We designed a retrospective study including patients from the five ICUs of the AC Camargo Cancer Center (São Paulo–Brazil).

For epidemiological surveillance, at the beginning of May 2017, the Hospital Infection and Prevention Control Department ordered a once-a-week rectal swab for routine screening of CR-GrN bacteria and VRE. Briefly, two separate rectal swabs were collected. For VRE detection, one swab was plated on ChromID® VRE (bioMérieux, Brazil). For carbapenem-resistant enterobacterales and acinetobacter detection, the other swab was plated on CHROMagar mSuperCarba (Plastlabor, Brazil). Both plates were incubated in air at 35 ± 1 °C for a maximum of 48 h. If colonies were observed, identification was carried out using the Microflex mass spectrometry system (Bruker, Germany) (further details on Additional file 1).

Every Tuesday, all patients admitted for more than 48 h in the ICU were screened, excluding patients at end-of-life care. There was no routine screening for MDRB colonization outside ICU.

Between May 2017 and May 2019, we included all adult patients (≥ 18 years old), with active solid tumors or hematologic malignancies screened for CR-GrN bacteria and VRE during ICU stay. If a patient had multiple ICU admissions, only the first was considered. However, if a patient had multiple ICU admissions during the same hospital admission, and not the first, but a subsequent admission revealed CR-GrN bacteria or VRE colonization, that was the admission included in the study.

Upon ICU admission, patient’s age and sex, days in hospital before ICU admission, Simplified Acute Physiology Score (SAPS 3) [12]; Eastern Cooperative Oncology Group (ECOG) performance status [13]; type of cancer (nonmetastatic or metastatic solid tumors, hematologic malignancies); ICU architectural design (single-bed room or multibed room), type of ICU admission (planned or unplanned); and source of ICU admission (emergency room, ward or operating room) were recorded.

We recorded the result of the rectal swab as follow: negative, positive for CR-GrN bacteria, positive for VRE; or positive for both bacteria. We also recorded the ICU and hospital length of stay.

In patients colonized by CR-GrN bacteria or VRE, we evaluated the conversion of colonization to clinical infection by the same bacteria up until one-year after hospital discharge. The conversion was confirmed when the same CR-GrN bacteria or VRE initially identified in the rectal swab and with the same antibiotic-resistance profile was isolated in any culture and the antibiotic regimen was initiated or changed accordingly. The conversion detection was performed by two investigators that independently examined the electronic health record and microbiological laboratory results.

The main outcomes of the study were in-hospital mortality and one-year survival. There were not missing values.

Categorical and continuous data were presented as absolute values (percentages) and median (25–75% interquartile range), respectively. Categorical variables were compared using the Chi-square test or Fisher’s exact test, as appropriate. Continuous variables were compared with the Mann–Whitney test.

A logistic regression model was used to evaluate the association of colonization with in-hospital mortality. We used directed acyclic graph to identify confounders [14]. We identified the following confounders: age, performance status, SAPS 3 score, type of ICU admission and type of cancer (Additional file 1: Fig. S1).

Odds ratio (OR) with 95% confidence interval was used to measure the association of colonization with in-hospital mortality. We presented the unadjusted and adjusted odds ratios.

We used the multivariate Cox proportional hazards regression to compare the survival curve adjusted for confounders identified using directed acyclic graph. We identified the following confounders: age, performance status, and type of cancer (Additional file 1: Fig. S2). The time zero of the one-year follow-up period was the day of rectal swab screening.

Hazard ratio (HR) with 95% confidence interval was used to measure the association of colonization with one-year survival. We presented the unadjusted and adjusted hazard ratios.

Statistical analyses were performed using SPSS software (Version 23.0. Armonk, NY: IBM Corp). The directed acyclic graphs were created using the browser-based environment DAGitty [15]. We followed the recommendations of the STROBE (Strengthening the Reporting of Observational Studies in Epidemiology). statement that guides the report of observational studies [16] and the guidance for control of confounding and reporting of results in causal inference studies from editors of respiratory, sleep, and critical care journals [14]. P values ≤ 0.05 were considered significant.