracle SQL) was used to integrate data from rectal swabs and BCs. All data were anonymised in compliance with data protection regulations.

MDRO colonisation was defined as a positive rectal swab for KPC, VRE, CRAB or NDM, while BSI was defined as a positive BC for the same MDRO previously identified in rectal swabs. The same MDRO was defined as the same species and resistance pattern, without molecular identification.

Specifically, BSIs were identified by matching positive BCs with rectal swab results from the same patient within a defined timeframe, namely from 7 days before the swab until hospital discharge. Cases with missing microbiological data were excluded.

Hospital wards were grouped into three categories: (i) ICU, (ii) surgical departments: general surgery, vascular surgery, thoracic surgery, orthopaedics, neurosurgery, otolaryngology, gynaecology and urology and (iii) medical departments: internal medicine, cardiology, pulmonology, nephrology, oncology, haematology, infectious diseases, neurology and geriatrics.

Metabolic syndrome was defined as the presence of diabetes, hypertension or cardiovascular diseases.

Rectal swabs were processed on selective chromogenic media (CHROMID® CARBA, CHROMID® ESBL, CHROMID® VRE bioMérieux, Marcy-l’Etoile France) and incubated at 35–37 °C for 18–48 h. Colony identification was performed by matrix-assisted laser desorption ionisation time-of-flight mass spectrometry (MALDI-TOF MS; Bruker Daltonics, Bremen, Germany) and analysed using the BioTyper version 3.1. Carbapenemase production was confirmed using the NG-Test® CARBA-5 immunoassay (NG Biotech Laboratories, Guipry, France) or the molecular-based test (Xpert Carba-R kit, Cepheid, Sunnyvale, CA, USA). The first isolate was subjected to antimicrobial susceptibility testing (AST) using the BD Phoenix 50™ instrument (Becton Dickinson, Franklin Lakes, NJ, USA). Minimum inhibitory concentration (MIC) values were interpreted according to the European Committee on Antimicrobial Susceptibility Testing (EUCAST). BCs were incubated into the BD BACTEC FX system, positive blood culture bottles were subjected to Gram staining, subculturing, species identification and antimicrobial susceptibility testing (AST) according to laboratory procedures.

Continuous variables were reported as median and interquartile range (IQR), while categorical variables were represented with absolute frequencies and percentages. Logistic regression models were used to assess the association between demographic, clinical and ward factors and the probability of BSI in patients with rectal MDRO colonisation. Starting from the full model (age, sex, ward, metabolic syndrome, obstructive pulmonary diseases, chronic kidney diseases, chronic liver diseases, rheumatic diseases, neoplasia, SOT, oncohaematology and BMT), a stepwise model-building approach was applied for covariate selection, where predictors were sequentially removed on the basis of their contribution to minimising the Akaike information criterion (AIC). Statistical analyses were conducted using the stepAIC function (MASS toolbox) in R (version 4.2.1), and statistical significance was set at p-values < 0.05.

Additionally, further analyses were performed on ‘prototypic patients’. Specifically, for each pathogen and mechanism, we used the estimated model to identify the ‘most favourable’ and ‘least favourable’ patient profiles. Model parameters were analysed, and the ‘most favourable’ patient is defined as having characteristics and comorbidities associated with negative coefficients, thus linked to a reduced risk of infection. Similarly, the ‘least favourable’ patient was defined as having the traits and comorbidities associated with positive coefficients and consequently with an increased risk of BSIs. This approach allowed us to define a range of risk associated with each mechanism, reflecting the variability in propensity to develop BSI across different patient scenarios based on both the MDRO characteristics and the patient profile extremes.

The study protocol was written in accordance with the ethical guidelines of the 1975 Declaration of Helsinki. The SkyNet Database was approved by the local ethical committee (Comitato Etico Pavia, no. prot. 20,210,068,914).