2.1 Study design

A retrospective cohort study was carried out by reviewing medical records of patients admitted to the ICU of a tertiary care hospital in Alexandria, Egypt—a tertiary referral hospital with 50 ICU beds—over a period of 4 years, from January 2017 to December 2020, to identify the prevalence, determinants, and changes in antimicrobial resistance patterns among adult medical and surgical critically ill patients with malignancies.

2.2 Inclusion criteria

By reviewing hospital medical records of ICUs, patients were included if:

- They were 18 years and older.

- They had a malignancy, confirmed by pathological and/or cytological tests.

- They had microbiological tests from suspected sites of infection, such as respiratory, urine, blood, aspirates, and swabs collected during their ICU stay.

If a patient had multiple admissions to a hospital ICU on more than one occasion, each admission was included separately in the analysis as an individual episode.

2.3 Data collection

Through patient records and the microbiology laboratory database, including the patient’s demographics, indication for ICU admission, laboratory results, administration of corticosteroids and/or antibiotics within the last 90 days, Charlson Comorbidity Index (CCI) score—used to classify patients according to their comorbidities, such as diabetes, heart failure, immunity profile, chronic obstructive pulmonary disease, cerebrovascular stroke, chronic liver and kidney disease—[12, 13], length of stay, causative organism and infection site, and antibiogram.

2.4 Operational definitions used throughout this study

Neutropenia: Absolute neutrophilic count (ANC) below 500 cells/mm3 or expected to decrease to less than 500 cells/mm3 during the next 48 h [14].

Adequate initial antibiotic therapy: Antimicrobial initiated was proven to have in vitro activity against the infecting strain according to antimicrobial susceptibility test results and the administration route and dosage were determined following current medical standards [15].

Prior antimicrobial therapy: Antimicrobial being administered within the last 3 months before the beginning of the infection episode [16, 17].

Multidrug-resistant bacterial infection (MDR): A gram-negative bacterium that is resistant to three different antibiotic groups. For gram-positive bacteria, methicillin resistance for Staphylococcus aureus (MRSA) and vancomycin resistance for Enterococcus species were considered MDR [18, 19].

Multidrug-resistant Candida infection: An isolate that is non-susceptible to ≥ 1 agent in ≥ 2 drug classes [20].

Septic shock: Sepsis accompanied by reduced organ perfusion and a need for a vasopressor administration to maintain blood pressure [21].

The Charlson Comorbidity Index (CCI) score: A list of 19 comorbid conditions, each having a weight assigned from 1 to 6 [22]. Scores of “1–2,” “3–4,” or “5 or more” are classified into “mild,” “moderate,” and “severe” illness, respectively. Moreover, survival rates over 5 years were 3.4%, 1.3%, and 1.3% in patients with CCI scores of mild, moderate, and severe, respectively [23, 24].

2.5 Methods

During their ICU stay, variable microbiological specimens were collected from cancer patients with purulent respiratory secretions, febrile neutropenia—as neutrophils play a vital role in protecting against infection—symptomatic urinary tract infection, or any of the following alarming signs of infection: fever (temperature > 38 °C), chills, hypotension, or reduced organ perfusion; as well as patients with elevated inflammatory response indicators—e.g., C reactive protein (CRP), procalcitonin, leukocytosis, and leukopenia. Anomalies of the patient’s inflammatory response may signify a higher risk of serious illness.

2.6 Interpretation of antimicrobial susceptibility testing results

Microbiological specimens were collected and cultured on blood, chocolate, and MacConkey agar. Additionally, blood samples were aerobically tested using the BACT/ALERT three-dimensional microbial detection system. Furthermore, bacterial identification and antimicrobial sensitivity tests were performed using the VITEK 2 compact automatic identification system. A modified Kirby Bauer disk diffusion method was performed to examine antibiotic susceptibility of ESBL gene-producing organisms and determine susceptibility to some of the antibiotics that were not included in the VITEK 2 antibiotic susceptibility testing (AST) panels.

Most of these approaches provide qualitative results—using the categories of susceptible, intermediate, or resistant—while some also yield quantitative data, given as the minimum inhibitory concentration (MIC) for each antibiotic—defined as the minimal antibiotic concentration that inhibits bacterial growth in a liquid medium.

For the cultures and AST, all the clinical samples were collected and analyzed. In the case of culture growth, the zones of antimicrobial inhibition were measured and interpreted according to the Clinical Laboratory Standard Institute (CLSI) 2017 breakpoints, to identify them as sensitive, intermediate, or resistant [25].

Isolates exhibiting the ESBL gene were identified by applying the double-disc synergy, looking for synergy between cephalosporin and clavulanic acid. ESBL production was detected if a ≥ eightfold reduction was observed in the MIC of a cephalosporin combined with clavulanic acid but not in that of cephalosporin alone [26]. Furthermore, CR for Enterobacteriaceae was considered with an MIC of > 8 for imipenem. Moreover, vancomycin resistance for Enterococcus species was considered with an MIC of > 4 mg/L [25].

2.7 Statistical analysis

The statistical analysis was performed using the Statistical Package for Social Sciences (SPSS) version 20.0. Only the records that satisfied the inclusion criteria were used for data analysis. The descriptive statistics are presented as mean, standard deviation (SD), and median for the quantitative variables. Additionally, categorical variables are presented as frequency and proportions. The chi-square test (χ2) was used for categorical comparisons.

We examined the prognostic factors by univariate analyses; variables with a P value of < 0.05 in the univariate analysis were candidates for multivariate analysis. Logistic multivariate regression analysis was used to develop a clinical prediction model to estimate the probability of developing MDR among malignancy patients in the ICU.

Patient demographics, comorbidities, type of malignancy, malignancy treatment, hospital length of stay, and recent corticosteroids or antibiotic use were analyzed in relation to developing MDR.

All statistical tools were two-tailed and the level of significance was set at p < 0.05. The odds ratios (OR) and 95% confidence intervals (CI) were calculated to evaluate the strength of any association that emerged.