Epidemiological characteristics and risk factors associated with bacteremia caused by intrinsically colistin-resistant gram-negative microorganisms: a case–control study

Abstract

Objective

To identify risk factors and clinical characteristics of bacteremia caused by intrinsically colistin-resistant Gram-negative microorganisms (ICRMs) in intensive care unit (ICU) patients and to compare them with Klebsiella pneumoniae bacteremia.

Methods

This retrospective case–control study was conducted in the ICUs of a tertiary-care hospital in Türkiye between January 2023 and February 2024. Adult patients with ICRM-positive blood cultures were compared with those with colistin-susceptible or colistin-resistant Klebsiella pneumoniae bacteremia, and demographic, clinical, and outcome data were analyzed.

Results

In total, 439 patients were included: 148 with ICRM bacteremia, 191 with colistin-susceptible K. pneumoniae bacteremia, and 100 with colistin-resistant K. pneumoniae bacteremia. Exposure to colistin or polymyxin B within the previous three months was independently associated with ICRM bacteremia (OR = 2.87, 95% CI = 1.29–6.39, p = 0.010). Polymicrobial bacteremia was identified in 20% of ICRM cases (30/148), most commonly involving Enterococcus spp. (7/30). New-onset BSIs within fourteen days of the initial bacteremia were more frequent in the ICRM group (16.9%, 25/148) than in both K. pneumoniae groups (p < 0.05), with K. pneumoniae as the most common pathogen (40%, 10/25). ICRM bacteremia cases were more often community-acquired or healthcare-associated and were associated with lower Sequential Organ Failure Assessment (SOFA) scores and mortality rates than both K. pneumoniae groups (each p < 0.05).

Conclusion

ICRM bacteremia poses a growing threat in critically ill patients due to intrinsic and increasing acquired resistance. Prior colistin or polymyxin-B exposure was identified as an independent risk factor, and these infections were more often polymicrobial and community- or healthcare-associated than K. pneumoniae bacteremia. Given the limited ICU-focused evidence on ICRM bacteremia, these findings highlight prior polymyxin exposure as a potentially modifiable risk factor and support antimicrobial stewardship efforts. Therefore, empirical therapy in high-risk patients should consider potential ICRM involvement.

Graphical Abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s12879-026-13762-0.

Background

Antimicrobial resistance represents a major global health threat, as resistant infections increasingly limit available treatment options [1]. As carbapenem resistance increases in Gram-negative microorganisms, colistin has re-emerged as an alternative therapy. Colistin is active against most Gram-negative microorganisms; however, members of the Tribe Proteeae (Proteus spp., Morganella morganii, Providencia spp.), Serratia spp., and Burkholderia cepacia are intrinsically resistant to colistin. Enterobacteriaceae intrinsically resistant to colistin show reduced tigecycline susceptibility, while Tribe Proteeae exhibit decreased imipenem susceptibility [2–4]. These organisms also acquire resistance mechanisms, such as ESBL and carbapenemases, further restricting treatment options [2, 5, 6]. Bloodstream infections are a major cause of morbidity and mortality, and bacteremia caused by intrinsically colistin-resistant microorganisms (ICRMs), though less frequent, is clinically significant due to their potential to cause life-threatening infections. Increased colistin use has contributed to higher rates of colonization and infection by microorganisms intrinsically resistant to colistin [7–10]. While ICRM infections are studied, data on bacteremia remain limited. Moreover, intensive care unit (ICU)-specific evidence regarding risk factors and clinical outcomes of ICRM bacteremia is scarce, limiting risk stratification and empiric management strategies. We hypothesized that specific risk factors, including prior colistin exposure, are associated with the development of ICRM-related bacteremia. To address this, we aimed to identify these risk factors through a case–control study comparing ICRM bacteremia with colistin-susceptible and colistin-resistant Klebsiella pneumoniae bacteremia, in order to provide clinically relevant data supporting empirical antimicrobial decision-making in ICU patients. K. pneumoniae was selected as the comparator pathogen, as it represents a well-defined and clinically relevant Gram-negative microorganism in which colistin resistance is commonly encountered.

Methods

Study design

This retrospective case–control study was conducted in the ICUs of a 3,704-bed tertiary referral hospital in Türkiye between January 2023 and February 2024, including patients ≥ 18 years. The hospital serves a large urban population and functions as a tertiary referral center admitting patients from across the country and internationally, with a total of 958 ICU beds across multiple medical, surgical, and specialized units, including cardiovascular surgery, transplantation, burn, neurosurgical, cardiology, general surgery, and neurology ICUs. ICUs in our institution are categorized as medical ICUs (managed by non-surgical specialties such as internal medicine, neurology, and cardiology), surgical ICUs (managed by surgical specialties), and anesthesiology ICUs, which function as tertiary-level general ICUs managed by anesthesiologists and admit a mixed population of critically ill patients, including both medical and surgical cases. In our institution, ICUs also manage patients with prolonged critical illness and extended lengths of stay, in addition to providing acute critical care.

During the study period, the prevalence of colistin resistance among K. pneumoniae bloodstream isolates in our ICUs was approximately 32%, while the reported national rate among healthcare-associated bloodstream infections was approximately 39% [11]. In our center, colistin is primarily used for the treatment of infections caused by carbapenem-resistant Gram-negative microorganisms and may be considered as empirical therapy in selected high-risk critically ill patients, particularly in those with prior colonization or infection with multidrug-resistant organisms, prolonged ICU stay, or recent exposure to broad-spectrum antibiotics. The case group included positive blood cultures for intrinsically colistin-resistant Gram-negative microorganisms (Proteus spp., Providencia spp., Morganella morganii, Serratia spp., and Burkholderia cepacia complex). The control group comprised K. pneumoniae bacteremia and was subdivided by colistin susceptibility using the broth microdilution method. K. pneumoniae was selected as the control pathogen due to its clinical relevance and its consistent identification as a major pathogen associated with colistin resistance in antimicrobial resistance surveillance data [12, 13], as well as to enable comparison with a single, well-defined Gram-negative pathogen. This approach allowed us to distinguish factors associated with intrinsic colistin resistance from those related to colistin resistance in general, while reducing additional microbiological heterogeneity.

The study received approval from the local ethics committee (E1-23-4016).

Data collection

Patients with ICRM or K. pneumoniae identified in blood cultures were retrieved from the hospital database, with only the first bacteremia episode occurring during the ICU stay included for each patient. Collected data included demographics, admission details, underlying diseases, Charlson Comorbidity Index (CCI), and pre-bacteremia risk factors (e.g., immunosuppression, invasive devices, PEG/ostomy, total parenteral nutrition, hemodialysis, recent antibiotics, and invasive procedures). Data on invasive procedures during hospitalization (e.g., tracheostomy, intubation, mechanical ventilation, central venous catheter, drains), bacteremia type, polymicrobial infection, antimicrobial susceptibility, new pathogen growth within 14 days, therapy, and 14- and 30-day mortality were recorded. Acute Physiology and Chronic Health Evaluation II (APACHE II) scores were calculated within 24 h of ICU admission, and Sequential Organ Failure Assessment (SOFA) scores were calculated during bacteremia. For the purposes of this study, data on all bloodstream infections across intensive care units during the same period were obtained from the Infection Control Committee.

Microbiological methods

Blood cultures were processed using the automated BACT/ALERT 3D system (bioMérieux, France). Bacterial identification was performed using matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS) with the VITEK^® MS system (bioMérieux, France). Antimicrobial susceptibility testing was carried out using the VITEK^® 2 automated system (bioMérieux, France), and results were interpreted according to the European Committee on Antimicrobial Susceptibility Testing (EUCAST) criteria. Colistin susceptibility was determined using the broth microdilution method, in accordance with EUCAST recommendations [14].

Definitions

Bloodstream infections (BSIs) are defined as positive blood cultures with infection signs and classified as community-acquired, healthcare-associated, or nosocomial. Nosocomial BSIs were those detected > 48 h after admission without prior infection symptoms. Healthcare-associated BSIs were defined as those occurring at admission or within 48 h in patients with recent intravenous therapy or dialysis (within 30 days), hospitalization for ≥ 2 days (within 90 days), or residence in a nursing home or long-term care facility. Remaining cases were classified as community-acquired. Bacteremia was classified as primary or secondary according to Centers for Disease Control and Prevention (CDC) criteria [15]. Persistent bacteremia was defined as positive blood cultures for the same microorganism persisting after at least 48 h of appropriate antibiotic therapy. Recurrent bacteremia was defined as the reappearance of bacteremia after treatment completion and clinical improvement in patients who had demonstrated negative blood cultures following the initial episode [16, 17].

Polymicrobial bacteremia was defined as the isolation of more than one microorganism from blood cultures obtained during the same bacteremic episode.

Antibiotic use against Gram-negative microorganisms was categorized as broad-spectrum or other, with broad-spectrum agents including third- and fourth-generation cephalosporins, piperacillin–tazobactam, novel β-lactam/β-lactamase inhibitors, carbapenems, polymyxins, tigecycline, and fosfomycin.

Statistical analysis

Categorical variables were compared using the chi-square test or Fisher’s exact test, as appropriate, and continuous variables were analyzed using the Mann–Whitney U test. Multivariate analysis was performed using conditional logistic regression.Variables with p < 0.05 in univariable analysis or those approaching statistical significance, as well as variables selected based on clinical relevance and those considered to have potential confounding effects, were included in the multivariable model. To avoid multicollinearity, among highly correlated variables, one was selected based on clinical considerations. Some comorbidity variables were not included in the final model due to the limited number of events and the low frequency of some variables, in order to avoid model overfitting.

A two-tailed p value < 0.05 was considered statistically significant. Statistical analyses were conducted using SPSS version 24.0.

Results

A total of 439 bacteremia episodes were analyzed: 148 with ICRM bacteremia and 291 with K. pneumoniae bacteremia (191 colistin-susceptible and 100 colistin-resistant). During this period, 1,850 bacteremia episodes were recorded in ICU patients, 1,139 of which were due to Gram-negative microorganisms. Species distribution of ICRM was as follows: Proteus spp. (70, 47.3%), Serratia spp. (53, 35.8%), Burkholderia cepacia complex (14, 9.5%), M. morganii (2, 1.4%), and Providencia spp. (9, 6.1%). Across all participants, the median age was 73 years (IQR: 60–83), and 251 (57.2%) were men (Table 1 and Supplementary Table S1).

Table 1.

Characteristics of patients

Variables	ICRM bacteremia n = 148		Colistin-susceptible K. pneumoniae bacteremia n = 191		Colistin-resistant K. pneumoniae bacteremia n = 100
	n	%	n	%	n	%
Gender, male	75	50.7	111	58.1	65	65.0
Age, years, median(IQRb)	148	74 (64–83)	191	73 (61–84)	100	70 (52-81.75)
Any comorbidity	138	93.2	181	94.8	94	94.0
Hypertension	85	57.4	105	55.0	50	50.0
Diabetes	45	30.4	59	30.9	28	28.0
Chronic kidney disease	9	6.1	19	9.9	7	7.0
Heart failure	29	19.6	18	9.4	20	20.0
Coronary artery disease	29	19.6	47	24.6	23	23.0
Peripheral vascular disease	7	4.7	2	1.0	1	1.0
Chronic pulmonary disease	25	16.9	23	12.0	20	20.0
Hepatobiliary disease	20	13.5	25	13.1	11	11.0
Urological disease	25	16.9	26	13.9	25	25.0
Cerebrovascular disease	35	23.6	26	13.6	9	9.0
Hematological malignancy	1	0.7	5	2.6	1	1.0
Solid tumor	23	15.5	31	16.2	21	21.0
Use of immunosuppressive therapy	12	8.1	23	12.0	12	12.0
Charlson comorbidity index. median(IQRb)	148	5 (3.25-7)	191	5 (3–6)	100	5 (2.25-6)
Type of ICU
Medical ICU	42	28.4	63	33.0	27	27.0
Surgical ICU	21	14.2	29	15.2	9	9.0
Anesthesiology ICUa	85	57.4	99	51.8	64	64.0
Admission diagnosis
Malignancy	13	8.8	12	6.3	6	6.0
Infection	28	18.9	32	16.8	21	21.0
Cardiac pathologies	14	9.5	17	8.9	9	9.0
Neurological disease	39	26.4	38	19.9	17	17.0
Medical causes	37	25.0	51	26.7	26	26.0
Others	11	7.4	26	13.6	15	15.0
Intra-abdominal pathologies	6	4.1	15	7.9	6	6.0

^a Anesthesiology ICU represent tertiary-level general ICUs managed by anesthesiologists and typically care for more critically ill patients

^b IQR, interquartile range (25th-75th percentile)

ICU: Intensive care unit

Secondary bacteremia occurred in 52.7% of ICRM-related cases, most commonly originating from the urinary tract. Although nosocomial bacteremia was predominant across all groups, healthcare-associated and community-acquired bacteremia were more frequent in the ICRM-related group than in either K. pneumoniae groups (each p < 0.05). Polymicrobial bacteremia was observed in 30 cases (20.3%) of the ICRM group, compared with 20 cases (10.5%) in the colistin-susceptible K. pneumoniae group and eight cases (8.0%) in the colistin-resistant K. pneumoniae group. Among ICRM cases, Enterococcus spp. (seven cases) was the most common co-isolate (Supplementary Table S2). However, no polymicrobial bloodstream infections involving both ICRM and K. pneumoniae were observed during the study period. Development of bacteremia from a different microorganism within 14 days of the initial episode was more frequent in the ICRM group than in both K. pneumoniae groups (p < 0.05). K. pneumoniae was the most common cause of new bacteremia within 14 days after ICRM-related bacteremia (10/25, 40%).

Table 2 summarizes the invasive procedures performed. Compared with the K. pneumoniae groups, the ICRM group had significantly lower rates of central venous catheter use, mechanical ventilation, and drainage catheter placement (each p < 0.05).

Table 2.

Antibiotic exposure and clinical factors

Variables	ICRM bacteremia n = 148		Colistin-susceptible K. pneumoniae bacteremia n = 191		Colistin-resistant K. pneumoniae bacteremia n = 100		P-value 1a	P-value 2b
	n	%	n	%	n	%
Antibiotic exposure within past 3 months	111	75.0	172	90.1	94	94.0	< 0.001*	< 0.001*
Other antibiotics	35	23.6	44	23.0	40	40.0	0.895	0.006*
Broad-spectrum antibioticsc	106	71.6	168	88.0	91	91.0	< 0.001*	< 0.001*
Ceftriaxone	19	12.8	51	26.7	25	25.0	0.002*	0.014*
Ceftazidime	4	2.7	10	5.2	10	10.0	0.245	0.015*
Cefepime	4	2.7	10	5.2	4	4.0	0.245	0.718d
Piperacillin-tazobactam	69	46.6	127	66.5	64	64.0	< 0.001*	0.007*
Ceftazidime-avibactam	1	0.7	6	3.1	4	4.0	0.142d	0.161d
Cefoperazone-sulbactam	4	2.7	7	3.7	1	1.0	0.762d	0.651
Meropenem	70	47.3	108	56.5	74	74.0	0.091	< 0.001*
Tigecycline	22	14.9	39	20.4	21	21.0	0.187	0.211
Polymxin B/Colistin	48	32.4	59	30.9	47	47.0	0.762	0.021*
Fosfomycin	4	2.7	12	6.3	11	11.0	0.123	0.007*
≥3 broad-spectrum antibiotics	43	29.1	79	41.4	52	52.0	0.019*	< 0.001*
Duration of broad-spectrum antibiotic use, days, median (IQRf)
Meropenem	70	0 (0–12)	108	2 (0–14)	74	9 (0–17)	0.171e	< 0.001 e *
Polymxin B/Colistin	48	0 (0–7)	59	0 (0–2)	47	0 (0–10)	0.593e	0.017 e *
Tigecycline	22	0 (0–0)	39	0 (0–9)	21	0 (0–0)	0.398e	0.196e
Fosfomycin	4	0 (0–0)	12	0 (0–0)	11	0 (0–0)	0.248e	0.009 e *
Invasive procedure-related risk factors
PEG	25	16.9	30	15.7	21	21.0	0.769	0.414
Gastrostomy	3	2.0	11	5.8	3	3.0	0.087	0.688d
Tracheostomy	29	19.6	31	16.2	21	21.0	0.421	0.787
Mechanical ventilator	74	50.0	127	65.5	81	81.0	0.002*	< 0.001*
Hemodialysis	18	12.2	40	20.9	16	16.0	0.033*	0.389
Central venous catheter	84	56.8	156	81.7	81	81.0	< 0.001*	< 0.001*
TPN	15	10.1	27	14.1	7	7.0	0.267	0.394
Drainage catheter	12	8.1	33	17.3	19	19.0	0.014*	0.011*
Invasive devices	10	6.8	27	14.1	12	12.0	0.031*	0.154
Hepatobiliary surgery/invasive procedure past three months	4	2.7	13	6.8	3	3.0	0.086	> 0.999d
Urological surgery/invasive procedure past three months	3	2.0	8	4.2	3	3.0	0.360d	0.688d
Other surgery/invasive procedures past three months	34	23.0	53	27.7	36	36.0	0.318	0.025*

^a Comparison between ICRM bacteremia and colistin-susceptible K. pneumoniae bacteremia

^b Comparison between ICRM bacteremia and colistin-resistant K. pneumoniae bacteremia

^c Broad-spectrum antibiotics were defined as third- and fourth-generation cephalosporins, piperacillin–tazobactam, novel β-lactam/β-lactamase inhibitors, carbapenems, polymyxins, tigecycline, and fosfomycin

^d Fisher’s exact test was used

^e Mann–Whitney U test was used

^f IQR, interquartile range (25th-75th percentile)

* p < 0.05 was considered statistically significant

Antimicrobial resistance results are presented in Supplementary Table S3. ICRM isolates were generally more susceptible than both colistin-susceptible and colistin-resistant K. pneumoniae isolates, except to imipenem.

SOFA scores and both 14-day and 28-day mortality were significantly lower in ICRM BSIs than in K. pneumoniae BSIs (p < 0.05 for each control group).

Risk factors for ICRM-related bacteremia identified by univariate analysis are summarized in Tables 2 and 3.

Table 3.

Clinical severity and outcome parameters of patients

	ICRM bacteremia n = 148		Colistin-susceptible K. pneumoniae bacteremia n = 191		Colistin-resistant K. pneumoniae bacteremia n = 100		P-value 1a	P-value 2b
	n	%	n	%	n	%
Development of bacteremia due to a different microorganism within 14 days of the initial bacteremia	25	16.9	16	8.4	5	5.0	0.017*	0.005*
APACHE II score, median (IQRe)	148	21 (17–27)	191	22 (15–27)	100	21 (18–29)	0.996d	0.391d
SOFA score, median (IQRe)	148	5 (3–8)	191	7 (4–10)	100	8 (5–11)	< 0.001 d *	< 0.001 d*
Length of hospital stay, days, median (IQRe)	148	42 (22.25–73.75)	191	43 (21–69)	100	57.5 (27-93.25)	0.664d	0.119d
Length of hospital stay preceding the development of bacteremia, days, median (IQRe)	148	13 (5.25-14)	191	22 (6–45)	100	31 (12.25–57.75)	0.259d	0.001 d *
Recurrent bacteremia	2	1.6	9	5.6	6	8.8	0.120c	0.024 c *
Persistent bacteremia	10	8.0	23	14.4	5	7.4	0.095	0.873
Overall death	112	75.7	170	89.0	87	87.0	0.001*	0.028*
14-day mortality	52	35.4	105	55.0	49	49.0	< 0.001*	0.033*
28-day mortality	68	46.3	124	64.9	59	59.0	0.001*	0.049*

^a Comparison between ICRM bacteremia and colistin-susceptible K. pneumoniae bacteremia

^b Comparison between ICRM bacteremia and colistin-resistant K. pneumoniae bacteremia

^c Fisher’s exact test was used

^d Mann–Whitney U test was used

^e IQR, interquartile range (25th-75th percentile)

* p < 0.05 was considered statistically significant

Logistic regression showed that polymyxin B/colistin exposure in the past three months independently increased risk of ICRM-related bacteremia compared with colistin-susceptible K. pneumoniae bacteremia (OR 2.87, 95% CI 1.29–6.39, p = 0.010). Hospitalization in a surgical ICU was independently linked to ICRM-related versus colistin-resistant K. pneumoniae bacteremia (OR 4.35, 95% CI 1.06–17.81, p = 0.041). Central venous catheters were more common in the colistin-susceptible group (OR 0.40, 95% CI 0.22–0.76, p = 0.005), while drainage catheters were more frequent in the colistin-resistant group (OR 0.25, 95% CI 0.08–0.76, p = 0.015). Polymicrobial presentation was more frequent in ICRM BSIs than in both controls and remained independently associated with ICRM BSIs in multivariable analysis (vs. colistin-susceptible: OR 2.57, 95% CI 1.26–5.22, p = 0.009; vs. colistin-resistant: OR 5.66, 95% CI 1.83–17.51, p = 0.003) (Table 4).

Table 4.

Multivariate logistic regression analysis results

Variables	ICRM vs. Colistin-susceptible K. pneumoniae bacteremiab			ICRM vs. Colistin-resistant K. pneumoniae bacteremiac
	OR	%95 CI	P-value 1	OR	%95 CI	P-value 2
Gender, male	1.61	0.94–2.74	0.080	1.53	0.77–3.03	0.225
Age (years)	0.99	0.98–1.01	0.495	1.01	0.99–1.04	0.179
APACHE II score	1.03	0.99–1.07	0.154	1.03	0.98–1.08	0.321
SOFA score	0.94	0.85–1.03	0.191	0.87	0.76–0.98	0.025*
Length of hospital stay	1.00	1.00-1.01	0.933	-	-	-
Length of hospital stay preceding the development of bacteremia	1.01	1.00-1.01	0.132	-	-	-
Type of intensive care unit
Medical intensive care unit	1.00		0.470	1.00		0.123
Surgical intensive care unit	0.58	0.21–5.60	0.293	4.35	1.06–17.81	0.041*
Anesthesiology intensive care unit	1.06	0.58–1.94	0.860	1.41	0.64–3.12	0.392
Setting of infection acquisition
Community-acquired	1.27	0.34–4.76	0.721	0.79	0.13–4.99	0.802
Healthcare-associated	0.74	0.32–1.71	0.480	2.94	0.67–12.94	0.154
Nosocomial	1.00		0.676	1.00		0.341
Hospital stay within past 3 months	0.61	0.34–1.12	0.113	0.57	0.27–1.22	0.150
Invasive procedure-related risk factors
Mechanical ventilator	0.60	0.29–1.22	0.158	0.46	0.18–1.17	0.104
Hemodialysis	0.73	0.33–1.59	0.421	0.86	0.33–2.23	0.749
Central venous catheter	0.40	0.22–0.76	0.005*	0.75	0.29–1.90	0.541
Gastrostomy	0.47	0.11–2.11	0.325	1.15	0.14–9.52	0.900
PEG	0.79	0.38–1.64	0.517	0.78	0.33–1.85	0.566
Invasive device	0.43	0.18–1.04	0.061	0.47	0.14–1.57	0.218
Drainage catheter	0.46	0.19–1.12	0.088	0.25	0.08–0.76	0.015*
Other surgery/invasive procedure	1.09	0.52–2.27	0.820	0.73	0.32–1.66	0.448
Polymicrobial bacteremia	2.57	1.26–5.22	0.009*	5.66	1.83–17.51	0.003*
Antibiotic exposure within past three months	0.55	0.21–1.44	0.224	0.44	0.46–4.24	0.478
Broad-spectrum antibioticsa	-	-	-	1.32	0.14–12.14	0.807
≥ 3 broad-spectrum antibiotics	0.43	0.17–1.07	0.068	0.93	0.24–3.66	0.917
Other antibiotics	-	-	-	0.69	0.31–1.53	0.363
Ceftriaxone	0.51	0.25–1.04	0.064	0.37	0.13–1.01	0.053
Ceftazidime	-	-	-	0.31	0.07–1.41	0.128
Piperacillin-tazobactam	0.54	0.28–1.05	0.069	0.72	0.31–1.71	0.462
Meropenem	1.25	0.52-3.00	0.625	0.64	0.20–2.02	0.446
Colistin/polymyxin B	2.87	1.29–6.39	0.010*	2.39	0.86–6.64	0.094
Tigecycline	-	-	-	2.41	0.85–6.86	0.098
Fosfomycin	0.71	0.18–2.79	0.618	0.23	0.05-1.00	0.050
Duration of meropenem use	-	-	-	0.98	0.93–1.02	0.304
Duration of colistin/polymxin B use	0.99	0.94–1.04	0.638	-	-	-

Variables shown include all covariates entered into the multivariable logistic regression model

^a Broad-spectrum antibiotics were defined as third- and fourth-generation cephalosporins, piperacillin–tazobactam, novel β-lactam/β-lactamase inhibitors, carbapenems, polymyxins, tigecycline, and fosfomycin

^bHosmer–Lemeshow goodness-of-fit test: ICRM vs. Colistin-susceptible K. pneumoniae bacteremia, p = 0,252

^cHosmer–Lemeshow goodness-of-fit test: ICRM vs. Colistin-resistant K. pneumoniae bacteremia, p = 0,528

* p < 0.05 was considered statistically significant

In a subgroup analysis of ICRM cases with new BSIs occurring within 14 days, univariate analysis identified several significant variables; however, none remained significant in the multivariate model. Prior use of broad-spectrum antibiotics targeting Gram-negative bacteria within the preceding three months approached statistical significance (OR 3.17, 95% CI 0.86–11.67, p = 0.082).

Discussion

Bloodstream infections remain a major cause of morbidity and mortality among critically ill patients. Although intrinsically colistin-resistant Gram-negative microorganisms are less frequent, inherent and emerging resistance mechanisms severely limit therapeutic options. Data on the epidemiology and risk factors of ICRM-related BSIs are limited, and no studies have comprehensively assessed the overall proportion of ICRMs among Gram-negative bloodstream infections. This ICU-based case–control study quantified the burden of ICRM bacteremia among Gram-negative bloodstream infections and identified independent risk factors and distinct microbiological characteristics compared with both colistin-susceptible and colistin-resistant K. pneumoniae bacteremia, addressing a key gap in current knowledge.

Comparison of ICRM-related BSIs with colistin-susceptible and colistin-resistant K. pneumoniae BSIs revealed differences between groups. Consistent with previous studies reporting on intrinsically colistin-resistant Gram-negative microorganisms [8–10, 18], we identified recent colistin use (within the past three months) as an independent risk factor for ICRM-related bacteremia. The intrinsic resistance of these organisms to colistin suggests that antibiotic-driven selective pressure may promote their emergence by suppressing susceptible Gram-negative flora and thereby providing a selective advantage for intrinsically colistin-resistant microorganisms.

However, although Proteus spp., Providencia spp., Morganella morganii, and Serratia spp.—all included in the ICRM group—are intrinsically resistant to tigecycline, tigecycline use was not identified as an independent risk factor for ICRM-related bacteremia in univariate analysis. This suggests that not all antibiotics to which these organisms are intrinsically resistant exert the same degree of selective pressure. The lack of association with tigecycline may reflect differences in ecological impact, tissue distribution, and patterns of clinical use compared with polymyxins. Similar findings have been reported in previous studies, including analyses of intrinsically colistin-resistant organisms as a group as well as species-specific investigations such as Providencia infections [9, 19]. Further research is warranted to elucidate the mechanisms underlying this lack of association.

Polymicrobial bacteremia occurs in approximately 6–11% of all BSIs and has been associated with poorer clinical outcomes, including higher rates of septic shock and mortality [20, 21]. In our study, the significantly higher frequency of polymicrobial bacteremia observed in the ICRM group compared with both control groups represents a novel contribution to the literature, as this phenomenon has not previously been specifically investigated in ICRM-related bacteremia. Previous studies on Tribe Proteeae bacteremia similarly reported relatively high proportions of polymicrobial infections involving Gram-negative bacteria and Enterococcus spp [5, 22, 23]., paralleling our finding that within polymicrobial ICRM-related bacteremia, Gram-negative bacteria were the most frequent co-isolates, while Enterococcus spp. was the most common specific co-pathogen. Polymicrobial bacteremia is unlikely to represent a causal risk factor; rather, it may reflect the complexity of infection and has important implications for empirical antimicrobial therapy. Similarly, the observed association with surgical ICU stay should be interpreted with caution, as it likely reflects differences in patient characteristics and infection dynamics across ICU settings rather than a direct causal relationship. Surgical ICUs are typically associated with shorter lengths of stay, which may limit colonization and infection with highly resistant pathogens such as colistin-resistant K. pneumoniae.

The ICRM group experienced a higher frequency of new BSIs caused by different pathogens within fourteen days of the initial bacteremia compared with both control groups. K. pneumoniae was the most common secondary pathogen identified in the ICRM group, and this finding has not been previously examined in the context of ICRM-related bacteremia. No statistically significant risk factors were identified for the development of new BSIs within fourteen days. However, multivariate analysis indicated that prior use of broad-spectrum antibiotics targeting Gram-negative bacteria within the previous three months approached statistical significance. This finding may be attributable to the limited number of patients with new BSIs and the imbalance in group distribution. Further studies are required to clarify risk factors for subsequent infections following ICRM-related bacteremia.

In this study, most ICRM-related bacteremia cases were of nosocomial origin. This pattern likely reflects the ICU-based patient population, the high proportion of Serratia species, and the inclusion of Burkholderia cepacia complex bacteremia. However, community-acquired and healthcare-associated infections were also more frequent in the ICRM group than in the control groups, as demonstrated by univariate analysis. This finding may be attributable to members of the Tribe Proteeae being commonly isolated from patients residing in long-term care facilities and nursing homes [24], although this could not be confirmed, as nursing home residency was not specifically evaluated. Consistently, ICRM BSIs were less frequently associated with invasive procedures, including central venous catheterization, drainage procedures, and mechanical ventilation, and were characterized by lower SOFA scores, lower mortality rates, and higher antimicrobial susceptibility. Taken together, these findings suggest that, compared with K. pneumoniae BSIs, ICRM BSIs show features closer to healthcare-associated or community-acquired infections and might be associated with a milder clinical course.

In line with prior reports, ICRM isolates exhibited higher susceptibility than those in the control groups [10, 22]. Nevertheless, resistance rates among ICRM isolates at our center remained high, likely reflecting the ICU setting and the high proportion of nosocomial isolates. In these microorganisms, the increasing prevalence of acquired resistance mechanisms, including ESBL production and carbapenemase emergence, together with the limited intrinsic activity of agents such as imipenem, tigecycline, and colistin, leaves few effective therapeutic options [2, 6, 25, 26]. Early detection and the rapid implementation of effective infection control measures are therefore critical for managing these infections.

This study has several limitations. One limitation is the retrospective design. Colistin consumption was not evaluated during the study period, which prevented analysis of temporal trends or associations with ICRM incidence. Another limitation is the single-center design. However, the study was conducted in a large tertiary referral hospital with high bed capacity, admitting patients from all regions of the country and from abroad, which partially mitigates this limitation. At the same time, the single-center setting ensured methodological consistency and data reliability by reducing variability in microbiological and clinical practices. The inclusion of only ICU patients and the use of a control group consisting exclusively of K. pneumoniae bacteremia may limit the generalizability of our findings. However, focusing on a single, clinically relevant and commonly reported colistin-resistant pathogen such as K. pneumoniae may have reduced microbiological heterogeneity. In addition, the clinical setting of our study, characterized by prolonged ICU stay and a high antimicrobial resistance burden, should be considered when interpreting the findings. Prolonged hospitalization has been identified as a risk factor for colistin resistance, particularly in cohorts with a high proportion of critically ill patients [27]. Furthermore, large international cohort studies including ICU patients from multiple countries have demonstrated that resistant Gram-negative bloodstream infections represent a global problem in critically ill populations [28]. These findings suggest that our cohort reflects a high-resistance end of a broader global epidemiological spectrum rather than a unique local phenomenon.

Conclusion

To our knowledge, this study represents the first investigation in ICU patients to evaluate risk factors and mortality associated with ICRM-related bloodstream infections. Exposure to colistin or polymyxin B within the previous three months increased risk nearly threefold. Polymicrobial bloodstream infections occurred in 20% of cases, most commonly involving Enterococcus spp. New-onset BSIs caused by other pathogens within fourteen days were more frequent in the ICRM group, with K. pneumoniae identified in 40% of these cases. ICRM-related BSIs were more often community-acquired or healthcare-associated than K. pneumoniae BSIs and generally followed a milder clinical course. The use of the broth microdilution method for colistin susceptibility testing in the control group represents a methodological strength of the study. Limited therapeutic options resulting from intrinsic resistance and increasing acquired resistance remain concerning. Given the high prevalence of resistant Gram-negative bacteria, the potential for ICRM involvement should be considered when initiating empirical antimicrobial therapy, particularly in critically ill patients with recent polymyxin exposure, and should guide appropriate antibiotic selection.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

APACHE II

Acute Physiology and Chronic Health Evaluation II

BSI

Bloodstream infection

CCI

Charlson Comorbidity Index

CDC

Centers for Disease Control and Prevention

EUCAST

European Committee on Antimicrobial Susceptibility Testing

ICRM

intrinsically colistin-resistant microorganism

ICU

intensive care unit

SOFA

Sequential Organ Failure Assessment

Author contributions

C.B.K. contributed to conceptualization and methodology, conducted the investigation, curated the data, performed the formal analysis, created the visualizations, and wrote the original draft of the manuscript. A.B. contributed to conceptualization and methodology, curated the data, created the visualizations, provided resources, supervised the study, and reviewed and edited the manuscript. A.S. contributed to methodology, data curation, formal analysis, and software. H.B. contributed to conceptualization and methodology and reviewed and edited the manuscript. All authors read and approved the final manuscript.

Funding

The authors did not receive support from any organization for the submitted work.

Data availability

The datasets generated and/or analyzed during the current study are not publicly available due to institutional privacy regulations but may be provided by the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

This study was approved by the Ankara Bilkent City Hospital No. 1 Clinical Research Ethics Committee (Approval No: E1-23-4016). The requirement for informed consent was waived due to the retrospective nature of the study and the use of pseudonymized data. The study was conducted in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Declaration of Helsinki and its later amendments or comparable ethical standards.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.