Epidemiology and risk factors of multidrug-resistant Gram-negative bacilli in healthcare-associated infections: development of an exploratory risk assessment model in a Moroccan regional hospital Authors: Asmaa Dihmane, Rafik Aniba, Habiba Raqraq, Amina Ressmi, Kaotar Nayme, Mohammed Timinouni, Abouddihaj Barguigua Published: 2025 ABSTRACT AbstractBackgroundHealthcare-associated infections (HAIs) caused by multidrug-resistant Gram-negative bacilli (MDR-GNB) are a growing public health threat, particularly in hospitals. This study identified risk factors for HAIs due to MDR-GNB in a regional hospital via univariate and multivariate logistic regression analyses.MethodsWe conducted a case–control study at the Regional Hospital of Beni Mellal, Morocco (November 2021–April 2024). Adult inpatients with HAIs due to GNB were included. Cases were defined as MDR-GNB infections, and controls as non-MDR-GNB infections. Demographic, clinical, and exposure data—including prior antibiotic use, hospitalization history, and device utilization—were collected from medical records. Independent risk factors for MDR-GNB infection were identified using multivariate logistic regression. An exploratory bedside risk index was developed and internally validated by bootstrap resampling (1,000 iterations), with discrimination assessed via receiver operating characteristic curve analysis, and calibration was assessed using calibration intercept and slope.ResultsAmong 188 patients with culture-confirmed GNB infections, 56.4% (n = 106) were MDR-GNB. The most frequent isolates were Escherichia coli (45%), Acinetobacter baumannii (27.4%), and Klebsiella pneumoniae (19.8%). Multivariate analysis identified antibiotic therapy within the previous 90 days (aOR = 14.79; 95% CI: 4.6–47.56), prior hospitalization within 90 days (aOR = 3.91; 95% CI: 1.52–10.07), and Foley catheter use (aOR = 3.69; 95% CI: 1.19–11.47) as independent risk factors for MDR-GNB infection. Conversely, bloodstream infection (aOR = 0.09; 95% CI: 0.01–0.52) and E. coli infection (aOR = 0.28; 95% CI: 0.10–0.75) were associated with lower odds of MDR-GNB infection. A five-variable risk assessment score showed excellent discrimination (AUC = 0.88, bootstrap-corrected AUC = 0.87) and acceptable calibration, accurately stratifying patients into low-, intermediate-, and high-risk groups for MDR-GNB infection.ConclusionThese findings highlight the urgent need to implement strict infection prevention protocols, optimize antibiotic use through stewardship programs, and enhance surveillance to combat the spread of MDR-GNB infections at Beni Mellal Regional Hospital.Supplementary informationThe online version contains supplementary material available at 10.1186/s12879-025-12371-7.Keywords: Healthcare-associated infections, Risk factors, Gram-negative bacilli, Multidrug resistance, Logistic regression, Regional hospital of Beni Mellal Introduction Antimicrobial resistance (AMR) is a major global health threat, undermining the effectiveness of standard therapies and contributing to increased morbidity, mortality, and healthcare costs. It is estimated that over 700,000 patients die each year from infections caused by resistant bacteria, and without intensified prevention and stewardship measures, this figure could rise to 10 million annually by 2050 [1–3]. These projections underscore the urgent need to address AMR as a global health priority. Healthcare-associated infections (HAIs) constitute a major driver of AMR, particularly in hospitalized patients, who are frequently exposed to invasive devices, prolonged antibiotic therapy, and opportunities for cross-transmission. Gram-negative bacilli (GNB), including Escherichia coli, Klebsiella pneumoniae, Pseudomonas aeruginosa, and Acinetobacter baumannii, are of particular concern due to their ability to acquire and disseminate multidrug resistance through extended-spectrum β-lactamases (ESBLs), carbapenemases, and plasmid-mediated resistance mechanisms [4, 5]. Multidrug-resistant (MDR)-GNB are now recognized as leading causes of HAIs worldwide, especially in acute hospital settings and long-term care facilities, and have been listed by the World Health Organization as “critical priority” pathogens for research and new drug development [6]. While surveillance data and predictive risk models for MDR-GNB infections are increasingly available from Europe, North America, and Asia [7–11], evidence from North Africa remains limited and fragmented. Existing Moroccan studies have primarily been descriptive, restricted to single hospitals or wards, and have rarely employed multivariable approaches to identify independent risk factors [12–14]. Moreover, predictive scoring systems that could assist clinicians in early recognition of patients at high risk of MDR-GNB infection are largely absent in this context. This knowledge gap hinders the development of evidence-based guidelines for empirical therapy, infection control prioritization, and antimicrobial stewardship in Moroccan hospitals. To address these gaps, we conducted a case–control study at the Regional Hospital of Beni Mellal, Morocco, with the following objectives: (i) to estimate the prevalence of MDR-GNB in HAIs; (ii) to characterize their AMR patterns; (iii) to identify independent demographic and clinical risk factors; and (iv) to develop a simple bedside risk assessment tool to support early recognition and targeted preventive strategies in clinical practice. Materials and methods Study design and setting This study aimed to determine the prevalence, antibiotic resistance patterns, and risk factors associated with MDR-GNB causing HAIs and to develop an exploratory bedside risk assessment tool for predicting MDR-GNB infections among hospitalized patients. It was conducted at Beni Mellal Regional Hospital, a 420-bed facility in the Beni Mellal–Khenifra region of Morocco, serving as a primary referral center with specialized departments in surgery, internal medicine, pediatrics, obstetrics, oncology, and intensive care. The study covered the period from November 2021 to April 2024 and employed a case–control design. All MDR-GNB culture-positive samples from this period were included to capture data and represent HAIs across hospital wards, providing a comprehensive analysis of MDR-GNB infection dynamics (Fig. 1). Study design and sample selection. Abbreviation: MDR-GNB, multidrug-resistant Gram-negative bacilli Sample size determination All available GNB infection cases recorded during the study period were included in the analysis to ensure comprehensive representation of HAIs and resistance patterns within the hospital. Case and control definitions Cases were defined as patients with culture-confirmed MDR-GNB infections meeting the Centers for Disease Control and Prevention and European Centre for Disease Prevention and Control (ECDC) criteria for HAIs [15, 16]. In accordance with the ECDC guidance and the international consensus, MDR-GNB strains were defined as those resistant to at least one agent in three or more distinct antimicrobial classes, with resistance profiles specific to each bacterial species [17]. Controls were patients with cultures positive for GNB that did not meet the MDR definitions—specifically, isolates resistant to fewer than three antimicrobial categories. Inclusion and exclusion criteria Patients admitted to the Beni Mellal Regional Hospital with positive microbial culture results related to HAIs and who underwent antibiotic susceptibility testing were included in this study. If two or more bacteria of the same species with identical resistance profiles associated with HAIs were isolated, only one isolate from each pathological sample taken from the same patient was considered. The analysis was limited to medical records from the Beni Mellal Regional Hospital. Patients with incomplete data or without a medical history at the hospital during the specified period were excluded from the study. In this study, only HAIs were analyzed. Community-acquired infections, which are infections contracted outside the hospital or before admission, were excluded [18, 19]. Data collection To assess the risk factors for acquisition of MDR-GNB associated with nosocomial infections, a standardized form was used to collect data from paper and electronic medical records of patients admitted to Beni Mellal Regional Hospital. The collected variables comprised demographic characteristics (age, sex, employment status) and clinical characteristics (ward of admission, previous hospitalizations, current length of stay defined as the number of days from admission until the date of culture positivity, presence of underlying diseases, use of medical devices, and antibiotic treatments administered within the preceding 90 days, categorized as targeted curative, probabilistic curative, or preventive). Infection sites were classified by anatomical location. Bacterial isolation and identification Bacterial isolates were obtained from clinical samples sent to the medical microbiology laboratory for diagnosis. The samples included urine, bronchoalveolar lavage, blood, catheters, pus, and cerebrospinal fluid. Clinical samples were processed according to the laboratory’s routine diagnostic protocol, which included identification by morphological, biochemical, and culture characteristics. Purified isolates were identified via an analytical profile index gallery (BioMérieux, Marcy-l’Etoile, France). Antimicrobial susceptibility testing The susceptibility test was performed via the diffusion method on Mueller–Hinton agar (Biokar, Beauvais, France), following the recommendations of the Antibiogram Committee of the French Society for Microbiology/European Committee on Antimicrobial Susceptibility Testing [20]. The antibiotics tested included ticarcillin (75 µg/disk), piperacillin (100 µg/disk), tazobactam + piperacillin (36 µg/disk), ampicillin (10 µg/disk), amoxicillin + clavulanic acid (30 µg/disk), cephalothin (30 µg/disk), ceftriaxone (30 µg/disk), cefotaxime (30 µg/disk), ceftazidime (30 µg/disk), cefoxitin (30 µg/disk), cefepime (30 µg/disk), aztreonam (30 µg/disk), meropenem (10 µg/disk), ertapenem (10 µg/disk), imipenem (10 µg/disk), gentamicin (10 µg/disk), tobramycin (10 µg/disk), amikacin (10 µg/disk), ciprofloxacin (5 µg/disk), nalidixic acid (30 µg/disk), levofloxacin (5 µg/disk), minocycline (30 µg/disk), tetracycline (30 µg/disk), trimethoprim/sulfamethoxazole (25 µg/disk), and chloramphenicol (30 µg/disk). The control strains used in the study were P. aeruginosa ATCC 27853 and E. coli ATCC 25922 (American Type Culture Collection, Manassas, USA). The multiple antibiotic resistance (MAR) index of the GNB isolates was calculated via the following formula [21]: MAR index = (the number of antibiotics to which the GNB strains are resistant)/(total number of antibiotics tested). Statistical analysis Data were entered using Microsoft Excel 2019 and analyzed with IBM SPSS Statistics for Windows, version 23.0 (IBM Corp., Armonk, NY, USA). Categorical variables were compared using the chi-square test or Fisher’s exact test when expected cell counts were below 5. A p-value < 0.05 was considered statistically significant. Univariate logistic regression was performed to examine associations between potential risk factors and MDR-GNB infection, with results expressed as odds ratios (ORs) and 95% confidence intervals (CIs). In line with standard practice, variables with p < 0.05 in univariate analysis or with strong clinical plausibility were entered into the multivariate logistic regression model. To further explore predictive potential, we developed an exploratory bedside risk index for MDR-GNB infection among HAI cases. Model coefficients were estimated using logistic regression (solver = liblinear). To mitigate overfitting, we reported the number of MDR events and the events-per-variable (EPV), and applied internal validation via 1,000 bootstrap resamples to obtain optimism-corrected discrimination and calibration metrics. The optimal thresholds for classification were identified by analyzing sensitivity, specificity, and likelihood ratios across score cut-offs. Reporting followed the transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD) guidelines and incorporated a prediction model risk of bias assessment tool (PROBAST)-based risk-of-bias appraisal [22, 23]. As recommended by contemporary guidance, we explicitly acknowledge the single-center sample size, the exploratory nature of the tool, and the requirement for external validation before clinical implementation [24]. The MAR indices across bacterial species were compared using a Kruskal–Wallis ANOVA. Results Characteristics of patients with MDR-GNB infections From November 1, 2021, a total of 346 specimens tested positive for culture, including 208 patients who received their initial culture report. Altogether, 188 patients tested positive for GNB. Table 1 shows the demographic characteristics of patients with MDR-GNB and non-MDR-GNB infections. Among the 188 GNB-positive patients, the study maintained a one-to-one case-to-control ratio: 56.4% (106 patients) had MDR-GNB infections (cases), whereas 43.6% (82 patients) had non-MDR-GNB infections (controls). The median age was 51 years for patients with MDR-GNB infections, whereas it was 53 years for those with non-MDR-GNB infections, with no significant difference in sex distribution. The median length of current hospital stay for patients with MDR-GNB infection was 10 days, and 47.2% of cases were admitted to the surgical ward. Most cases also had at least one underlying disease (59 cases, 55.7%). Invasive medical devices were commonly used among MDR-GNB cases, particularly urinary catheters (51%) and endotracheal tubes (27.4%). Moreover, the majority of patients (61.3%) had received antibiotic therapy within the preceding 90 days. These descriptive findings provided the basis for the subsequent univariate logistic regression analysis to identify factors associated with MDR-GNB infections. Table 1Univariate logistic regression analysis of risk factors associated with MDR-GNB infectionsCharacteristicsMDR-GNBn = 106 (56.4%)NON-MDR-GNBn = 82 (43.6%)Odds Ratios[confidence intervals]p-valueDEMOGRAPHIC FACTORAgeMean age51 ± 18.7453 ± 19.240.48300GenderFemale50 (47.2%)49 (59.8%)1.69 [0.74 – 3.88]0.10500Male56 (52.8%)33 (40.2%)EmploymentEmployed53 (50%)34 (41.5%)1.41 [0.79 – 2.52]0.30200Unemployed53 (50%)48 (58.5%)0.71 [0.40 – 1.27]CLINICAL FACTORAntibiotic exposurePrevious antibiotic therapy in the past 90 days65 (61.3%)18 (22.0%)5.64 [2.93 – 10.83] < 0.00001Targeted antibiotic therapy14 (13.2%)11 (13.4%)0.98 [0.42 – 2.29]0.96690Empirical antibiotic therapy35 (33%)11 (13.4%)3.18 [1.5 – 6.76]0.00260Preventive antibiotic for prophylaxis15 (14.2%)9 (11%)1.34 [0.55 – 3.23]0.51860Site of infectionUrinary system66 (62.3%)51 (62.2%)1.00 [0.55 – 1.82]1.00000Abdominal11 (10.4%)18 (22.0%)0.41 [0.18 – 0.93]0.04010Blood2 (1.9%)16 (19.5%)0.08 [0.02 – 0.36]0.00006Respiratory system58 (54.7%)30 (36.6%)2.09 [1.16 – 3.78]0.01410Underlying diseaseLiver diseases11 (10.4%)1 (1.2%)9.38 [1.19 – 74.21]0.03390Peritonitis9 (8.5%)15 (18.3%)0.41 [0.17 – 1.00]0.05063Diabetes mellitus27 (25.5%)12 (14.6%)1.99 [0.94 – 4.23]0.07323Renal disease19 (17.9%)22 (26.8%)0.60 [0.30 – 1.20]0.14470Anemia12 (11.3%)9 (11%)1.04 [0.41 – 2.59]1.00000Hypertension5 (4.7%)3 (3.7%)1.30 [0.30 – 5.62]1.00000Cancer16 (15.1%)19 (23.2%)0.59 [0.28 – 1.23]0.18743Hospitalization historyPrevious hospitalization42 (39.6%)19 (23.2%)2.18 [1.14 – 4.14]0.01900Length of current stay10.57 ± 6.706.24 ± 6.32Ward admissionIntensive care unit31 (29.2%)6 (7.3%)5.24 [2.06 – 13.28]0.00010Surgical50 (47.2%)30 (36.6%)0.65 [0.36 – 1.17]0.18100Medical22 (20.8%)22 (20.8%)0.71 [0.36 – 1.41]0.38600Oncology14 (13.2%)19 (23.2%)0.50 [0.24 – 1.08]0.08400Maternity1 (0.9%)3 (3.7%)0.25 [0.03 – 2.46]0.31949Hemodialysis2 (1.9%)3 (3.7%)0.51 [0.08 – 3.10]0.65500Medical device usage in the past 90 daysFoley catheter54 (51%)9 (11%)8.42 [3.82 – 18.56] < 0.00001Tracheal intubation + ventilation29 (27.4%)1 (1.2%)30.51 [4.06 – 229.45]0.00090Central venous catheter13 (12.3%)16 (19.5%)0.58 [0.26 – 1.28]0.17570PATHOGEN-RELATED FACTORSE. coli48 (45%)65 (79%)0.23 [0.12 – 0.44]0.00003K. pneumoniae21 (19.8%)8 (9.8%)2.29 [0.90 – 5.17]0.06320P. aeruginosa6 (5.7%)2 (2.4%)2.40 [0.47 – 12.21]0.29160A. baumannii29 (27.4%)1 (1.2%)30.51 [4.06 – 229.45]0.00090Serratia marcescens02 (2.4%)0.99930Enterobacter cloacae2 (1.9%)4 (4.9%)0.38 [0.07 – 2.10]0.40640Abbreviation: MDR-GNB, multidrug-resistant Gram-negative bacilli Univariate logistic regression analysis of risk factors associated with MDR-GNB infections Abbreviation: MDR-GNB, multidrug-resistant Gram-negative bacilli Risk factors for MDR-GNB infections Univariate logistic regression analysis Several variables were analyzed to determine their association with MDR-GNB infections. A total of 11 variables were found to be significantly associated with MDR-GNB acquisition (Table 1). Among demographic characteristics, no significant associations were observed with age, gender, or employment status. Regarding clinical characteristics, significant predictors included a history of antibiotic therapy within the previous 90 days (OR = 5.64, 95% CI: 2.93–10.83, p < 0.001), the presence of liver disease (OR = 9.38, 95% CI: 1.19–74.21, p = 0.035), a respiratory site of infection (OR = 2.09, 95% CI: 1.16–3.78, p = 0.014), prior hospitalization within the past 90 days (OR = 2.18, 95% CI: 1.14–4.14, p = 0.018), and the use of invasive medical devices, including urinary catheters (OR = 8.42, 95% CI: 3.82–18.56, p < 0.001) and endotracheal tubes or mechanical ventilation (OR = 30.51, 95% CI: 4.06–229.45, p < 0.001). Concerning pathogen-related factors, infections caused by A. baumannii were strongly associated with MDR-GNB (OR = 30.51, 95% CI: 4.06–229.45, p < 0.001), underscoring the importance of stringent infection control measures to limit the spread of this high-risk organism. Multivariate logistic regression analysis All variables with p < 0.05 in univariate analysis or with strong clinical plausibility were entered into the multivariate logistic regression model. After backward stepwise selection, five independent predictors were retained. Previous antibiotic therapy in the last 90 days (adjusted OR = 14.79; 95% CI: 4.60–47.56), previous hospitalization in the last 90 days (adjusted OR = 3.91; 95% CI: 1.52–10.07), and Foley catheterization in the last 90 days (adjusted OR = 3.69; 95% CI: 1.19–11.47) were associated with increased odds of MDR-GNB infection. In contrast, bloodstream infection (adjusted OR = 0.09; 95% CI: 0.01–0.52) and infection caused by E. coli (adjusted OR = 0.28; 95% CI: 0.10–0.75) were independently associated with reduced odds of MDR-GNB infection (Table 2). Table 2Assessment of risk factors for MDR-GNB infections using multivariate logistic regression analysesVariablesOdds Ratios95% Confidence Intervalsp-valuePrevious antibiotic therapy in the past 90 days14.79[4.6 – 47.56] < 0.0001Empirical antibiotic therapy0.30[0.08 – 1.13]0.0754Abdominal0.75[0.22 – 2.59]0.6518Blood0.09[0.01 – 0.52]0.0077Respiratory system0.00010.00010.9970Liver diseases5.88[0.35 – 98.0]0.2174Previous hospitalization in the last 90 days3.91[1.52 – 10.07]0.0040Intensive care unit admission0.24[0.02 – 2.46]0.2380Foley catheter3.69[1.19 – 11.47]0.0240Tracheal intubation + ventilation30.3[0.91 – 1012.44]0.0567E. coli0.28[0.1 – 0.75]0.0111A. baumannii1.82[0.11 – 28.71]0.6717 Assessment of risk factors for MDR-GNB infections using multivariate logistic regression analyses Application of a risk assessment tool in the prediction of MDR-GNB infections A risk assessment tool for MDR-GNB infections was developed using five pre-specified clinical predictors: previous antibiotic therapy within the past 90 days (2 points), previous hospitalization in the last 90 days (1 point), Foley catheter use (1 point), bloodstream infection (−2 points), and infection with E. coli (−1 point). The total score ranged from −3 to +4. Table 3 summarizes the scoring system and assigned weights. Table 3Risk scores for MDR-GNB infection risk factorsRisk FactorAssigned PointsPrevious antibiotic therapy in the past 90 days2Previous hospitalization in the last 90 days1Foley catheter1Bloodstream infection−2E. coli infection−1Total possible score−3 to +4 Risk scores for MDR-GNB infection risk factors The tool demonstrated excellent discrimination, with an apparent area under the curve (AUC) of 0.876 and an optimism-corrected AUC of 0.866 (bootstrap, 1,000 iterations). Calibration was acceptable (intercept −0.033, slope 1.248), indicating good agreement between predicted and observed risks. Stratification into three risk groups showed clear separation in MDR prevalence: 13.7% in the low-risk group (−3 to −1 points), 53.5% in the intermediate-risk group (0–1 points), and 92.4% in the high-risk group (2–4 points) (Fig. 2, supplementary file S1). ROC curve and calibration plot for the MDR-GNB risk assessment tool (five predictors). (a) ROC curve with an AUC of 0.88, showing good discrimination (optimal cut-off = 0.69; sensitivity = 72.6%; specificity = 90.2%). (b) calibration plot illustrating the agreement between predicted and observed probabilities of MDR-GNB infection. Abbreviations: ROC, receiver operating characteristic; AUC, area under the curve; MDR-GNB, multidrug-resistant Gram-negative bacilli Analysis of cut-off points and classification accuracy The diagnostic performance of the risk score was evaluated across different thresholds (Table 4). A cut-off of ≥ 2 points provided the best balance between sensitivity (82.1%) and specificity (68.3%), with a positive likelihood ratio (LR+) of 2.59 and a negative likelihood ratio (LR-) of 0.26. A threshold of ≥ 3 points markedly increased specificity to 95.1% and LR+ to 10.6, though sensitivity decreased to 51.9%. Higher cut-offs (≥4 points) yielded excellent specificity (98.8–100%) and strong LR+ values ( > 20), but at the expense of sensitivity ( < 30%). Table 4Analysis of breakpoints and classification accuracyScore Cut-Off (≥)Sensitivity (%)Specificity (%)Correctly Classified (%)LR+LR-186.848.868.11.690.27282.168.376.62.590.26351.995.173.410.640.51430.298.866.024.750.71513.210057.4–0.87Abbreviations: LR+, positive likelihood ratio; LR-, negative likelihood ratio Analysis of breakpoints and classification accuracy Abbreviations: LR+, positive likelihood ratio; LR-, negative likelihood ratio These findings suggest that ≥ 2 points may be appropriate for screening purposes (ruling out low-risk patients), while ≥ 3 points are better suited for confirming high-risk patients in clinical practice. Bacterial species identification and AMR profiles Among the 188 Gram-negative isolates recovered from clinical samples, E. coli was the most frequently identified species (n = 85; 45%), followed by A. baumannii (n = 52; 27.4%), K. pneumoniae (n = 37; 19.8%), P. aeruginosa (n = 11; 5.7%), and E. cloacae (n = 3; 1.9%). Two isolates of S. marcescens (1.2%) were also detected. The antimicrobial susceptibility testing results are summarized in Table 5. E. coli isolates exhibited high resistance to ampicillin (87%) and amoxicillin + clavulanic acid (71%), with moderate resistance to first- and third-generation cephalosporins (32–55%). Fluoroquinolone resistance ranged from 38% to 45%, whereas carbapenem resistance remained below 10%. K. pneumoniae showed extensive resistance to β-lactams, including amoxicillin + clavulanic acid (82%) and third-generation cephalosporins (71%), as well as to fluoroquinolones (68%). Resistance to carbapenems ranged between 11% and 18%, while intrinsic resistance to ampicillin was excluded from the analysis. A. baumannii isolates exhibited very high resistance to all β-lactams and aminoglycosides, including carbapenems (93–100%). P. aeruginosa demonstrated resistance to piperacillin (87.5%), ceftazidime (75%), cefepime (87.5%), and fluoroquinolones (87.5%), with aminoglycoside resistance varying from 25% to 62.5%, while all isolates remained susceptible to carbapenems. E. cloacae displayed moderate resistance to cephalosporins and aztreonam (33%) but was fully susceptible to carbapenems. Finally, S. marcescens isolates were fully susceptible to all tested antibiotics. Table 5Antibiotic resistance rate of GNB species isolated from clinical samplesGNB speciesResistance to antibiotic (%)AMPTICPIPAMCTZPKFFOXCTXCROCAZATMFEPETPMEMIMPGENAKTOBNALCIPLVXTETMNOSXTCK. pneumoniae (n = 29)---82717539717171717114181154764716864--7121E. coli (n = 113)87--71-5527323234323298622935534538--3610E. cloacae (n = 6)-50--50--333333333300033033331717--500P. aeruginosa (n = 8)-10087.5-87.5----755087.500062.52562.5-87.587.5----A. baumannii (n = 30)-100100-100----97-931009793000-97930012-S. marcescens (n = 2)-00-0--000-0000000-00000-Abbreviations: AMP: Ampicillin; TIC: Ticarcillin; PIP: Piperacillin; AMC: Amoxicillin + Clavulanic acid; TZP: Tazobactam + Piperacillin; KF: Cephalothin; FOX; Cefoxitin; CTX: Cefotaxime; CRO: Ceftriaxone; CAZ: Ceftazidime; ATM: Aztreonam; FEP: Cefepime; ETP: Ertapenem; MEM: Meropenem; IMP: Imipenem; GEN: Gentamicin; AK: Amikacin; TOB; Tobramycin; NAL: Nalidixic acid; CIP: Ciprofloxacin; LVX: Levofloxacin; TET: Tetracycline; MNO: Minocycline; SXT: Trimethoprim/Sulfamethoxazole; C: Chloramphenicol Antibiotic resistance rate of GNB species isolated from clinical samples Abbreviations: AMP: Ampicillin; TIC: Ticarcillin; PIP: Piperacillin; AMC: Amoxicillin + Clavulanic acid; TZP: Tazobactam + Piperacillin; KF: Cephalothin; FOX; Cefoxitin; CTX: Cefotaxime; CRO: Ceftriaxone; CAZ: Ceftazidime; ATM: Aztreonam; FEP: Cefepime; ETP: Ertapenem; MEM: Meropenem; IMP: Imipenem; GEN: Gentamicin; AK: Amikacin; TOB; Tobramycin; NAL: Nalidixic acid; CIP: Ciprofloxacin; LVX: Levofloxacin; TET: Tetracycline; MNO: Minocycline; SXT: Trimethoprim/Sulfamethoxazole; C: Chloramphenicol In total, 122 isolates (67.7%) had a MAR index above 0.3, with the highest index recorded at 1. When comparing the MAR indices across six different groups of Gram-negative bacteria, A. baumannii had the highest average MAR index at 0.95, followed by P. aeruginosa (0.87) and E. cloacae (0.83). K. pneumoniae had a MAR index of 0.74, whereas E. coli isolates had a relatively lower MAR index, with an average of 0.33. Owing to the variations in the MAR indices among the bacterial species, a Kruskal‒Wallis ANOVA was conducted, which produced an H-value of 103.607 with p < 0.001, indicating statistically significant differences between the groups. Pairwise comparisons using Dunn’s method showed significant differences in the MAR indices (p < 0.05) between E. coli and K. pneumoniae, E. coli and E. cloacae, E. coli and P. aeruginosa, E. coli and A. baumannii, and K. pneumoniae and A. baumannii (Fig. 3). Box plot of the MAR index of different groups of bacteria. Abbreviation: MAR, multiple antibiotic resistance A forest plot of the MAR index, which uses mean, minimum, and maximum values for each risk factor, revealed high MAR indices for respiratory system infections (0.99 [0.86, 1.00]) and for patients who had undergone invasive procedures such as intubation and ventilation (0.96 [0.42, 1.00]). These results highlight the significant level of resistance to multiple antibiotics among the different patient groups studied (Fig. 4). Forest plot of the MAR index of different risk factors studied. MAR, multiple antibiotic resistance Monthly evolution of MDR-GNB prevalence Monthly surveillance of MDR-GNB at the Beni Mellal Regional Hospital during our study period revealed an increase in the prevalence of MDR-GNB, particularly between April 2022 and April 2023. The number of A. baumannii and P. aeruginosa MDR strains began to increase in the second quarter of 2022. Conversely, there was a slight decrease in the prevalence of MDR-GNB between May 2023 and April 2024 (Fig. 5). Monthly surveillance of MDR-GNB prevalence in the regional hospital of Beni Mellal Discussion AMR is recognized as one of the most urgent threats to global health, with an estimated 1.27 million deaths attributable to bacterial resistance in 2019 and projections showing a continued rise if prevention and stewardship efforts are not intensified [25, 26]. Analyzing the risk factors for acquiring MDR-GNB is essential to mitigating the burden of AMR in HAIs, as these pathogens represent a significant threat to patient outcomes and complicate clinical management. Our study adds to this global evidence by documenting the epidemiology of MDR-GNB in a Moroccan regional hospital, providing novel data from a setting where systematic information remains limited. The prevalence of MDR-GNB among HAIs in our cohort was 56.4%, which is higher than that reported in a tertiary hospital in Saudi Arabia (≈42%) [27] and in a rural Thai hospital (≈31%) [9], but comparable to Moroccan ICU surveys reporting rates between 40% and 60% [13, 28]. These results highlight that the MDR-GNB burden in Moroccan hospitals is similar to that observed in middle-income countries with high antibiotic consumption, underlining the urgent need for tailored surveillance and stewardship interventions. Multivariate modeling confirmed recent antibiotic exposure as the strongest predictor of MDR-GNB infection, a finding consistent with previous studies in Europe and Latin America showing that broad-spectrum antibiotic use substantially increases the risk of resistant Gram-negative pathogens [29–31]. Prior hospitalization within the preceding 90 days was also independently associated with MDR-GNB, reflecting repeated exposure to hospital flora and selective pressure, as previously reported in Moroccan ICUs [13] and in multicenter European cohorts [32–34]. Device-related exposures, particularly urinary catheterization, were strongly associated with MDR acquisition, echoing results from the International Nosocomial Infection Control Consortium, which consistently links invasive devices to increased risk of HAIs in low- and middle-income countries [35]. Interestingly, bloodstream infection and E. coli infection were independently associated with reduced odds of MDR-GNB. This may be explained by lower antibiotic selection pressure in bloodstream infections in our cohort and by the fact that E. coli generally exhibits lower resistance rates compared with non-fermenters such as A. baumannii and P. aeruginosa. Non-fermenters are dominant in ICU settings and more prone to multidrug resistance, whereas E. coli often causes community-onset infections [26, 36]. The highest MAR index values were observed in A. baumannii, corroborating reports of endemic carbapenem-resistant A. baumannii in Moroccan and regional ICUs [37, 38]. Elevated MAR indices among respiratory infections and patients undergoing invasive procedures reflect the combined impact of antibiotic pressure and cross-transmission in critical care wards. These observations highlight the need for antimicrobial stewardship and rigorous device-care bundles to control the spread of high-risk clones. Another contribution of this study is the development of an exploratory bedside risk index for predicting MDR-GNB infections. Based on a small set of readily available clinical predictors, the tool demonstrated excellent discrimination (AUC 0.88) and good calibration. Compared with previously published scores, which often achieved very high specificity but very low sensitivity, our tool provides more balanced performance and remains simple enough for bedside application [9, 39]. Its integration into daily practice could improve the early identification of high-risk patients, optimize empirical antibiotic therapy, and support targeted infection prevention strategies. Nevertheless, we explicitly acknowledge the limitations related to single-center design and sample size and emphasize the need for external validation in larger multicenter cohorts. Taken together, our findings confirm that recent antibiotic exposure, repeated hospitalization, and invasive devices are key drivers of MDR-GNB infections, while non-fermenters, especially A. baumannii, remain the most problematic organisms in Moroccan hospitals. The introduction of a pragmatic, exploratory risk assessment tool, parallel with antimicrobial stewardship and infection control reinforcement, could represent an effective strategy to mitigate MDR-GNB burden and improve patient outcomes in resource-limited healthcare systems. Limitations of this study This study has several limitations that should be considered when interpreting the findings. First, the sample size was relatively small, which may limit the generalizability of the proposed risk model; therefore, its performance should be evaluated in larger and more diverse patient populations. In addition, the observation that E. coli appeared to act as a “protective factor” requires cautious interpretation. Although this pattern may relate to the microorganism’s generally lower resistance rates and its frequent involvement in community-onset rather than ICU-acquired infections, these explanations remain hypothetical. Future studies with broader sampling and more detailed clinical and microbiological data are needed to determine whether this association reflects a true protective effect or arises from contextual factors specific to our cohort. Conclusion This study highlights the urgent need for continuous surveillance and proactive management of MDR-GNB in healthcare settings. The observed 56.4% prevalence of MDR-GNB infections at the Beni Mellal Regional Hospital underscores the necessity of strengthening infection control measures and optimizing antibiotic stewardship. Key determinants included prior antibiotic exposure, recent hospitalization, and the use of invasive devices such as Foley catheters, all of which significantly increase infection risk. Importantly, the MAR index provided an additional dimension to resistance assessment by quantifying cumulative antibiotic selection pressure. The high MAR values observed in A. baumannii and among patients with respiratory or device-associated infections confirm the index’s utility as a complementary epidemiological indicator for identifying high-risk environments and guiding targeted interventions. Integrating MAR index monitoring into routine hospital surveillance could therefore support early detection of emerging resistance trends and inform more effective infection prevention strategies. Electronic supplementary material Below is the link to the electronic supplementary material. Supplementary Material 1 Acknowledgements The authors would like to thank the staff of Beni Mellal Regional Hospital for their assistance in data collection and the members of the microbiology laboratory for their technical support. Abbreviations Acinetobacter baumannii Antimicrobial Resistance Area Under the Curve Confidence Intervals Enterobacter cloacae Escherichia coli European Centre for Disease Prevention and Control Events Per Variable Extended-Spectrum Beta-Lactamases Gram-Negative Bacilli Healthcare-Associated Infections Klebsiella pneumonia Positive Likelihood Ratio Negative Likelihood Ratio Multiple Antibiotic Resistance Multidrug-Resistant Gram-Negative Bacilli Odds Ratios Pseudomonas aeruginosa Prediction model Risk Of Bias Assessment Tool Receiver Operating Characteristic Serratia marcescens Transparent Reporting of a multivariable prediction model for Individual Prognosis or Diagnosis Author contributions AD contributed to writing the original draft, validation, software, methodology, formal analysis, and data curation. RA contributed to resources, investigation, formal analysis, and data curation. HR contributed to methodology, investigation, and data curation. AR contributed to software, resources, methodology, and funding acquisition. KN contributed to visualisation, validation, supervision, and conceptualisation. MT contributed to visualisation, validation, and supervision. AB contributed to writing, review, and editing. All authors read and approved the final manuscript. Funding This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors. Data availability The datasets generated and/or analysed during the current study are available from the corresponding author upon reasonable request. Declarations Ethics approval and consent to participate The study protocol was approved by the Comité d’Éthique pour la Recherche Biomédicale d’Oujda, affiliated with the Faculty of Medicine and Pharmacy, Mohammed First University, Oujda, Morocco (Approval No. 22/2021). and the National Commission for the Protection of Personal Data (authorization No A-RS-255/2021). All procedures involving human participants were conducted in accordance with the ethical standards of these institutions, national regulations, and the 1964 Helsinki Declaration and its amendments or equivalent ethical standards. Written informed consent was obtained from all participants prior to sample collection and data analysis. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1 Data Availability Statement The datasets generated and/or analysed during the current study are available from the corresponding author upon reasonable request.