Clinical and microbiological characteristics of bacteremia in pediatric intensive care unit: a retrospective analytical observational study

Abstract

Purpose

Children with bacteremia in the pediatric intensive care unit (PICU) are often in critical condition. Early recognition and treatment by clinicians are crucial to prevent the progression of bacteremia to fatal outcomes.

Methods

This single-center, retrospective analytical observational study included 9,814 pediatric patients in the PICU, ranging from 28 days to 18 years of age. Data were drawn from the pediatric intensive care (PIC) database between 2010 and 2018. Clinical characteristics, organisms isolated from blood cultures, drug resistance patterns, and factors associated with mortality were analyzed.

Results

Among the 9,814 patients, 630 (6.42%) had bacteremia, and 80 (12.70%) died during their hospital stay. Patients with bacteremia had lower levels of platelet (PLT) count, hemoglobin, potential of hydrogen (pH), arterial oxygen partial pressure (PaO₂), sodium, and albumin, and higher levels of lactic dehydrogenase (LDH), alanine transaminase (ALT), activated partial thromboplastin time (APTT), prothrombin time (PT), D-dimer, C-reactive protein (CRP), and lactate (all P < 0.05). These patients also had increased hospital and ICU stays, along with higher in-hospital mortality compared to those without bacteremia. Age, APTT, CRP, and albumin were independent factors significantly associated with bacteremia in the PICU. A total of 728 pathogenic strains were isolated, including Gram-positive bacteria (70.05%), Gram-negative bacteria (23.90%), and fungi (6.04%). The highest case fatality rate (CFR) was observed in children with fungal septicemia (27.27%), followed by Gram-negative bacteremia (22.41%) and Gram-positive bacteremia (8.88%). The CFR was also higher in cases with multiple infections (20.34%) and multidrug-resistant (MDR) infections (14.43%). Multivariable logistic regression analysis revealed that PLT, APTT, and lactate were independent factors significantly associated with mortality in patients with bacteremia.

Conclusions

Bacteremia in the PICU is associated with an increased mortality rate. Clinicians must identify and manage the risk factors associated with poor clinical prognosis early to improve the survival prospects of patients with bacteremia.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12887-025-06282-4.

Introduction

Bacteremia is a significant cause of morbidity and mortality in pediatric intensive care units (PICUs), imposing a substantial burden on healthcare systems. It is also a major contributor to overall hospital-related morbidity and mortality, with serious consequences for children’s health and well-being [1]. Bacteremia is the presence of bacteria in the bloodstream, while sepsis is a severe response to infection leading to organ dysfunction. Not all cases of bacteremia progress to sepsis, and their management differs. Several studies have demonstrated that the mortality rate among children with bacteremia ranges from 11% to 81.8% [1–3]. A study conducted in Israel reported 47.12 bacteremia events per 1,000 PICU admissions, with an elevated mortality rate among affected patients [4]. Furthermore, severe complications are common outcomes of infections associated with bacteremia [5]. These complications can lead to extended hospital stays, substantial healthcare costs, and significant mortality. A study across six PICUs showed that different pathogens, types of infections, and additional patient-related factors influenced treatment duration and disease outcomes [6]. However, despite these studies, there remains limited evidence on the clinical and laboratory predictors of bacteremia in PICU settings.

The risk factors for bacteremia in PICU patients are multifactorial, including congenital conditions, invasive procedures (e.g., central venous catheters), immunosuppressive treatments (e.g., in oncology patients), and prolonged hospitalization [7–9]. Central line-associated bloodstream infections (CLABSIs) are particularly significant and potentially lethal complications, especially in children undergoing hematopoietic stem cell transplantation [10, 11]. A recent study indicated that pediatric patients with fragile immune systems, such as those undergoing surgery for complex congenital heart disease, are at increased risk for central venous catheter-associated bacteremia, which can result in substantial mortality [11]. Additionally, the immature innate and adaptive immune systems in children contribute to an increased risk, further exacerbating the severity and duration of infections in this population [12, 13].

The most frequently identified pathogens in PICU patients with bacteremia are Gram-negative bacilli and Gram-positive cocci. One study found that 75% of bacteremia cases were attributed to Gram-negative bacilli, while 25% were caused by Gram-positive cocci [4]. Recently, antimicrobial resistance has emerged as a critical area of research. The treatment of antimicrobial-resistant infections is particularly challenging in the PICU. A study conducted in an intensive care unit (ICU) setting found that multidrug-resistant (MDR) pathogens were prevalent in pediatric patients with bacteremia, complicating treatment [14]. Certain pathogenic bacteria, including Staphylococcus, Pseudomonas aeruginosa, and Escherichia coli, are the primary causative agents of sepsis, and the multidrug resistance of these pathogens is associated with poor prognosis [14–17].

The evidence presented highlights the complexity and severity of bacteremia in pediatric ICU patients. Further research and improvement strategies are crucial to enhancing prognosis and reducing the incidence of bacteremia in this vulnerable population. Our study was designed to describe the incidence of bacteremia, pathogen species, drug resistance, mortality in pediatric patients with bacteremia, and risk factors predicting mortality in the PICU, providing valuable insights to help pediatricians facilitate prompt diagnosis and treatment.

Methods

Study design

This is a single-center, retrospective analytical observational study included 9,814 pediatric patients in the PICU from the Children’s Hospital of Zhejiang University School of Medicine (Hangzhou, China) between 2010 and 2018, ranging from 28 days to 18 years of age. This study employed a retrospective design based on the following considerations: (1) Feasibility: the pediatric intensive care electronic medical record system contains a complete record of clinical data of children in PICU, allowing efficient data extraction; (2) Exploratory needs: rapid identification of risk factors is needed to guide clinical alerts; (3) Ethical adaptation: to avoid additional interventions in critically ill children.

Study participants

Clinical data were gathered from the pediatric intensive care (PIC) database (version 1.2.0), which includes hospital clinical records from the Children’s Hospital of Zhejiang University School of Medicine (Hangzhou, China) between 2010 and 2018 [18]. As a tertiary academic children’s hospital, the Children’s Hospital of Zhejiang University School of Medicine is the largest pediatric medical care in Zhejiang Province. The PICU is a 24-bed unit with approximately 1,100 admissions annually, serving as a regional referral center for critically ill children. To ensure data fidelity, we have retained the original Chinese data and manually reviewed most of the translated English terms. However, the ICD-10 English terms were considered to be of a sufficiently high quality to be exempt from review. With the exception of the symptoms that were extracted from narrative documents by Natural Language Processing technology, all data had undergone minimal post-processing. All eligible patients over the study period were included in our study. A comprehensive medical history was obtained for each patient, along with standard laboratory values and vital signs. Individuals aged ≤ 28 days or without laboratory data were excluded. The final cohort comprised 9,814 patients, of whom 9,184 exhibited no evidence of bacteremia, while 630 demonstrated bacteremia (Fig. 1). This project was approved by both the Institutional Review Board of the Children’s Hospital of Zhejiang University School of Medicine and the Institutional Review Board of Qilu Hospital of Shandong University (KYLL-202202-027−1). Since the study findings do not impact clinical practice, individual consent was not required. All confidential health data were anonymized.

Fig. 1.

Flow chart of study participants. ICU, intensive care unit

Data collection

The research data are all derived from the PIC electronic database by automated recording. The following clinical data were collected for this study: demographic characteristics (age and sex), vital signs [temperature, heart rate (HR), respiratory rate (RR), systolic blood pressure (SBP), and diastolic blood pressure (DBP)], laboratory data, and outcomes. Infection-related variables included pathogen species identified from positive blood cultures and antimicrobial resistance testing. All laboratory data presented were obtained from initial blood tests drawn within 12 h of admission to the PICU. Blood cultures were obtained from all patients with confirmed or suspected infection within 12 h of PICU admission and before the initiation of any antibiotics, by trained healthcare professionals via sterile venipuncture. Diagnosis was made based on criteria outlined in the 10th edition of the International Statistical Classification of Diseases and Related Health Problems (ICD-10). Specifically, patients were categorized into major diagnostic groups based on their primary admission diagnosis and clinical records, including congenital disorders, hematological disorders, circulatory disorders, neurological disorders, digestive disorders, neoplasm, respiratory disorders, trauma, and other conditions. The primary outcome was mortality during hospitalization. The implementation of disparate clinical information systems at varying temporal points resulted in the incomplete data from certain systems prior to their implementation. Consequently, the completeness of the various data tables varies and missing data does exist [18].

Missing data

The completeness of our dataset was assessed and the missing data were minimal. All key outcome and exposure variables (including bacteremia status and in-hospital mortality) were fully recorded for 100% of the cohort. Laboratory variables also had very low rates of missingness. No variable in our multivariable analysis showed more than 5% missing values. Given this low level of missing data, we did not perform imputation for missing values.

Definitions and classification of infections

Bacteremia was defined as the presence of a positive pathogen, excluding contamination, in blood cultures. Coagulase-negative staphylococci (CoNS) and other commensal bacteria were classified as true pathogens under certain criteria. Specifically, the presence of at least two positive cultures or the manifestation of clinical signs indicative of sepsis, in conjunction with targeted antibiotic treatment, was requisite for the diagnosis. New infections were defined as pathogenic microorganisms initially isolated from the patients. The same episode of infection was classified as the same organism being isolated from different blood samples within a seven-day period. Persistent infection was defined as the same organism being isolated again between the seventh and thirtieth days following the initial isolation and within the same hospital admission. Bacterial isolates were classified as MDR according to the international consensus definition by Magiorakos et al. [19], i.e., non-susceptibility to at least one agent in three or more antimicrobial categories. In borderline cases where isolates showed resistance to fewer than three classes, they were not classified as MDR. Intermediate susceptibility was counted as non-susceptibility when determining MDR status. This standardized definition was applied across all isolates to ensure consistency and comparability. Multiple infections were defined as concurrent infections by ≥ 2 distinct pathogens.

Strain identification and drug sensitivity test

Enrichment cultures were processed using a fully automated blood culture system (BACTEC FX400, BD) and accompanying blood culture bottles. Bacterial and fungal identification was performed using a fully automated microbiological identification and drug sensitivity analysis system (Vitek 2 compact, bioMerieux, France) or matrix-assisted laser desorption ionization-time of flight mass spectrometry (MALDI-TOF-MS, Bruker). Bacterial susceptibility testing was performed using a fully automated microbiological identification and susceptibility analysis system (Vitek 2 compact, bioMerieux, France) and the susceptibility testing of Candida was performed using a yeast-like fungal susceptibility kit (FUNGUS 3, bioMerieux, France) and the results were interpreted with reference to the Clinical and Laboratory Standards Institute (CLSI) for the current year. We conducted a comprehensive assessment of the identified suspected contaminants, especially coagulase-negative staphylococci and streptococci, in order to exclude the possibility of false positives due to contamination and so on.

Statistical analysis

All statistical analyses were conducted using R version 3.4.3. Continuous variables with a normal distribution were expressed as the mean ± standard deviation (SD) and analyzed using the unpaired Student’s t-test. Continuous variables with a non-normal distribution were expressed as the median (interquartile range, IQR) and analyzed using the Mann-Whitney U test. Categorical variables were analyzed using chi-square or Fisher’s exact test and presented as frequencies (proportions). The mortality among patients with Gram-positive bacteria, Gram-negative bacteria, and fungal septicemia were compared using chi-square or Fisher’s exact test. To identify factors associated with poor outcomes, univariate and multivariable logistic regression models were used. For multivariable model building, stepwise logistic regression was used. We began with all candidate variables that showed an association with the outcome in univariate analysis (P < 0.05). Using a backward elimination process, variables were removed one at a time from the full model if they did not maintain significance (P < 0.05) and if their removal did not substantially change the coefficients of remaining variables, which resulting in a parsimonious final model containing only independent predictors of the outcome. Multivariable logistic regressions were adjusted for baseline confounders, including sex, underlying diagnostic category, and vital signs. Age was treated as an independent predictor in the bacteremia model and as an adjusting covariate in the mortality model. Adjusted odds ratios (ORs) with 95% confidence intervals (CIs) were reported from the final models. The potential interactions between key variables were evaluated during model building, but none reached statistical significance. To assess the robustness of our findings, we repeated the multivariable logistic regression after excluding extreme outliers for the main predictors and by restricting the model to core clinically relevant variables only. A two-sided P-value of < 0.05 was considered statistically significant.

Results

Clinical and laboratory characteristics of the study population

The final cohort consisted of 9,814 patients, of whom 9,184 had no evidence of bacteremia, while 630 had bacteremia. Among the 630 patients with bacteremia, 550 survived and were discharged, while 80 died (Fig. 1). Table 1 provides a visual representation of the clinical and laboratory characteristics of patients with and without bacteremia. The average age of patients with bacteremia was 9.16 (IQR 2.68–38.45) months, while that of patients without bacteremia was 18.23 (IQR 5.23–58.59) months (P < 0.001), indicating a statistically significant difference. Bacteremia incidence was lower in females than in males (39.52% vs. 60.48%). Among patients with bacteremia, the most prevalent underlying disease was circulatory (19.37%), followed by respiratory (13.97%), neurologic (10.95%), and digestive (10.48%) diseases. A comparison of vital signs revealed significant differences in temperature, HR, RR, SBP, and DBP between patients with and without bacteremia (all P < 0.05). Laboratory data showed lower levels of platelet count (PLT), hemoglobin, potential of hydrogen (pH), arterial oxygen partial pressure (PaO₂), sodium, and albumin in patients with bacteremia (all P < 0.05). Additionally, patients with bacteremia exhibited significantly elevated levels of lactate dehydrogenase (LDH, 338.00 vs. 304.00 U/L), alanine transaminase (ALT, 27.00 vs. 20.00 U/L), activated partial thromboplastin time (APTT, 34.80 vs. 31.10 s), prothrombin time (PT, 12.50 vs. 11.90 s), D-dimer (1.05 vs. 0.44 mg/L), C-reactive protein (CRP, 8.94 vs. 6.00 mg/L), and lactate (1.90 vs. 1.60 mmol/L) compared to patients without bacteremia (all P < 0.05). The bacteremia group also had longer hospital stays [18.85 (10.00, 34.99) vs. 11.57 (6.99, 17.90), P < 0.001], longer ICU stays [5.88 (1.65, 20.76) vs. 1.81 (0.87, 5.35), P < 0.001], and higher in-hospital mortality (12.70% vs. 4.99%, P < 0.001).

Table 1.

Comparison of clinical characteristics between patients with and without bacteremia

	Non-bacteremia (n = 9184)	Bacteremia (n = 630)	P value
Age (month)	18.23 (5.23, 58.59)	9.16 (2.68, 38.45)	< 0.001
Gender			0.027
Male (%)	5138 (55.95)	381 (60.48)
Female (%)	4046 (44.05)	249 (39.52)
Primary diagnosis on ICU admission, n (%)			< 0.001
Congenital	1379 (15.02)	64 (10.16)
Hematological	324 (3.53)	54 (8.57)
Circulation	2244 (24.43)	122 (19.37)
Neurologic	958 (10.43)	69 (10.95)
Digestive	799 (8.70)	66 (10.48)
Neoplasm	847 (9.22)	39 (6.19)
Respiratory	1088 (11.85)	88 (13.97)
Trauma	479 (5.22)	32 (5.08)
Others	1066 (11.61)	96 (15.24)
Vital signs
Temperature (oC)	36.83 ± 0.68	36.91 ± 0.93	0.025
HR (beats/min)	124.28 ± 28.25	133.68 ± 24.63	< 0.001
RR (breaths/min)	28.00 (24.00, 36.00)	34.00 (28.00, 46.00)	< 0.001
SBP (mmHg)	99.73 ± 18.39	95.26 ± 17.04	< 0.001
DBP (mmHg)	58.00 ± 14.08	55.36 ± 14.39	< 0.001
WBC (×109/L)	9.02 (6.89, 12.04)	9.17 (6.17, 13.19)	0.566
PLT (×109/L)	321.00 (241.00, 405.00)	282.00 (165.00, 394.00)	< 0.001
Hemoglobin (g/L)	116.00 (102.00, 126.00)	107.00 (92.00, 120.00)	< 0.001
LDH (U/L)	304.00 (252.00, 410.50)	338.00 (268.00, 511.00)	0.001
pH	7.40 (7.35, 7.44)	7.39 (7.33, 7.44)	0.020
PaCO2 (mmHg)	36.20 (32.10, 41.20)	36.30 (31.80, 43.00)	0.130
PaO2 (mmHg)	131.00 (66.43, 182.00)	104.50 (55.00, 163.75)	< 0.001
ALT (U/L)	20.00 (13.00, 34.00)	27.00 (15.00, 57.00)	< 0.001
CK-MB (U/L)	31.00 (22.00, 44.00)	30.00 (20.00, 50.75)	0.629
Creatinine (umol/L)	43.00 (37.00, 51.00)	42.75 (35.00, 52.40)	0.295
Sodium (mmol/L)	136.54 ± 5.13	135.91 ± 6.29	0.003
APTT (s)	31.10 (27.60, 36.50)	34.80 (28.80, 44.60)	< 0.001
PT (s)	11.90 (11.20, 12.90)	12.50 (11.30, 14.60)	< 0.001
D-dimer (mg/L)	0.44 (0.20, 1.41)	1.05 (0.35, 2.90)	< 0.001
Fib (g/L)	2.03 (1.62, 2.56)	1.99 (1.44, 2.62)	0.961
CRP (mg/L)	6.00 (3.00, 25.66)	8.94 (4.00, 41.00)	< 0.001
Lactate (mmol/L)	1.60 (1.10, 2.50)	1.90 (1.20, 3.10)	< 0.001
Albumin (g/L)	42.00 (37.50, 45.30)	38.30 (32.60, 42.80)	< 0.001
Vasopressor use during ICU stay time, n (%)	3250 (35.39)	232 (36.83)	0.700
Hospital days	11.57 (6.99, 17.90)	18.85 (10.00, 34.99)	< 0.001
ICU days	1.81 (0.87, 5.35)	5.88 (1.65, 20.76)	< 0.001
In-hospital mortality	458 (4.99)	80 (12.70)	< 0.001

ICU Intensive care unit, HR Heart rate, RR Respiratory rate, SBP Systolic pressure, DBP Diastolic pressure, WBC White blood cell, PLT Platelet, LDH Lactic dehydrogenase, pH Potential of hydrogen, PaCO₂ Arterial partial pressure of carbon dioxide, PaO₂, Arterial oxygen partial pressure, ALT Alanine transaminase, CK-MB Creatine kinase-MB, APTT Activated partial thromboplastin time, PT Prothrombin time, Fib Fibrinogen, CRP C reactive protein

Risk factors for incidence of bacteremia in PICU

Univariate analysis showed that age, sex, HR, RR, SBP, DBP, PLT, hemoglobin, LDH, PaO₂, sodium, APTT, PT, D-dimer, CRP, lactate, and albumin were associated with bacteremia in the PICU. After adjusting for confounding factors, multivariable logistic regression analysis identified age (OR 0.99, 95% CI 0.99–1.00.99.00; P < 0.001), APTT (OR 1.01, 95% CI 1.00–1.02.00.02; P = 0.038), CRP (OR 1.00, 95% CI 1.00–1.01.00.01; P = 0.029), and albumin (OR 0.93, 95% CI 0.91–0.95; P < 0.001) as independent factors significantly associated with bacteremia in the PICU (Table 2). Among them, APTT and CRP showed a positive association with bacteremia, while age and albumin showed a negative association with bacteremia. The performance of these risk factors in predicting bacteremia in the PICU is shown in Supplementary Table 1. CRP values in different subgroups of bacteremia are showed in Supplementary Table 2, exhibiting a significantly elevated level of CRP in patients with Gram-negative bacteria compared to patients with Gram-positive bacteria (P = 0.013).

Table 2.

Logistic regression analysis for associated factors of bacteremia

	Univariate analysis		Multivariable analysis
	OR (95% CI)	P value	OR (95% CI)	P value
Age (month)	1.00 (0.99, 1.00)	< 0.001	0.99 (0.99, 1.00)	< 0.001
Male (vs. female)	1.20 (1.02, 1.42)	0.027	1.11 (0.87, 1.41)	0.397
Temperature (oC)	1.00 (0.98, 1.02)	0.998
HR (beats/min)	1.01 (1.01, 1.02)	< 0.001	1.00 (1.00, 1.00)	0.336
RR (breaths/min)	1.01 (1.00, 1.01)	< 0.001	1.00 (1.00, 1.01)	0.118
SBP (mmHg)	0.98 (0.98, 0.99)	< 0.001	0.99 (0.98, 1.01)	0.375
DBP (mmHg)	0.99 (0.98, 0.99)	< 0.001	1.01 (0.99, 1.02)	0.337
PLT (×109/L)	1.00 (1.00, 1.00)	< 0.001	1.00 (1.00, 1.00)	0.113
Hemoglobin (g/L)	0.98 (0.98, 0.99)	< 0.001	1.00 (0.99, 1.01)	0.647
LDH (U/L)	1.00 (1.00, 1.00)	0.003	1.00 (1.00, 1.00)	0.269
pH	1.00 (1.00, 1.00)	0.964
PaO2 (mmHg)	1.00 (1.00, 1.00)	< 0.001	1.00 (1.00, 1.00)	0.402
ALT (U/L)	1.00 (1.00, 1.00)	0.333
Sodium (mmol/L)	0.98 (0.96, 0.99)	0.003	1.00 (0.97, 1.02)	0.844
APTT (s)	1.02 (1.01, 1.02)	< 0.001	1.01 (1.00, 1.02)	0.038
PT (s)	1.02 (1.01, 1.03)	< 0.001	0.97 (0.94, 1.01)	0.115
D-dimer (mg/L)	1.05 (1.03, 1.06)	< 0.001	1.01 (0.98, 1.03)	0.676
CRP (mg/L)	1.01 (1.00, 1.01)	< 0.001	1.00 (1.00, 1.01)	0.029
Lactate (mmol/L)	1.08 (1.05, 1.10)	< 0.001	1.06 (0.99, 1.13)	0.088
Albumin (g/L)	0.93 (0.92, 0.94)	< 0.001	0.93 (0.91, 0.95)	< 0.001

OR Odds ratio, CI Confidence interval, HR Heart rate, RR Respiratory rate, SBP Systolic pressure, DBP Diastolic pressure, PLT Platelet, LDH Lactic dehydrogenase, pH Potential of hydrogen, PaO₂ Arterial oxygen partial pressure, ALT Alanine transaminase, APTT Activated partial thromboplastin time, PT Prothrombin time, CRP C reactive protein

Etiology and multidrug resistance

The organisms isolated from blood cultures are described in Table 3; Fig. 2. A total of 728 bacterial or fungal pathogens were isolated from blood cultures of 630 patients. Among these, 510 were Gram-positive bacteria (484 patients), 174 were Gram-negative bacteria (174 patients), and 44 were fungi (44 patients). The most frequently isolated Gram-positive bacteria were Staphylococcus species (55.91%) and Streptococcus species (5.35%). The most common Gram-negative bacteria isolated were Klebsiella species (4.67%), Escherichia coli (4.53%), Acinetobacter baumannii (2.88%), and Pseudomonas aeruginosa (2.61%). Candida was the most frequently identified fungal organism (5.91%), followed by Aspergillus species (0.14%). A total of 728 pathogens were identified, 100 (13.74%) of which were MDR infections in 97 patients. Of the 510 Gram-positive pathogens, 53 (10.39%) were MDR, with Streptococcus species being the most common (20.51%). Among patients with MDR Staphylococcus species, 4 cases resulted in death, whereas no deaths occurred in patients with MDR Streptococcus species. Of the 174 Gram-negative pathogens, 47 (27.01%) were MDR, with Acinetobacter baumannii being the most common (52.38%), and 2 deaths occurred in children. There were 24 patients with persistent infections and 59 with multiple infections (Table 5).

Table 3.

Organisms isolated from blood culture

	No. of infections (%)	No. of patients	No. of deaths (%)	No. of multidrug-resistant infections (%)	No. of deaths in children with multidrug-resistant infections
Gram-positive bacteria
Staphylococcus species	407 (55.91)	385	37 (9.61)	35 (8.60)	4
Streptococcus species	39 (5.35)	39	2 (5.13)	8 (20.51)	0
Others	64 (8.79)	60	4 (6.67)	10 (15.63)	0
Gram-negative bacteria
Escherichia coli	33 (4.53)	33	6 (18.18)	11 (33.33)	3
Pseudomonas aeruginosa	19 (2.61)	19	5 (26.32)	6 (31.58)	1
Klebsiella species	34 (4.67)	34	7 (20.59)	10 (29.41)	2
Stenotrophomonas maltophilia	7 (0.96)	7	3 (42.86)	2 (28.57)	1
Acinetobacter baumannii	21 (2.88)	21	4 (19.05)	11 (52.38)	2
Others	60 (8.24)	60	14 (23.33)	7 (11.67)	1
Fungus
Candida	43 (5.91)	43	12 (27.91)	0 (0)	0
Aspergillus	1 (0.14)	1	0 (0)	0 (0)	0

Fig. 2.

Distribution of the isolated pathogens in blood culture samples (%)

Table 5.

Case fatality rates for population subgroups

	No. of patients	No. of deaths (%)
Overall	9814	538 (5.48)
Bacteria	630	80 (12.70)
Gram-positive bacteria	484	43 (8.88)
Gram-negative bacteria	174	39 (22.41)
Fungal septicemia	44	12 (27.27)
New infection	630	80 (12.70)
Persistent infection	24	6 (25)
Multiple infections	59	12 (20.34)
Multidrug-resistant infections	97	14 (14.43)

Comparison of clinical characteristics between survivors and non-survivors

In the non-survivor group, 70% were male, with an average age of 14.70 (IQR 3.26–56.75) months (Table 4). Hematological diseases (21.25%) were the most common among the non-survivor cohort, while circulation diseases (20.73%) were more prevalent in the survivor group. No significant differences were observed in vital signs between survivors and non-survivors, including temperature, HR, RR, SBP, and DBP. The PLT (163.00 vs. 293.50 × 10⁹/L, P < 0.001), pH (7.37 vs. 7.39, P < 0.001), and albumin (33.10 vs. 38.70 g/L, P < 0.001) levels were significantly lower in the non-survivor group than in the survivor group. In contrast, LDH (468.00 vs. 334.00 U/L, P = 0.004), ALT (35.50 vs. 26.00 U/L, P = 0.017), APTT (39.25 vs. 34.40 s, P < 0.001), PT (13.10 vs. 12.30 s, P = 0.002), D-dimer (2.49 vs. 0.94 mg/L, P = 0.007), CRP (22.50 vs. 8.00 mg/L, P = 0.028), and lactate (2.60 vs. 1.80 mmol/L, P < 0.001) were significantly higher in the non-survivor group. The non-survivor group also exhibited a higher rate of vasopressor utilization (55.00% vs. 34.18%, P < 0.001) compared to the survivor group. Additionally, patients who died had significantly shorter hospital stays (10.23 days, IQR 1.60–31.72) than survivors (19.05 days, IQR 10.95–35.69; P < 0.001).

Table 4.

Comparison of clinical characteristics between survivors and non-survivors of children with bacteremia

	Survivors (n = 550)	Non-survivors (n = 80)	P value
Age (month)	8.87 (2.63, 37.25)	14.70 (3.26, 56.75)	0.052
Gender			0.062
Male (%)	325 (59.09)	56 (70.00)
Female (%)	225 (40.91)	24 (30.00)
Primary diagnosis on ICU admission, n (%)			< 0.001
Congenital	58 (10.55)	6 (7.50)
Hematological	37 (6.73)	17 (21.25)
Circulation	114 (20.73)	8 (10.00)
Neurologic	60 (10.91)	9 (11.25)
Digestive	63 (11.45)	3 (3.75)
Neoplasm	36 (6.55)	3 (3.75)
Respiratory	76 (13.82)	12 (15.00)
Trauma	25 (4.55)	7 (8.75)
Others	81 (14.73)	15 (18.75)
Vital signs
Temperature (oC)	36.94 ± 0.86	36.57 ± 1.49	0.123
HR (beats/min)	133.58 ± 24.32	134.93 ± 29.05	0.777
RR (breaths/min)	34.00 (28.00, 47.00)	36.00 (29.50, 43.00)	0.710
SBP (mmHg)	95.00 ± 16.44	98.52 ± 23.40	0.286
DBP (mmHg)	55.07 ± 14.25	58.97 ± 15.90	0.161
WBC (×109/L)	9.18 (6.29, 12.92)	8.96 (2.92, 15.79)	0.836
PLT (×109/L)	293.50 (185.25, 401.00)	163.00 (75.00, 324.00)	< 0.001
Hemoglobin (g/L)	108.00 (93.00, 120.75)	107.00 (83.00, 116.50)	0.081
LDH (U/L)	334.00 (267.25, 481.75)	468.00 (298.50, 932.50)	0.004
pH	7.39 (7.34, 7.44)	7.37 (7.30, 7.44)	< 0.001
PaCO2 (mmHg)	36.45 (31.80, 43.00)	35.10 (30.27, 42.52)	0.336
PaO2 (mmHg)	107.00 (57.05, 166.00)	89.55 (44.82, 151.25)	0.104
ALT (U/L)	26.00 (15.00, 52.00)	35.50 (16.25, 106.75)	0.017
CK-MB (U/L)	31.00 (22.00, 50.50)	24.00 (13.50, 51.00)	0.063
Creatinine (umol/L)	42.00 (35.00, 52.00)	43.00 (36.17, 61.50)	0.634
Sodium (mmol/L)	135.85 ± 5.92	136.30 ± 8.42	0.551
APTT (s)	34.40 (28.80, 44.05)	39.25 (29.23, 58.55)	< 0.001
PT (s)	12.30 (11.30, 14.22)	13.10 (11.80, 18.70)	0.002
D-dimer (mg/L)	0.94 (0.32, 2.42)	2.49 (0.88, 4.84)	0.007
Fib (g/L)	1.99 (1.50, 2.60)	1.94 (0.98, 2.94)	0.885
CRP (mg/L)	8.00 (4.00, 38.75)	22.50 (4.00, 61.00)	0.028
Lactate (mmol/L)	1.80 (1.10, 2.90)	2.60 (1.78, 5.63)	< 0.001
Albumin (g/L)	38.70 (33.50, 43.00)	33.10 (29.05, 40.20)	< 0.001
Vasopressor use during ICU stay time, n (%)	188 (34.18)	44 (55.00)	< 0.001
Hospital days	19.05 (10.95, 35.69)	10.23 (1.60, 31.72)	< 0.001
ICU days	5.81 (1.74, 17.62)	9.82 (1.54, 29.17)	0.758

ICU Intensive care unit, HR Heart rate, RR Respiratory rate, SBP Systolic pressure, DBP Diastolic pressure, WBC White blood cell, PLT Platelet, LDH Lactic dehydrogenase, pH Potential of hydrogen, PaCO₂ Arterial partial pressure of carbon dioxide, PaO₂ Arterial oxygen partial pressure, ALT Alanine transaminase, CK-MB Creatine kinase-MB, APTT Activated partial thromboplastin time, PT Prothrombin time, Fib Fibrinogen, CRP C reactive protein

Mortality

Table 5 provides a comprehensive description of the case fatality rates (CFRs) of the PICU population, including clinical and microbiological subgroups. A total of 9814 patients were admitted to our study, of whom 538 (5.48%) died in the PICU. Of these deceased patients, 14.87% (80/538) suffered from bacteremia. Children in the PICU with bacteremia had a CFR 2.5 times higher than that observed for children without bacteremia. The highest CFR (27.27%) was found in children with fungal septicemia in the PICU, compared to 8.88% for Gram-positive bacteremia and 22.41% for Gram-negative bacteremia with significant differences across the three groups (χ² = 28.0, df = 2, P < 0.001). A total of 25% of patients with persistent infections died during their hospital stay. Additionally, the CFR was high in cases with multiple infections (20.34%) and MDR infections (14.43%). Clinical characteristics of patients with new infection, persistent infection, and multidrug-resistant infection is shown in Supplementary Table 3.

Risk factors for mortality in patients with bacteremia in PICU

As shown in Table 6, PLT, LDH, pH, ALT, APTT, PT, D-dimer, CRP, lactate, albumin, and vasopressor use during ICU stay were identified as factors associated with mortality in patients with bacteremia. A multivariable logistic regression analysis demonstrated that PLT (OR 1.00, 95% CI 0.99–1.00; P = 0.009), APTT (OR 1.02, 95% CI 1.00–1.04; P = 0.049), and lactate (OR 1.15, 95% CI 1.00–1.32; P = 0.043) were independent factors significantly associated with mortality in these patients. Among them, APTT and lactate showed a positive association with the mortality in patients with bacteremia, while PLT showed a negative association.

Table 6.

Logistic regression analysis for associated factors of death in patients with bacteremia

	Univariate analysis		Multivariable analysis
	OR (95% CI)	P value	OR (95% CI)	P value
PLT (×109/L)	1.00 (1.00, 1.00)	<0.001	1.00 (0.99, 1.00)	0.009
LDH (U/L)	1.00 (1.00, 1.00)	0.037	1.00 (1.00, 1.00)	0.088
pH	0.02 (0.00, 0.12)	< 0.001	0.36 (0.01, 9.33)	0.537
ALT (U/L)	1.00 (1.00, 1.00)	0.039	1.00 (0.99, 1.00)	0.096
APTT (s)	1.02 (1.01, 1.03)	< 0.001	1.02 (1.00, 1.04)	0.049
PT (s)	1.04 (1.01, 1.07)	0.005	0.95 (0.89, 1.02)	0.131
D-dimer (mg/L)	1.05 (1.01, 1.09)	0.010	1.02 (0.96, 1.07)	0.554
CRP (mg/L)	1.01 (1.00, 1.01)	0.031	1.00 (0.99, 1.01)	0.668
Lactate (mmol/L)	1.22 (1.14, 1.31)	< 0.001	1.15 (1.00, 1.32)	0.043
Albumin (g/L)	0.94 (0.91, 0.97)	< 0.001	1.00 (0.95, 1.05)	0.925
Vasopressor use during ICU stay time	5.24 (2.02, 13.61)	< 0.001	2.37 (0.66, 8.52)	0.187

OR Odds ratio, CI Confidence interval, PLT Platelet, LDH Lactic dehydrogenase, pH Potential of hydrogen, ALT Alanine transaminase, APTT Activated partial thromboplastin time, PT Prothrombin time, CRP C reactive protein, ICU Intensive care unit

Discussion

Bacteremia, or bloodstream infection, is a serious condition caused by the entry of pathogenic microorganisms into the bloodstream. In severe cases, it can lead to sepsis and septic shock, potentially resulting in organ dysfunction and even death. Our study presented a 9-year analysis of bacteremia in the PICU of a hospital in Hangzhou, China, encompassing 9,814 patients, 630 of whom had bacteremia. We analyzed the epidemiology, characteristics of antibiotic resistance, and investigated the principal risk factors influencing the incidence and prognosis of patients with bacteremia in the PICU.

In our cohort, the average age of patients with bacteremia was 9.16 (IQR 2.68–38.45) months, younger than that of patients without bacteremia. Several studies have demonstrated that younger patients have a greater likelihood of developing bacteremia. A study among Australian children under the age of 18, which recorded 19.7 million cases between 2000 and 2019, found a significantly higher prevalence of bacteremia among infants and newborns compared to older children [20]. Another study noted that newborns and infants are more susceptible to bacteremia caused by Enterobacteriaceae due to their immature immune systems, and these infections are often associated with a higher incidence of antibiotic resistance [21].

Our research revealed that patients with bacteremia exhibited poorer coagulation function. Increasing evidence suggests that sepsis-associated coagulopathy (SAC), characterized by microvascular thrombosis and subsequent multiple organ failure (MOF), is closely linked to severe infection [22–25]. Prolonged coagulation times are common in critically ill patients. Additionally, previous research has demonstrated that PT and APTT can serve as predictors of sepsis and mortality in this population [26, 27].

CRP has been established as a biomarker of inflammation and is frequently used to help diagnose and treat sepsis based on extensive scientific research [28]. Regular testing of CRP has been shown to identify the efficacy of antibiotic therapy in critically ill patients, thereby reducing the duration of treatment [29]. In our study, CRP was identified as an independent factor significantly associated with bacteremia in the PICU. Similarly, previous studies have also shown that CRP may serve as an indicator of bacteremia in young infants and children [30, 31]. We also found that decreased albumin levels were an independent factor associated with bacteremia in the PICU. Serum albumin serves as an indicator of the body’s overall nutritional status and organ functional capacity. The underlying inflammatory state in bacteremia stimulates the release of inflammatory factors, which in turn cause the liver to produce less albumin. Patients with sepsis often experience capillary leakage, which can also lead to hypoalbuminemia [32]. Moreover, albumin has been shown to be a reliable diagnostic tool for predicting mortality in the PICU [33]. In a multi-center study, the albumin cut-off of 3.785 g/dl at admission to the PICUs was a sensitive and specific predictor of mortality and prognosis [34]. Although multiple potential predictors were analyzed, their individual AUC values, as shown in Supplementary Table 1 were relatively low. This reflects the current limitations of clinical prediction for bacteremia and underscores the ongoing challenge in developing more accurate diagnostic tools.

In our study, 728 pathogenic strains were isolated, including Gram-positive bacteria (70.05%, 510/728), Gram-negative bacteria (23.90%, 174/728), and fungi (6.04%, 44/728), similar to a six-year study from a children’s medical center in Eastern China [35]. Gram-positive bacteria, primarily Staphylococcus species, were the principal pathogens causing bacteremia in the PICU, with an incidence consistently exceeding that of Gram-negative bacteria and fungi. An increasing number of studies have demonstrated a rising incidence of bacteremia caused by Gram-positive bacteria [36–38]. Conversely, some studies have indicated that Gram-negative bacteria represent the predominant pathogenic agents [39, 40]. The observed differences in pathogen detection among patients with bacteremia may be attributed to several factors, such as the timing of the study, geographical location, and the specific objective of the study. Consequently, the results reflect the incidence within the research institution during that particular period. Furthermore, in accordance with previous studies [41], the most prevalent Gram-negative bacterium responsible for bacteremia in our study was Klebsiella species. Another study by Lee et al. revealed an increasing incidence of fungemia caused by fungi in recent years, with Candida being the most common [42]. Our study showed a lower incidence of fungal infection in children compared to bacterial infection, with Candida emerging as the predominant infectious pathogen. We also found a markedly elevated incidence of fungemia, exceeding the previously documented range of 7.8–15% [3, 43]. The risk factors for high candidemia prevalence includes prolonged mechanical ventilation, central venous catheter, extended hospital stay, parenteral nutrition support, and elevated Pediatric Risk of Mortality III (PRISM III) score [44].

Among the 630 patients with bacteremia, 571 were infected with a single pathogen (90.63%, 571/630), while 59 were infected with multiple pathogens (9.37%, 59/630). This multiple infection rate is comparable to that reported in previous studies, which ranged from 6% to 13% [45–47]. Multiple infections represent the most complex and serious form of infection in sepsis, often associated with inadequate treatment and poor prognosis [47]. MDR in the PICU is also a growing concern. A total of 100 MDR bacteria were isolated from the 728 pathogens identified in this study, including 35 strains of Staphylococcus species, 11 strains of Escherichia coli, 11 strains of Acinetobacter baumannii, 10 strains of Klebsiella species, 8 strains of Streptococcus species, 6 strains of Pseudomonas aeruginosa, 2 strains of Stenotrophomonas maltophilia, and 17 other strains. Consistent with our findings, the SENTRY antimicrobial surveillance program indicates that Staphylococcus aureus and Escherichia coli have been the primary pathogens associated with bacteremia globally over the past two decades. There has been a discernible decline in the proportion of Gram-positive organisms exhibiting drug resistance, including methicillin-resistant Staphylococcus aureus (MRSA) and vancomycin-resistant enterococci (VRE). Conversely, a notable increase has been observed in the incidence of Gram-negative organisms with MDR [48]. Our findings regarding multidrug-resistant organisms (MDROs) are clinically relevant, as they may influence the choice of empirical antibiotic therapy in critically ill pediatric patients. The observed resistance patterns highlight the need for careful consideration of local antibiotic susceptibility profiles when initiating empirical treatment. However, we did not further characterize specific resistance mechanisms such as extended-spectrum β-lactamase (ESBL) production, carbapenem resistance, or MRSA genotypes, which represents a limitation of our study.

We found 630 patients with bacteremia among 9,814 patients (6.42%), with 80 patients dying in hospital (12.70%). These data are consistent with findings from other studies that demonstrated a correlation between bacteremia and elevated mortality [1–3]. In our study, the highest CFR was observed in children with fungal septicemia (27.27%), compared to 22.41% for Gram-negative bacteremia and 8.88% for Gram-positive bacteremia. This CFR is comparable to the mortality rate of candidemia reported in a cohort study, which ranged from 10% to 49% [49]. These results highlight the critical need to focus efforts on treating children infected with fungi in the PICU.

Our analysis revealed that PLT, APTT, and lactate were independent factors significantly associated with mortality in patients with bacteremia. It has been shown that during the early stage of bacterial infection, there is a significant increase in PLT count, which later decreases to an excessively low level [50]. We found that decreased PLT levels were associated with increased mortality among patients with bacteremia. Similarly, Agrawal et al. observed a reduction in PLT numbers in pediatric patients, which was associated with poor clinical outcomes [51].

Benediktsson et al. demonstrated that prolonged APTT at the time of ICU admission was associated with elevated mortality in patients with sepsis [52]. A recent study indicated that APTT on ICU admission was significantly associated with acute kidney injury (AKI) in patients with septic shock caused by intra-abdominal infection (IAI) and was an independent predictor of 30-day mortality [53]. These findings provide compelling evidence supporting APTT as a biomarker of mortality in bacteremia.

Our study found that increased lactate was associated with mortality among patients with bacteremia. The severity of infection and the reversibility of organ dysfunction significantly impact the outcomes of patients with sepsis. The prognostic value of lactate has been evaluated in several studies [54, 55]. It has been demonstrated that serum lactate level is a valuable predictor of fatal outcomes in patients with critical illness [56]. Moreover, a recent study found that the lactate/albumin ratio is a useful predictor of mortality in children with nosocomial infections [57].

The following limitations are inherent in this study: [1] the retrospective design meant that data were collected from clinical records, rather than being specifically gathered for the defined variables, which may have led to inaccuracies and reporting bias; [2] as the included cases were derived from a single center, the results cannot be generalized to other regions; [3] the study did not analyze the resistance of major bacteria and fungi to commonly used antibiotics, and antibiotic treatments were not included; and [4] there were potential selection bias and residual confounding risk in our study. Furthermore, this study did not include risk factors and scoring data specific to the PICU, which may limit our ability to conduct a thorough analysis of the unique risk factors specific to the PICU.

In this large single-center retrospective study of PICU patients, we found that younger age, prolonged APTT, higher CRP, and lower albumin were independently associated with an increased risk of bacteremia. Among patients with bacteremia, lower platelet count, prolonged APTT, and elevated lactate were significant predictors of mortality. These findings underscore the importance of incorporating routinely available laboratory parameters into risk assessment to support timely recognition and management of bacteremia in critically ill children. Prospective multicenter studies are warranted to validate these predictors and to explore strategies that may improve outcomes in pediatric bacteremia.

Abbreviations

ALT

Alanine transaminase

APTT

Activated partial thromboplastin time

CFR

Case fatality rate

CI

Confidence interval

CLABSI

Central line-associated bloodstream infection

CRP

C-reactive protein

DBP

Diastolic blood pressure

HR

Heart rate

IAI

Intra-abdominal infection

ICU

Intensive care unit

IQR

Interquartile range

LDH

Lactate dehydrogenase

MDR

Multidrug-resistant

MOF

Multiple organ failure

MRSA

Methicillin-resistant Staphylococcus aureus

OR

Odds ratio

PaO2

Arterial oxygen partial pressure

pH

Potential of hydrogen

PIC

Pediatric intensive care

PICU

Pediatric intensive care unit

PLT

Platelet count

PT

Prothrombin time

RR

Respiratory rate

SAC

Sepsis-associated coagulopathy

SBP

Systolic blood pressure

SD

Standard deviation

VRE

Vancomycin-resistant enterococci

Authors’ contributions

GW and JLS contributed to the study conception and design, wrote the first draft, and revised the manuscript. YAF, YG and JQS performed the data analysis. RX and CZ contributed to data collection. All authors have read and approved the final manuscript.

Funding

This work was supported by the Natural Science Foundation of Shandong Province (ZR2023QH065).

Data availability

The datasets used or analyzed during the current study are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

This project was approved by both the Institutional Review Board of the Children’s Hospital of Zhejiang University School of Medicine and the Institutional Review Board of Qilu Hospital of Shandong University (KYLL-202202-027-1). Since the study findings do not impact clinical practice, individual consent was not required, which was also approved by the Institutional Review Board of the Children’s Hospital of Zhejiang University School of Medicine. All confidential health data were anonymized.

Competing interests

The authors declare no competing interests.