Clinical characteristics, causative organisms, and antimicrobial resistance profiles of hospital-acquired invasive gram-negative bacterial infections at a tertiary-care center in Saudi Arabia: a five-year retrospective cohort study (2020–2024)

Abstract

Background

Hospital-acquired gram-negative bacterial (HA-GNB) infections are a leading cause of nosocomial morbidity and mortality, yet local epidemiological data to guide empirical therapy remain limited.

Methods

We conducted a retrospective cohort study of 377 adult patients with culture-confirmed invasive gram-negative infection acquired ≥ 48 h after admission to King Abdulaziz Medical City (KAMC), Jeddah, Saudi Arabia (2020–2024). Demographics, comorbidities, infection type, microbiology, susceptibility, complications, and outcomes were analyzed using chi-square/Fisher’s exact tests and t-tests/Mann–Whitney U tests; multivariable logistic regression was used to identify independent predictors of multidrug-resistant (MDR) infection and in-hospital mortality. Infection types were classified according to U.S. Centers for Disease Control and Prevention / National Healthcare Safety Network (CDC/NHSN) surveillance definitions.

Results

Mean age was 64.6 ± 17.7 years (59.4% male). Hypertension (61.0%), diabetes (54.6%), and cardiovascular disease (40.8%) were the commonest comorbidities. Pneumonia predominated (47.1%), followed by bloodstream infection (33.0%), surgical-site infection (15.7%), and ventilator-associated pneumonia (9.6%). Klebsiella pneumoniae (41.9%), Pseudomonas aeruginosa (34.0%), and Escherichia coli (15.4%) were the predominant pathogens. Resistance was highest for ciprofloxacin (44.4%), ceftazidime (40.2%), and cefepime (39.9%), and 34.3% for meropenem. Overall morbidity was 68.6% and mortality 39.9%. Mortality was significantly higher in patients with cardiovascular disease (48.8% vs. 34.3%, p = 0.011), hypertension (45.4% vs. 31.5%, p = 0.015), and COPD (100.0% vs. 39.2%, p = 0.025). Diabetes was strongly associated with multidrug-resistant infection (OR 3.60, 95% CI 1.44–9.04; p = 0.004) and remained an independent predictor in multivariable logistic regression adjusting for age, sex, cardiovascular disease, and pneumonia (adjusted OR 2.82, 95% CI 1.11–7.17).

Conclusions

HA-GNB infections at this Saudi tertiary center are dominated by K. pneumoniae and P. aeruginosa, carry a roughly two-in-five case fatality, and show worrying resistance to commonly used empirical agents. Within the constraints of a single-center, retrospective design, these findings support local antibiogram-informed review of empirical protocols, continued antimicrobial stewardship, and heightened infection-control attention to high-risk subgroups, particularly patients with diabetes and cardiopulmonary comorbidities; confirmation in prospective, multi-center studies is warranted.

Introduction

Hospital-acquired infections (HAIs), also termed healthcare-associated infections, are infections that are neither present nor incubating at the time of admission and that develop during a patient’s stay in a healthcare facility, usually manifesting at least 48 h after admission [1, 2]. The Centers for Disease Control and Prevention (CDC) includes in this definition complications related to indwelling devices and invasive procedures – namely surgical-site infections (SSI), central-line-associated bloodstream infections (CLABSI), catheter-associated urinary tract infections (CAUTI), and ventilator-associated pneumonia (VAP) – which together form the surveillance backbone of the National Healthcare Safety Network (NHSN) [3, 4].

Globally, HAIs remain a major patient-safety problem. Point-prevalence studies have estimated that roughly one in twenty hospitalized patients carries an active HAI on any given day, and intensive-care units bear a disproportionate share of this burden [5, 6]. The case mix is increasingly dominated by Gram-negative bacilli, which now account for more than 30% of all HAIs reported to NHSN and for nearly half of cases of VAP and urinary tract infection [4, 7]. Lower respiratory tract and bloodstream infections are also the most lethal HAI syndromes [7].

The clinical importance of gram-negative HAIs is amplified by the rapid expansion of antimicrobial resistance. Escherichia coli, Klebsiella pneumoniae, Enterobacter spp., Pseudomonas aeruginosa, and Acinetobacter baumannii are the organisms most frequently isolated, and the latter three are listed by the World Health Organization as critical-priority pathogens for which new therapeutic options are urgently needed [8, 9]. Resistance mechanisms such as extended-spectrum beta-lactamase (ESBL) and carbapenemase production now circulate widely, narrowing empirical options and lengthening hospital stays [10]. Risk factors for acquisition include advanced age, prolonged hospitalization, ICU admission, mechanical ventilation, recent invasive procedures, indwelling devices, multiple comorbidities, and prior antibiotic exposure [2, 11].

Clinical presentations of gram-negative HAI vary by site. K. pneumoniae causes pneumonia, bacteremia, urinary tract infection, pyogenic liver abscess, and wound or burn infection [10]. Both E. coli and K. pneumoniae are frequent urinary pathogens; K. pneumoniae predominates in catheter-associated urinary tract infection, while E. coli is most often associated with urolithiasis, neurogenic bladder, and obstructive uropathy [12]. VAP typically presents more than 48 h after intubation with fever, dyspnea, tachycardia, copious secretions, hypoxemia, leukocytosis, and new pulmonary infiltrates [13]. Bacteremia most often presents with fever, whereas UTIs are characterized by dysuria, fever, and gastrointestinal symptoms [14, 15].

Despite the international literature on HA-GNB, contemporary, organism-resolved data from Saudi Arabian tertiary care centers remain limited, and the local clinical phenotype, resistance patterns, and mortality risk profile have not been comprehensively described during the post-COVID-19 period. The present study was therefore designed to characterize the demographic, clinical, microbiological, and outcome features of adult patients who developed invasive gram-negative HAI at King Abdulaziz Medical City (KAMC), Jeddah, between 2020 and 2024, with the aim of informing empirical therapy, stewardship, and infection control practices in the region.

Materials and methods

Study design and setting

This was a retrospective cohort study conducted at King Abdulaziz Medical City (KAMC), Jeddah – a tertiary-care facility of the Ministry of National Guard Health Affairs (MNG-HA) in the Western Region of Saudi Arabia. A cohort design was used because patients were identified by exposure (hospital-acquired gram-negative infection) and followed forward from their index admission to a defined outcome (survival or in-hospital death), rather than assessed at a single cross-sectional time point. All hospital-acquired infections were defined and classified according to the U.S. Centers for Disease Control and Prevention / National Healthcare Safety Network (CDC/NHSN) surveillance definitions, including the standard criteria for central line-associated bloodstream infection (CLABSI), catheter-associated urinary tract infection (CAUTI), ventilator-associated pneumonia/ventilator-associated event (VAP/VAE), and surgical-site infection (SSI) and its superficial, deep, and organ/space subtypes. Pneumonia that did not meet ventilator-associated criteria was classified as hospital-acquired pneumonia per the NHSN PNEU/VAP protocol. Medical records were reviewed for the five-year period from January 2020 to December 2024. The study was reviewed and approved by the King Abdullah International Medical Research Center (KAIMRC) Institutional Review Board (project number SP-2024-02-M12B-23; College of Medicine Research Committee endorsement, 03 June 2024). Patient identifiers were not collected, and a waiver of informed consent was granted due to the retrospective design and the use of a de-identified dataset.

Participants

Adult patients (≥ 18 years) admitted during the study period who developed a culture-confirmed gram-negative bacterial infection more than 48 h after admission – satisfying the CDC/NHSN definition of healthcare-associated infection – were eligible for inclusion. Patients with infections attributable to gram-positive organisms only, those with infection diagnosed within the first 48 h of admission (community-acquired), and those with incomplete microbiological data were excluded. Eligible patients were first identified from the microbiology laboratory information system as having a gram-negative organism recovered from a clinically significant specimen (blood, urine, sputum, endotracheal aspirate, wound or tissue, or sterile body fluid) collected more than 48 h after admission. Each candidate record was then reviewed against the corresponding NHSN site-specific definition to confirm that the positive culture represented a true hospital-acquired infection rather than colonization or contamination; specifically, pneumonia required compatible clinical, radiographic, and microbiological criteria per the NHSN PNEU protocol, and bloodstream, urinary, and surgical-site infections were confirmed against the LCBI, UTI, and SSI definitions, respectively. Pneumonia meeting ventilator-associated criteria (infection arising > 48 h after endotracheal intubation) was recorded separately as VAP and is reported as a distinct category, although such cases are also counted within the broader pneumonia group where indicated. The complete process of identification, screening, exclusion, and final inclusion of the 377-patient cohort is summarized in the study flow diagram (Fig. 1).

Fig. 1.

Study flow diagram showing identification of gram-negative isolates, screening against CDC/NHSN surveillance definitions, application of exclusion criteria, probability-based selection of eligible records, and assembly of the final analytic cohort (N = 377) followed forward to in-hospital outcome

Sample size

Assuming a target source population of approximately 20,000 admissions over the study window, a 95% confidence level, a 5% margin of error, and a conservative response distribution of 50%, the minimum required sample was estimated at 377 patients using Raosoft^® (Raosoft Inc., Seattle, WA). Because the number of culture-confirmed eligible cases over the five-year period exceeded this minimum, the a priori calculation was used only to establish the smallest defensible sample; the final cohort was then assembled by random selection of records from the full pool of eligible patients identified through the laboratory information system, using a computer-generated random sequence, until the target of 377 was reached. This probability-based selection from all eligible cases was intended to minimize selection bias relative to convenience sampling. We acknowledge that not enrolling all eligible patients is a limitation and that the potential for residual selection bias is addressed in the Limitations section.

Data sources and variables

Clinical and microbiological data were extracted from the BESTCare 2.0 electronic medical record system using a standardized, pre-piloted data abstraction sheet. Variables collected included: (i) demographics (age, sex); (ii) comorbidities (hypertension, diabetes mellitus, cardiovascular disease, chronic obstructive pulmonary disease (COPD), cancer, chronic kidney disease, anemia, and others); (iii) risk factors (ICU admission, age ≥ 65 years, presence of any chronic disease); (iv) infection type and site (pneumonia, VAP, hospital-acquired pneumonia (HAP), bloodstream infection (BSI), urinary tract infection (UTI), surgical-site infection (SSI), wound infection); (v) specimen type (blood, urine, sputum, endotracheal aspirate, wound swab, tissue, sterile fluid); (vi) causative organism; (vii) antimicrobial susceptibility (interpreted per current Clinical and Laboratory Standards Institute (CLSI) breakpoints); (viii) clinical signs and symptoms; (ix) complications including sepsis, septic shock, multidrug-resistant (MDR) infection, abscess formation, and respiratory failure; and (x) outcome at end of admission (alive vs. deceased). MDR was defined according to the international expert proposal of Magiorakos et al. as acquired non-susceptibility to at least one agent in three or more antimicrobial categories [16].

Statistical analysis

Continuous variables were summarized as mean ± standard deviation (SD) or as median with interquartile range when not normally distributed; categorical variables were summarized as counts and percentages. Categorical comparisons used the Pearson chi-square test or Fisher’s exact test where expected cell counts were below five. Continuous comparisons used the independent-samples t-test or the Mann-Whitney U test, as appropriate. For binary outcomes, crude odds ratios (OR) with 95% confidence intervals (CI) were calculated. To identify independent predictors and to adjust for potential confounding, multivariable binary logistic regression models were constructed for the two principal outcomes, MDR infection and in-hospital mortality. Candidate covariates were selected a priori on clinical grounds and included age, sex, infection type, and the comorbidities significant at p < 0.10 on univariable testing; adjusted odds ratios (aOR) with 95% CIs are reported, and model discrimination was assessed using the area under the receiver-operating-characteristic curve. Antimicrobial susceptibilities were also tabulated separately for the three predominant organisms (Klebsiella pneumoniae, Pseudomonas aeruginosa, and Escherichia coli) to provide organism-specific antibiograms. Given the large number of exploratory comparisons performed, the possibility of type I error (false-positive associations) was considered when interpreting borderline findings, and results from small subgroups with wide confidence intervals were treated as hypothesis-generating rather than confirmatory. A two-sided p-value of less than 0.05 was considered statistically significant. All analyses were performed using IBM SPSS Statistics version 26 (IBM Corp., Armonk, NY).

Ethical statement

This study was approved by the King Abdullah International Medical Research Center (KAIMRC) Institutional Review Board (project number NRJ24/008/7). As this is a retrospective study using existing health records and no patient contact is required, patient consent was waived in accordance with local regulations and the Declaration of Helsinki.

Results

Demographics

A total of 377 adult patients with culture-confirmed hospital-acquired gram-negative infection were included over the five-year period 2020–2024. The mean age was 64.6 ± 17.7 years (range, 18 − 110), and 56.2% of the cohort were aged 65 years or older. Males accounted for 224 patients (59.4%) and females for 153 (40.6%). Age was significantly higher in patients who developed pneumonia (67.94 ± 17.51 vs. 61.66 ± 17.68 years, p = 0.001), urinary tract infection (73.44 ± 12.84 vs. 63.94 ± 18.02 years, p = 0.008), and ventilator-associated pneumonia (71.11 ± 12.21 vs. 63.93 ± 18.23 years, p = 0.022). In contrast, surgical-site infections occurred in significantly younger patients (58.51 ± 17.53 vs. 65.76 ± 17.71 years, p = 0.004). Sex was associated with pneumonia (p = 0.011; OR 0.585, 95% CI 0.385–0.887), although interpretation of this odds ratio depends on the chosen reference category. Baseline demographic characteristics and the principal age – infection-type associations are summarized in Table 1.

Table 1.

Baseline demographic characteristics of the study cohort (N = 377), with significant age-based associations across the major infection types

Characteristic	Value
Total patients, n	377
Age, years — mean ± SD (range)	64.6 ± 17.7 (18–110)
Age ≥ 65 years, n (%)	212 (56.23%)
Male sex, n (%)	224 (59.41%)
Female sex, n (%)	153 (40.59%)
Mean age in pneumonia vs. no pneumonia, years	67.94 ± 17.51 vs. 61.66 ± 17.68 (p = 0.001)
Mean age in UTI vs. no UTI, years	73.44 ± 12.84 vs. 63.94 ± 18.02 (p = 0.008)
Mean age in VAP vs. no VAP, years	71.11 ± 12.21 vs. 63.93 ± 18.23 (p = 0.022)
Mean age in SSI vs. no SSI, years	58.51 ± 17.53 vs. 65.76 ± 17.71 (p = 0.004)

Comorbidities and risk factors

Most patients had at least one chronic condition. Hypertension (61.01%), diabetes mellitus (54.64%), and cardiovascular disease (40.84%) were the most prevalent comorbidities, followed by cancer (22.07%). The full prevalence profile of the major comorbidities is shown in Fig. 2. Common predisposing factors beyond comorbidity included chronic disease of any type (70.55%), age ≥ 65 years (56.23%), and ICU admission during the index stay (55.43%). Diabetes mellitus showed a strong, statistically significant association with multidrug-resistant infection: 80% of patients with MDR isolates had diabetes, compared with a lower proportion in the non-MDR group (p = 0.004; OR 3.60, 95% CI 1.44–9.04). Cardiovascular disease was significantly associated with the development of both sepsis (49.7%; p = 0.010; OR 1.76, 95% CI 1.15–2.67) and septic shock (24.1%; p = 0.028; OR 1.84, 95% CI 1.09–3.10). Hypertension, by contrast, was not significantly associated with septic shock (p = 0.79) and is therefore not reported as a significant predictor.

Fig. 2.

Prevalence of major comorbidities in the study cohort (N = 377). Cardiometabolic disease dominated the comorbidity profile

Type of infection and clinical presentation

Pneumonia was the dominant syndrome, identified in 47.1% of patients, followed by bloodstream infection (33.0%), surgical-site infection (15.7%), ventilator-associated pneumonia (9.6%), and urinary tract infection (7.2%); a smaller number of wounds, central nervous system, and intra-abdominal infections were also recorded. The full distribution of infection sites is presented in Table 2.

Table 2.

Distribution of infection sites in the cohort

Type of infection	n	%
Pneumonia (any)	177	47.07
Bloodstream infection (BSI)	124	32.98
Surgical-site infection (SSI)	59	15.69
Ventilator-associated pneumonia (VAP)	36	9.57
Urinary tract infection (UTI)	27	7.18
Wound infection	23	6.12
Other (HAP, CNS, intra-abdominal, etc.)	< 5 each	—

Patients could contribute to more than one site, so totals exceed 100%

Constitutional and localizing symptoms dominated the presentation. Fever (49.9%), localized pain or swelling (29.2%), and pus or discharge (26.5%) were the three most commonly documented features, followed by hemodynamic and neurological signs, including hypotension (17.8%) and altered mental status (15.1%). The full ranking of presenting symptoms recorded at the time of infection diagnosis is shown in Fig. 3.

Fig. 3.

Most frequent presenting signs and symptoms at the time of infection diagnosis (n = 377). Only features documented in ≥ 5% of the cohort are shown

Several signs were significantly associated with downstream complications. Septic shock was independently associated with hypotension (29.7% vs. 13.5%, p = 0.002; OR 2.73, 95% CI 1.51–4.94). Tachycardia, tachypnoea, and dyspnea showed only non-significant trends toward septic shock (all p > 0.05) and are not reported as significant associations. Sepsis was strongly associated with subsequent septic shock (32.9% vs. 8.1%, p < 0.001; OR 5.53, 95% CI 3.08–9.95) and reduced urine output (8.4% vs. 3.2%, p = 0.026; OR 2.80, 95% CI 1.09–7.19). Bacteremia was associated with hypotension (60% vs. 17.3%, p = 0.041; OR 7.20, 95% CI 1.18–43.94) and hypoxia (40% vs. 1.3%, p = 0.003; OR 48.80, 95% CI 6.64–358.79); these wide confidence intervals reflect the relatively small number of bacteremia events with concurrent vital-sign documentation. Pleural effusion was associated with bradycardia (33.3% vs. 0.8%, p = 0.032; OR 61.67, 95% CI 4.33–877.87) and with tachypnoea/dyspnea (66.7% vs. 7.0%, p = 0.015; OR 26.69, 95% CI 2.34–304.20). Pneumonia was significantly associated with cough (23.7% vs. 5.0%, p < 0.001; OR 5.88, 95% CI 2.78–11.52) and pneumothorax (5.9%, p = 0.008). Pneumonia complicated by lung collapse was associated with pleural effusion (100% vs. 0.5%, p = 0.008) and thick white secretions (100% vs. 0.3%, p = 0.005).

Morbidity and mortality

Overall morbidity – defined for the purposes of this study as the development of one or more documented infection-related complications during the index admission (sepsis, septic shock, MDR infection, abscess formation, or respiratory failure) – was 68.6%. The most frequent complications were sepsis (41.2%), septic shock (18.4%), and MDR infection (8.0%). The all-cause in-hospital mortality rate was 39.9% (129 deaths among 323 patients with a documented outcome). Mortality varied substantially by comorbidity status: it was significantly higher among patients with cardiovascular disease (48.8% vs. 34.3%, p = 0.011), hypertension (45.4% vs. 31.5%, p = 0.015), and COPD (100.0% vs. 39.2%, p = 0.025), as illustrated in Fig. 4. Cancer was also more frequent among non-survivors but did not reach significance (47.6% vs. 38.1%, p = 0.197). To explore whether mortality differed by factors other than comorbidity, in-hospital death was additionally examined across causative organism, primary infection site, MDR status, and age group (Table 3). Mortality was significantly higher in patients aged ≥ 65 years (48.3% vs. 30.2%, p = 0.001) and in those with pneumonia (50.3% vs. 30.0%, p < 0.001), whereas it was significantly lower in patients with surgical-site infection (19.2% vs. 44.0%, p = 0.001) and in those with Escherichia coli infection (26.9% vs. 42.2%, p = 0.044). Mortality did not differ significantly by MDR status (42.9% vs. 39.7%, p = 0.82) or between K. pneumoniae (41.5%) and P. aeruginosa (46.5%). On multivariable logistic regression adjusting for age, sex, and the comorbidities and infection types significant on univariable testing, the independent predictors of in-hospital mortality were older age (adjusted OR 1.41 per 10-year increase, 95% CI 1.19–1.67), cardiovascular disease (adjusted OR 1.75, 95% CI 1.05–2.91), pneumonia (adjusted OR 2.29, 95% CI 1.39–3.80), and cancer (adjusted OR 3.16, 95% CI 1.60–6.21); hypertension and diabetes were no longer significant after adjustment.

Fig. 4.

In-hospital mortality stratified by selected comorbidities. Bars show mortality (%) in patients with versus without each comorbidity; the rightmost pair denotes the cohort-wide rate (39.9%) for reference

Table 3.

In-hospital mortality stratified by subgroup (n = 323 patients with a documented outcome)

Subgroup	Died/Total	Mortality	p-value
Age ≥ 65 years	84/174	48.3%	0.0010
Age < 65 years	45/149	30.2%	0.0010
Pneumonia	76/151	50.3%	0.0004
Bloodstream infection (BSI)	46/109	42.2%	0.6310
Surgical-site infection (SSI)	10/52	19.2%	0.0010
Ventilator-associated pneumonia (VAP)	15/30	50.0%	0.2466
Urinary tract infection (UTI)	8/23	34.8%	0.6642
Klebsiella pneumoniae	56/135	41.5%	0.6465
Pseudomonas aeruginosa	46/99	46.5%	0.1390
Escherichia coli	14/52	26.9%	0.0441
MDR infection	9/21	42.9%	0.8200
Non-MDR infection	120/302	39.7%	0.8200

Mortality is deaths/total within each subgroup; p-values compare each subgroup against all other patients (Fisher exact test). No intensive care unit indicator was available in the dataset, so ICU stratification could not be performed

An unexpected finding was that no MDR infections were observed among patients who developed septic shock (0% vs. 19.9% in non-MDR cases; p = 0.003). Given the small absolute number of MDR cases and the multiple potential confounders – including earlier escalation of empirical therapy, source control, and selection of sicker patients into the non-MDR comparison group – this inverse association should be interpreted with considerable caution.

Causative organisms and antimicrobial susceptibility

Klebsiella pneumoniae (41.9%), Pseudomonas aeruginosa (34.0%), and Escherichia coli (15.4%) together accounted for more than 90% of all isolates. The full distribution of identified gram-negative pathogens, including the less common Acinetobacter, Enterobacter, and Proteus species, is shown in Fig. 5. Pseudomonas aeruginosa was significantly associated with pneumonia (OR 3.95), sputum (OR 2.29), and endotracheal aspirate isolation (OR 5.34), and was negatively associated with bloodstream infection (OR 0.48). Escherichia coli was significantly associated with urinary tract infection (OR 3.69) and surgical-site infection (OR 3.04) and was negatively associated with pneumonia. Some of these associations carried wide confidence intervals, reflecting limited precision in the smaller subgroups. Candida species, when identified, were predominantly co-isolated from sputum, but their association with any specific syndrome did not reach statistical significance.

Fig. 5.

Distribution of causative gram-negative pathogens identified across the cohort. K. pneumoniae and P. aeruginosa together accounted for over 75% of all isolates

Antimicrobial susceptibility testing demonstrated substantial resistance across multiple classes of agents. As detailed in Table 4, resistance was highest to ciprofloxacin (44.4%), with high rates also observed for ceftazidime (40.2%), cefepime (39.9%), piperacillin-tazobactam (36.4%), meropenem (34.3%), and amoxicillin (31.1%). Resistance to gentamicin and trimethoprim-sulfamethoxazole was intermediate (26.3% each) and to ceftriaxone, 24.2%, while the lowest resistance rate was observed for amikacin (16.5%), which retained activity against most isolates. Because aggregate rates across all gram-negative organisms may obscure clinically important species-level differences, susceptibility was also analyzed separately for the three predominant pathogens. Organism-specific resistance profiles for Klebsiella pneumoniae, Pseudomonas aeruginosa, and Escherichia coli are presented in Table 5. Carbapenem (meropenem) resistance was high in K. pneumoniae (43.2%) and P. aeruginosa (45.7%) but low in E. coli (10.3%); amikacin retained the best activity overall, with resistance of 31.0%, 12.1%, and 3.4% in K. pneumoniae, P. aeruginosa, and E. coli, respectively. With respect to multidrug resistance, MDR was identified at the isolate level by applying the Magiorakos et al. criteria (non-susceptibility to ≥ 1 agent in ≥ 3 antimicrobial categories) to each organism’s full susceptibility panel; the modest cohort-wide MDR proportion (8.0%; 30/376) relative to several individual-agent resistance rates reflects that single-agent resistance does not by itself satisfy the multi-category MDR threshold. MDR prevalence was highest in K. pneumoniae (12.3%; 19/155), intermediate in P. aeruginosa (7.8%; 9/116), and lowest in E. coli (3.4%; 2/58).

Table 4.

Antimicrobial resistance profile across the cohort

Antimicrobial agent	Resistance rate (%)
Ciprofloxacin	44.41
Piperacillin–tazobactam	36.44
Amoxicillin–clavulanate	31.12
Ceftazidime	40.16
Cefepime	39.89
Meropenem	34.31
Ceftriaxone	24.20
Gentamicin	26.33
Trimethoprim–sulfamethoxazole	26.33
Amikacin	16.49

Denominators correspond to the number of isolates tested against each individual agent

Table 5.

Organism-specific antimicrobial resistance (antibiograms) for the three predominant gram-negative pathogens

Antimicrobial agent	Klebsiella pneumoniae (n = 155)	Pseudomonas aeruginosa (n = 116)	Escherichia coli (n = 58)
Ciprofloxacin	83/155 (53.5%)	53/116 (45.7%)	25/58 (43.1%)
Piperacillin Tazobactam	63/155 (40.6%)	50/116 (43.1%)	19/58 (32.8%)
Amoxicillin	76/155 (49.0%)		25/58 (43.1%)
Ceftazidime	72/155 (46.5%)	51/116 (44.0%)	21/58 (36.2%)
Cefepime	75/155 (48.4%)	46/116 (39.7%)	23/58 (39.7%)
Meropenem	67/155 (43.2%)	53/116 (45.7%)	6/58 (10.3%)
Ceftriaxone	62/155 (40.0%)		22/58 (37.9%)
Gentamicin	59/155 (38.1%)	28/116 (24.1%)	6/58 (10.3%)
Trimethoprim-sulfamethoxazole	54/155 (34.8%)	17/116 (14.7%)	22/58 (37.9%)
Amikacin	48/155 (31.0%)	14/116 (12.1%)	2/58 (3.4%)
Imipenem	51/155 (32.9%)	45/116 (38.8%)	5/58 (8.6%)
Ampicillin	76/155 (49.0%)	20/116 (17.2%)	32/58 (55.2%)
Tigecycline	21/155 (13.5%)	14/116 (12.1%)	1/58 (1.7%)
Cefoxitin	38/155 (24.5%)	8/116 (6.9%)	12/58 (20.7%)

Each cell shows the number of resistant / number of isolates of that organism (% resistant). Escherichia coli retained markedly greater susceptibility to the carbapenems and amikacin than Klebsiella pneumoniae and Pseudomonas aeruginosa. Cells for amoxicillin and ceftriaxone against P. aeruginosa are left blank because P. aeruginosa is intrinsically resistant to these agents; they are therefore not routinely tested or reported for this organism

Significant clinical and microbiological associations

Table 6 consolidates the principal statistically significant associations identified in the cohort, integrating findings across comorbidities, presenting features, and clinical outcomes for ease of cross-reference.

Table 6.

Principal statistically significant associations between baseline characteristics, presenting features, and clinical outcomes (chi-square or Fisher’s exact test; OR with 95% CI for binary outcomes)

Variable / association	Comparison	OR (95% CI)	p-value
Diabetes mellitus → MDR infection	80% MDR vs. non-MDR	3.60 (1.44–9.04)	0.004
Cardiovascular disease → sepsis	49.7% vs. reference	1.76 (1.15–2.68)	0.010
Cardiovascular disease → septic shock	24.1% vs. reference	1.84 (1.09–3.10)	0.028
Sepsis → septic shock	32.9% vs. 8.1%	5.53 (3.08–9.95)	< 0.001
Hypotension → septic shock	32.8% vs. 13.5%	2.73 (1.51–4.94)	< 0.001
Cough → pneumonia	80.8% vs. 13.9%	5.88 (2.78–11.52)	< 0.001
Cardiovascular disease → mortality	48.8% vs. 34.3%	—	0.011
Hypertension → mortality	45.4% vs. 31.5%	—	0.015
COPD → mortality	100% vs. 39.2%	—	0.025

Discussion

This five-year retrospective cohort study describes the clinical, microbiological, and outcome profiles of 377 adults with hospital-acquired invasive gram-negative bacterial infections at a large tertiary care center in the Western Region of Saudi Arabia. To our knowledge, it is one of the largest contemporary single-institution descriptions of HA-GNB to span the full COVID-era window (2020–2024) in this region, and it complements existing Saudi reports by jointly characterizing organism distribution, resistance phenotype, and clinical risk factors for sepsis, septic shock, and death within the same cohort.

Demographic and clinical profile

The demographic skew towards older men (mean age 64.6 years; 59.4% male) and a heavy cardiometabolic comorbidity burden mirrors the patient profile reported in other Saudi HA-GNB cohorts. Al Bshabshe et al., in a 2018–2019 ICU study from Aseer Central Hospital, found that patients aged 50–90 years accounted for 67% of K. pneumoniae-attributable deaths [17], while the national MOH outbreak analysis by Kilani et al. identified advanced age as a significant independent predictor of HAI mortality [18]. The age-related vulnerability to lower respiratory and urinary infection observed in our cohort – and the converse finding of younger age in patients with surgical-site infection – is consistent with both Saudi and international datasets and most likely reflects the combined contributions of immunosenescence, multimorbidity, and increased exposure to invasive devices in older patients [7, 11, 17].

Site distribution and causative organisms

Pneumonia (47.1%) was the dominant infection site in our cohort, followed by bloodstream infection (33.0%) and surgical-site infection (15.7%). The Saudi systematic review by Alshammari et al., which synthesized 28 studies on community- and hospital-acquired pneumonia in the Kingdom, identified Acinetobacter spp. and Pseudomonas aeruginosa as the leading HAP pathogens nationally [19] – a finding that is in close agreement with the prominence of P. aeruginosa in our respiratory specimens (OR 3.95 for pneumonia and OR 5.34 for endotracheal aspirate). Klebsiella pneumoniae, Pseudomonas aeruginosa, and Escherichia coli together accounted for more than 90% of isolates in our cohort, a near-identical top three to that reported in essentially every contemporary Saudi descriptive series [17–19].

A notable difference from the Aseer ICU cohort of Al Bshabshe et al. lies in the relative organism mix: their study reported K. pneumoniae at 39% and Acinetobacter spp. 30%, P. aeruginosa 10%, and E. coli 7% [17], whereas our cohort had a markedly higher P. aeruginosa share (~ 34%) and a substantially smaller Acinetobacter share (~ 5%). This divergence likely reflects three factors: differences in case mix (ICU-only versus whole-hospital sampling), regional microbial ecology (southern versus western Saudi Arabia), and hospital-specific selection pressures driven by local antibiotic prescribing patterns. The MOH outbreak dataset of Kilani et al., in which gram-negative bacteria accounted for 73.8% of all outbreak cases with K. pneumoniae predominant in ICU settings [18], reinforces the consistent national pattern of K. pneumoniae predominance – a finding our data reproduces.

Antimicrobial resistance in the regional context

The resistance pattern observed in our cohort is sobering, with rates exceeding 20% for ciprofloxacin, piperacillin-tazobactam, amoxicillin-clavulanate, third- and fourth-generation cephalosporins, and even the carbapenem meropenem (34.3%). However, set against the regional benchmark, our center’s resistance is intermediate rather than extreme. The 12-year longitudinal surveillance of K. pneumoniae at Aseer Central Hospital, reported by Alshehri et al., documented a striking escalation in carbapenem resistance from 12.4% in 2012 to 69.3% by 2021, with a prevalence of difficult-to-treat resistance of 50.7% by the end of the observation window [20]. The earlier Al Bshabshe ICU study from the same institution reported meropenem resistance of 61.7% and ertapenem resistance of 65.2% [17] – roughly twofold higher than our figure of 34.3% meropenem resistance. By contrast, our center’s data fall below these southern-Saudi peer institutional rates but well above the threshold at which empirical carbapenem monotherapy can be confidently recommended. The relative position of our center likely reflects a combination of more conservative carbapenem prescribing, a different patient case-mix, and possibly the impact of stewardship interventions implemented over the study period; nonetheless, the WHO classification of carbapenem-resistant Enterobacterales as critical-priority pathogens [9] applies fully to our setting and supports annual antibiogram updates and active carbapenemase surveillance in high-risk units.

Comorbidity, MDR risk, and mortality

Diabetes mellitus emerged in our cohort as a strong, statistically significant predictor of multidrug-resistant infection (OR 3.60, 95% CI 1.44–9.04). This effect size is consistent with the case-control study of MDR Pseudomonas aeruginosa in immunocompromised patients reported by Hernández-Jiménez et al., who identified diabetes mellitus as an independent risk factor for MDR-PA infection, with an OR of 4.74 (95% CI 1.63–13.79) [21]. The mechanistic link between hyperglycemia, impaired neutrophil function, microvascular compromise, and increased colonization of the gut and skin by resistant Enterobacterales has been supported across multiple international cohorts [21]. Our finding that cardiovascular disease and hypertension were independently associated with sepsis, septic shock, and mortality echoes the principal predictors of HAI mortality reported in the Saudi MOH outbreak analysis – advanced age, ICU admission, COVID-19 co-infection, and invasive procedures [18].

The all-cause in-hospital mortality of 39.9% observed in our cohort is broadly comparable to the 33.3% rate reported by Al Bshabshe et al. in the Aseer ICU [17], reinforcing that HA-GNB carries a high case-fatality across geographically distinct Saudi tertiary-care centers; the somewhat higher figure in our cohort may reflect differences in case mix and the exclusion of patients with undocumented outcomes. By contrast, the 47.5% case-fatality rate reported by Kilani et al. across MOH outbreaks nationally [18] is appreciably higher; this difference is consistent with the fact that their dataset was restricted to declared outbreaks (an enriched-severity sample) and included a substantial proportion of COVID-19 co-infections, both of which would be expected to bias mortality upward relative to a routine-care HA-GNB cohort. The strikingly elevated mortality observed in our COPD subgroup (100%, based on only four patients with COPD and a known outcome) is broadly consistent with the international literature linking chronic respiratory disease to poor HAI outcomes and warrants confirmation in larger, prospective samples.

Vital-sign and symptom associations

The vital-sign and symptom associations identified – hypotension and tachycardia with septic shock, cough with pneumonia, hypoxia with bacteremia, and reduced urine output with sepsis – are clinically intuitive but reinforce the value of structured early-warning scoring at the bedside. Several of these associations, however, carried very wide confidence intervals (notably the bacteremia-hypoxia and pleural effusion-bradycardia signals) and should be interpreted as hypothesis-generating rather than confirmatory.

The MDR-septic shock paradox

One counterintuitive finding warrants comment: no patients in the MDR subgroup developed septic shock (0% vs. 19.9% in non-MDR cases, p = 0.007). This inverse association almost certainly reflects confounding rather than a true protective effect of resistance. Possible explanations include earlier and broader empirical antibiotic escalation in patients flagged as MDR-risk, more aggressive source control, lower baseline severity in the small MDR subgroup, and the selection of sicker patients into the non-MDR comparison arm. Of note, Hernández-Jiménez et al. observed the opposite directional signal in their immunocompromised cohort, where septic shock was an independent risk factor for MDR-PA acquisition (OR 3.73, 95% CI 1.36–10.21) [21]; the contrast underscores the importance of cohort composition and adjustment for empirical-therapy timing in interpreting MDR-outcome relationships, and a larger, prospectively adjusted study will be needed to clarify the signal observed here.

Implications for practice

Taken together, our findings reproduce the core organism profile and mortality estimate reported elsewhere in Saudi Arabia [17–19], while showing substantially lower carbapenem resistance than that from peer southern Saudi institutions [17, 20] and a national outbreak series [18]. They extend prior local work by jointly quantifying the organism mix, resistance patterns, and comorbidity-driven outcomes within a well-defined cohort. The actionable signals – a high P. aeruginosa burden in respiratory specimens, resistance exceeding 20% to first-line empirical agents including meropenem, and a 3.6-fold higher MDR risk in patients with diabetes – provide a reasonable basis for local review of the institutional empirical-therapy guideline against the current antibiogram, for continued stewardship attention to carbapenem use, and for closer infection-control monitoring of cardiometabolic and chronic-respiratory subgroups. Because these signals derive from a single-center, retrospective cohort with predominantly univariable and exploratory analyses, they should be regarded as locally hypothesis-generating and confirmed in adjusted, prospective, multi-center work before being generalized or translated into firm practice mandates.

Limitations

Several limitations should be acknowledged. First, the single-center, retrospective cohort design limits the generalizability of the findings to other Saudi institutions and to other healthcare systems with differing resistance ecologies and stewardship practices. Second, reliance on the BESTCare 2.0 electronic record introduces the risk of missing or misclassified data, particularly for clinical signs and symptoms that may not have been systematically charted. Third, antimicrobial-susceptibility denominators differ across agents because not every isolate was tested against every drug; therefore, reported resistance rates reflect the tested population for each drug rather than the full cohort. Fourth, the study did not capture the timing of empirical therapy, source control, or treatment duration, all of which are important confounders of outcomes and may help explain part of the unexpected inverse association between MDR status and septic shock. Fifth, although the dataset spans 2020–2024, year-by-year stratification of resistance trends was not feasible due to subgroup sizes, and the influence of the COVID-19 pandemic on antibiotic-prescribing patterns and HAI epidemiology could not be formally disentangled. Sixth, prior antimicrobial exposure—a well-established, arguably one of the most important, risk factors for multidrug-resistant infection—was not captured in the dataset and could not be incorporated into the analysis; its omission may confound the observed predictors of MDR infection and should be explicitly weighed when interpreting the diabetes–MDR association. Seventh, because the cohort was assembled by random selection from eligible cases rather than by enrolling every eligible patient, residual selection bias cannot be fully excluded, although probability-based sampling was used to mitigate it. Eighth, a large number of statistical comparisons were performed; the resulting risk of type I error (false-positive associations) means that borderline and subgroup findings, particularly those with very wide confidence intervals, should be regarded as hypothesis-generating, and formal correction for multiple comparisons was not applied. Finally, although multivariable logistic regression was undertaken for the principal outcomes to adjust for key confounders, the models were constrained by the available covariates and event numbers, and several reported univariable odds ratios remain crude estimates that should be interpreted accordingly. Future prospective, multi-center studies with adjusted analyses and time-resolved antibiotic exposure data are warranted.

Conclusion

In this five-year retrospective cohort of 377 adults with culture-confirmed hospital-acquired gram-negative infections at a Saudi tertiary care center, pneumonia was the dominant syndrome, and Klebsiella pneumoniae, Pseudomonas aeruginosa, and Escherichia coli accounted for more than 90% of isolates. One in three patients died in the hospital, and resistance to ciprofloxacin, piperacillin-tazobactam, amoxicillin-clavulanate, and the third- and fourth-generation cephalosporins exceeded thresholds usually considered acceptable for empirical monotherapy. Diabetes was a strong predictor of multidrug resistance, and cardiovascular disease, hypertension, and COPD were significantly associated with mortality. Interpreted within the limits of a single-center, retrospective cohort, these data support local, antibiogram-informed review of empirical protocols, continued antimicrobial stewardship, and targeted infection-control attention to the cardiometabolic- and respiratory-comorbid populations that appear to carry the highest risk of poor outcome. Prospective, multicenter studies with multivariable adjustment and time-resolved antibiotic exposure data are needed to confirm these associations before broader practice recommendations can be made.

Abbreviations

aOR

Adjusted odds ratio

BSI

Bloodstream infection

CAUTI

Catheter-associated urinary tract infection

CDC

Centers for Disease Control and Prevention

CI

Confidence interval

CKD

Chronic kidney disease

CLABSI

Central line-associated bloodstream infection

CLSI

Clinical and Laboratory Standards Institute

CNS

Central nervous system

COPD

Chronic obstructive pulmonary disease

ESBL

Extended-spectrum beta-lactamase

HA-GNB

Hospital-acquired gram-negative bacterial

HAI

Healthcare-associated (hospital-acquired) infection

HAP

Hospital-acquired pneumonia

ICU

Intensive care unit

IQR

Interquartile range

IRB

Institutional Review Board

KAIMRC

King Abdullah International Medical Research Center

KAMC

King Abdulaziz Medical City

KSAU-HS

King Saud bin Abdulaziz University for Health Sciences

MDR

Multidrug-resistant

MNG-HA

Ministry of National Guard Health Affairs

NHSN

National Healthcare Safety Network

OR

Odds ratio

SD

Standard deviation

SSI

Surgical-site infection

SPSS

Statistical Package for the Social Sciences

TMP-SMX

Trimethoprim–sulfamethoxazole

UTI

Urinary tract infection

VAE

Ventilator-associated event

VAP

Ventilator-associated pneumonia

WHO

World Health Organization

Author contributions

A.A.A. conceived the study and reviewed and edited the manuscript; A.A.Als, A.O.E., A.F.B., M.T.A, A.A.Alg., and S.F.S collected clinical data and drafted the manuscript; M.A.K performed statistical analysis and validated data. All authors 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

Data supporting the findings of this study are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

This study was approved by the Institutional Review Board of the King Abdullah International Medical Research Center (KAIMRC), Ministry of National Guard Health Affairs, Jeddah, Saudi Arabia (project number NRJ24/008/7). Because the study was retrospective and used existing, de-identified electronic health records with no direct patient contact, the requirement for informed consent to participate was waived by the KAIMRC Institutional Review Board. All methods were performed in accordance with the principles of the Declaration of Helsinki and the relevant institutional and national guidelines and regulations.

Declaration of generative AI and AI-assisted technologies in the manuscript preparation process

During the preparation of this work, the authors used Claude in order to improve the language, grammar, and coherence. After using this tool/service, the authors reviewed and edited the content as needed and take full responsibility for the content of the published article.

Consent for publication

Not applicable. This manuscript does not contain any individual person’s identifiable data (including images, videos, or personal details); all data were fully de-identified.

Competing interests

The authors declare no competing interests.