Clinical, laboratory, and microbiological differences between early- and late-onset neonatal sepsis in preterm infants: a cross-sectional study in Vietnam

Abstract

Background

Although neonatal sepsis remains a significant cause of morbidity and mortality in preterm infants, few studies in Vietnam have compared the characteristics of early-onset sepsis (EOS) and late-onset sepsis (LOS).

Objective

To compare the clinical, laboratory, and microbiological characteristics of EOS and LOS in preterm neonates at a tertiary obstetric hospital in Vietnam.

Methods

This cross-sectional study included 106 preterm infants with culture-confirmed sepsis admitted from August 2022 to May 2024. Clinical data, laboratory results and blood cultures were analyzed. Statistical tests and univariate logistic regression were used to explore differences between EOS and LOS.

Results

Of 106 neonates, 24 (22.6%) had EOS, and 82 (77.4%) had LOS. LOS predominated in extremely preterm, very low birth weight infants and was associated with more severe symptoms, including respiratory distress and abnormal heart rate. EOS cases were more likely to have birth weight ≥ 1500 g and maternal conditions such as intrapartum fever, genital infections, contaminated amniotic fluid and prolonged amniotic fluid leakage (> 18 h). Laboratory analysis revealed higher levels of leukopenia, platelet and cerebrospinal fluid protein in EOS, as well as elevated C-reactive protein and blood protein in LOS. Gram-negative bacteria were predominant in both groups. Escherichia coli was most common in EOS (46%), whereas Klebsiella pneumoniae predominated in LOS (38%).

Conclusion

EOS and LOS in preterm neonates exhibit distinct patterns in perinatal factors, clinical manifestations, laboratory findings and microbiological profiles. Differentiating these two entities may aid in timely diagnosis and guide empirical treatment in settings with limited resources.

Introduction

Neonatal sepsis, defined as a severe systemic infection occurring within the first 28 days of life, remains a significant cause of neonatal morbidity and mortality worldwide, particularly in low- and middle-income countries [1, 2]. This critical condition can be caused by a wide range of pathogens, including bacteria, viruses, and fungi, and its development is influenced by both maternal and neonatal risk factors [2].

Clinically, neonatal sepsis is categorized into two distinct entities: early-onset sepsis (EOS), defined as infection occurring within the first 72 h of life, and late-onset sepsis (LOS), occurring after 72 h. These two forms differ significantly in terms of risk factors, sources of infection, causative organisms, clinical presentation, and outcomes. Specifically, EOS is often associated with vertical transmission from the mother, while LOS is more frequently related to postnatal or nosocomial acquisition. Understanding these distinctions is essential for optimizing prevention strategies, guiding empirical antibiotic selection, and refining treatment protocols.

In Vietnam, neonatal sepsis poses a substantial health burden, with reported rates of culture-confirmed sepsis ranging from 4 to 7% in hospitalized neonates and case—fatality rates reaching as high as 46% in certain populations [3–6]. However, most prior studies have focused on the overall sepsis burden without providing a detailed comparison between EOS and LOS. This represent a significant gap because, unlike high-income countries—where EOS is often associated with maternal factors and Group B Streptococcus (GBS)—studies in Vietnam have shown that Gram-negative bacteria such as Klebsiella pneumoniae, Acinetobacter spp., and Escherichia coli, predominante in both EOS and LOS cases, with GBS rarely isolated [3–5]. Furthermore, the high incidence of LOS and nosocomial infections in Vietnamese neonatal intensive care units highlight potential differences in infection control practices, antibiotic stewardship, and healthcare resources compared to developed countries.

Despite this critical insight into the local epidemiology, there remains a paucity of comprehensive, comparative data on the clinical characteristics, laboratory findings, and microbial profiles of EOS versus LOS in Vietnam, particularly among preterm infants—a high-risk population with unique vulnerabilities. Previous research in Vietnam has primarily examined sepsis in broader neonatal cohorts at pediatric hospitals, often overlooking the specific dynamics in obstetric settings where high-risk deliveries and immediate postpartum care predominate [3–6]. Therefore, this study aimed to addresses this gap by providing the first detailed comparative analysis of the clinical, laboratory, microbiological characteristics and to identify associated risk factors of early-onset and late-onset neonatal sepsis in preterm infants at a tertiary obstetric hospital in Vietnam. The specific aims are:

• Comparing the clinical, laboratory, and microbiological characteristics of EOS and LOS in preterm neonates to elucidate their distinct patterns.

• To identify maternal and neonatal risk factors associated with EOS versus LOS

Methods

Study design

This is a cross-sectional study conducted at the Hanoi Obstetrics and Gynecology Hospital, Vietnam, from August 2022 to May 2024. Hanoi Obstetrics and Gynecology Hospital is a tertiary hospital located in Hanoi, Vietnam, with a capacity of 1,000 beds. The hospital’s neonatal department has 100 beds and provides care for neonates with various conditions, including sepsis.

Study population

The study population was comprised of 106 preterm neonates admitted to the Department of Neonatology and diagnosed with neonatal sepsis using a convenient sampling technique during the study period. Inclusion criteria were neonates with age from 0–30 days at the time of sepsis evaluation and had symptoms of sepsis based on the 2010 European Medicines Agency criteria which required the presence of at least two clinical symptoms and at least two laboratory signs, in the context of a suspected or proven infection confirmed using positive blood culture [7]. The clinical criteria includes: (1) Abnormal body temperature: core > 38.5 °C or < 36 °C, and/or temperature instability, (2) Cardiovascular instability: bradycardia (heart rate < 10th percentile for age) or unexplained tachycardia (heart rate > 2 SD above normal) persistent for 0.5–4 h, (3) Hypotension: mean arterial pressure < 5th percentile for age or need for vasopressor support, (4) Reduced urine output: < 1 mL/kg/h, (5) Mottled skin or impaired peripheral perfusion, (6) Skin or subcutaneous lesions: petechiae, sclerema, or other suggestive lesions, (7) Respiratory instability: apnea episodes, tachypnea (> 2 SD above normal), increased oxygen requirement, or need for mechanical ventilation, (8) Gastrointestinal signs: feeding intolerance, abdominal distension, vomiting, or poor sucking, (9) Central nervous system signs: lethargy, irritability, hypotonia, or seizures. The laboratory criteria includes: (1) White blood cell (WBC) count: < 4,000 × 10⁹/L or > 20,000 × 10⁹/L, (2) Immature to total neutrophil (I/T) ratio: > 0.2, (3) Platelet count: < 100,000 × 10⁹/L, (4) C-reactive protein (CRP): > 15 mg/L, (5) Procalcitonin (PCT): ≥ 2 ng/mL (cut-off not fully validated in neonates), (6) Glucose intolerance: hyperglycemia (> 180 mg/dL or > 10 mmol/L) or hypoglycemia (< 45 mg/dL or < 2.5 mmol/L), (7) Metabolic acidosis: base excess < − 10 mEq/L or serum lactate > 2 mmol/L. Exclusion criteria were neonates with positive blood cultures due to external contamination. Neonates were classified as having EOS (sepsis within 72 h of birth) or LOS (sepsis after 72 h).

Data and sample collection

Data were collected from medical records and included demographic information, clinical signs and symptoms, laboratory results, and treatment outcomes. Neonatal risk factors recorded were gender, gestational age, and birth weight; maternal factors included type of delivery, multiple pregnancy, Apgar score, intrapartum fever, maternal genital infection, prolonged amniotic fluid leakage (> 18 h), and presence of foul-smelling amniotic fluid. Gestational age was calculated based on the first day of the mother’s last menstrual period, or if unavailable, estimated from first-trimester ultrasonography. Multiple pregnancy was defined as the presence of two or more fetuses. The Apgar score was recorded at 5 min after birth. Maternal genital infection was identified as diagnosed or receiving treatment for genital infection.

Clinical symptoms included body temperature, coma state, heart rate, shock, shortness of breath, Chest retraction, Apnea, Poor sucking reflex, Abdominal distension, jaundice, Scleroderma, Subcutaneous hemorrhage. Fever (hyperthermia) was defined as a core body temperature greater than 38.5 °C, while hypothermia was defined as a core temperature below 36 °C or marked temperature instability, characterized by fluctuations exceeding 2 °C within 24 h. Bradycardia was defined as a heart rate below the 10th percentile for age, or a persistent rate of less than 100 beats per minute in neonates, sustained for at least 0.5 to 4 h, and not attributable to transient causes such as apnea or procedures. In contrast, tachycardia was defined as a heart rate exceeding 2 standard deviations above the normal range for age, or greater than 180 beats per minute, persisting for at least 0.5 to 4 h and not explained by factors such as fever, crying, or medications.

Laboratory results comprised white blood cell count, neutrophil count, platelet count, C-reactive protein level, blood glucose level, blood pH, blood lactate level, base excess, cerebrospinal fluid white blood cell count, cerebrospinal fluid protein level, cerebrospinal fluid glucose level, and blood culture results to identify causative pathogens. All the neonates identified with sepsis will receive lumbar puncture according to the hospital’s protocol. However, there were cases where children are in very serious conditions such as septic shock, severe respiratory failure, severe blood clotting disorders, which do not meet the conditions and safety for cerebrospinal fluid testing.

Microbiological investigations

Blood samples (1–2 mL, depending on infant weight and clinical condition) were collected under sterile conditions, inoculated into aerobic culture bottles (BACT/ALERT PF plus), and transported to the microbiology laboratory for bacterial culture. In cases suspected fungal infection, an additional bottle (BACT/ALERT PF plus) was used for fungal culture. All inoculated bottles (BACT/ALERT) were incubated in an automated system until a positive growth signal was detected. Once flagged positive, a Gram stain was performed, and a portion of the sample was sub-cultured onto appropriate agar media to allow for colony isolation. The resulting colonies were subsequently identified using the VITEK 2 Compact system (BioMérieux, France) in accordance with the manufacturer’s instructions.

Statistical analysis

Statistical analyses were performed using R software (version 4.2.2). Descriptive statistics summarized categorical variables as frequencies and percentages and continuous variables as medians with interquartile ranges. The chi-square test or Fisher’s exact test compared categorical variables between groups, and the Wilcoxon rank-sum test compared continuous variables. Univariate logistic regression identified factors associated with EOS and LOS. A p-value < 0.05 was considered statistically significant.

Ethical considerations

The study was approved by the Ethics Committee of Hanoi Obstetrics and Gynecology Hospital (Approval No. CS/PSHN/DD/23/14). Informed consent was obtained from the parents or guardians of all neonates before data collection. All data were anonymized to protect participant privacy.

Results

Table 1 presents the characteristics of neonates with sepsis. Among 106 neonates with sepsis, 82 had LOS while 24 had EOS. There was no significant difference in gender distribution between the groups. However, significant differences were observed in gestation age and birth weight. Neonates with LOS were more commonly born preterm and with lower birth weight in comparison with EOS. The mode of delivery (vaginal vs. cesarean section) did not significantly differ between the groups. About 22% of neonates didn’t survive.

Table 1.

Characteristics of neonates with sepsis, Hanoi Obstetrics and Gynecology hospital, Vietnam (2022–2024)

Characteristic	Overall N = 1061	Late sepsis N = 821	Early sepsis N = 241	p-value2
Gender				0.2
Male	65 (61%)	53 (65%)	12 (50%)
Female	41 (39%)	29 (35%)	12 (50%)
Gestation age				0.027
< 32 weeks	88 (83%)	72 (88%)	16 (67%)
> = 32 weeks	18 (17%)	10 (12%)	8 (33%)
Weight at birth				0.018
< 1500 g	81 (76%)	67 (82%)	14 (58%)
> = 1500 g	25 (24%)	15 (18%)	10 (42%)
Type of delivery				0.6
Vaginal delivery	67 (63%)	53 (65%)	14 (58%)
Cesarean section	39 (37%)	29 (35%)	10 (42%)
Treatment result (died)	23 (22%)	21 (26%)	2 (8.3%)	0.071

¹n (%)

²Pearson's Chi-squared test; Fisher's exact test

The risk factors associated with EOS and LOS were presented in Table 2. In the univariate logistic regression model comparing EOS to LOS, several perinatal and maternal factors were independently associated with EOS. Higher gestational age (≥ 32 weeks) was significantly associated with early-onset sepsis (OR = 3.60, 95% CI: 1.21–10.6, p = 0.022), as was higher birth weight (≥ 1500 g; OR = 3.19, 95% CI: 1.18–8.60, p = 0.023). Among maternal factors, fever during labor (OR = 9.24, 95% CI: 2.80–33.9, p < 0.001), gynecological inflammation (OR = 5.79, 95% CI: 2.00–17.3, p = 0.001), prolonged amniotic fluid leakage > 18 h (OR = 4.61, 95% CI: 1.64–13.1, p = 0.004), and the presence of contaminated amniotic fluid (OR = 4.05, 95% CI: 1.03–16.0, p = 0.045) were also significantly associated with early-onset sepsis. In contrast, gender, type of delivery, multiple pregnancy, and Apgar score did not show significant associations (p > 0.05).

Table 2.

Univariate logistic regression analysis of risk factors associated with EOS and LOS among neonates with sepsis, Hanoi Obstetrics and Gynecology hospital, Vietnam (2022–2024)

Early vs Late onset sepsis	N	OR	95% CI	p-value
Gender (Female vs Male)	106	1.83	0.73, 4.63	0.2
Gestational age (≥ 32 weeks vs < 32 weeks)	106	3.60	1.21, 10.6	0.022
Birth weight (≥ 1500 g vs < 1500 g)	106	3.19	1.18, 8.60	0.023
Type of delivery (Cesarean vs Vaginal)	106	1.31	0.51, 3.29	0.6
Multiple pregnancy (Yes vs No)	106	2.23	0.78, 6.12	0.1
Apgar score	106	1.04	0.78, 1.43	0.8
Maternal fever during labor (Yes vs No)	106	9.24	2.80, 33.9	< 0.001
Maternal gynecological inflammation (Yes vs No)	106	5.79	2.00, 17.3	0.001
Amniotic fluid leakage > 18 h (Yes vs No)	106	4.61	1.64, 13.1	0.004
Contaminated amniotic fluid (Yes vs No)	106	4.05	1.03, 16.0	0.045

Abbreviations: CI Confidence Interval, OR Odds Ratio

A summary of the clinical symptoms is provided in Table 3. Clinical presentations differed between neonates with EOS and LOS. Although body temperature abnormalities were not significantly different between groups (p = 0.10), hypothermia was observed only in LOS cases (7.3%), whereas fever was more common in LOS (33%) than in EOS (17%). The distribution of coma states did not differ significantly (p = 0.5), with lethargy and irritability seen in both groups. Heart rate abnormalities showed a significant difference (p = 0.035), with tachycardia being more frequently observed in LOS (44%) compared to EOS (17%), while bradycardia was rare in both groups. Shock was observed in 26% of the total cohort, with no significant differences between groups (p = 0.5).

Table 3.

Clinical symptom characteristics of neonates with sepsis, Hanoi Obstetrics and Gynecology hospital, Vietnam (2022–2024)

Characteristic	Overall N = 1061	Late sepsis N = 821	Early sepsis N = 241	p-value2
Body temperature				0.1
Normal	69 (65%)	49 (60%)	20 (83%)
Hypothermia	6 (5.7%)	6 (7.3%)	0 (0%)
Fever	31 (29%)	27 (33%)	4 (17%)
Coma state				0.5
Normal	43 (41%)	31 (38%)	12 (50%)
Lethargic	43 (41%)	33 (40%)	10 (42%)
Irritable	13 (12%)	12 (15%)	1 (4.2%)
Convulsions	7 (6.6%)	6 (7.3%)	1 (4.2%)
Heart rate				0.035
Normal	61 (58%)	42 (51%)	19 (79%)
Bradycardia	5 (4.7%)	4 (4.9%)	1 (4.2%)
Tachycardia	40 (38%)	36 (44%)	4 (17%)
Shock	28 (26%)	23 (28%)	5 (21%)	0.5
Shortness of breath	58 (55%)	41 (50%)	17 (71%)	0.071
Chest retraction	45 (42%)	28 (34%)	17 (71%)	0.001
Apnea	50 (47%)	45 (55%)	5 (21%)	0.003
Poor sucking reflex	97 (92%)	78 (95%)	19 (79%)	0.027
Abdominal distension	31 (29%)	27 (33%)	4 (17%)	0.1
Jaundice	62 (58%)	44 (54%)	18 (75%)	0.062
Scleroderma	18 (17%)	16 (20%)	2 (8.3%)	0.4
Subcutaneous hemorrhage	16 (15%)	14 (17%)	2 (8.3%)	0.5

¹n (%)

²Fisher's exact test; Pearson's Chi-squared test

Signs of respiratory distress varied depending on the onset of sepsis. Shortness of breath was observed more common in EOS (71%) than LOS (50%), though the difference was not statistically significant (p = 0.071). However, chest retraction (71% vs. 34%, p = 0.001) and apnea (21% vs. 55%, p = 0.003) were significantly more frequent in the LOS group. A poor sucking reflex was more commonly observed in LOS (95%) compared to EOS (79%), with the difference reaching statistical significance (p = 0.027). Other symptoms, including abdominal distension (p = 0.12), jaundice (p = 0.062), scleroderma (p = 0.4), and subcutaneous hemorrhage (p = 0.5), did not show significant differences between the groups.

As illustrated in Table 4, neonates with EOS had lower neutrophil counts (median 2,990 vs. 6,527 n/mm³, p = 0.022), higher platelet counts (230,000 vs. 128,500 n/mm³, p = 0.001), and lower C-reactive protein levels (3 mg/L vs. 29 mg/L, p < 0.001) compared to those with LOS. Blood glucose was also significantly lower in the EOS group (3.8 mmol/L vs. 7.5 mmol/L, p = 0.001). 81 neonates received lumbar puncture for cerebrospinal fluid examination. In which, cerebrospinal fluid protein levels were higher in EOS (1.41 g/L vs. 1.14 g/L, p = 0.046), while cerebrospinal fluid white blood cell and glucose concentrations did not differ significantly between groups (p = 0.8 and p = 0.10, respectively). No significant differences were observed in white blood cell (p = 0.058), blood pH (p = 0.8), blood lactate (p = 0.3), or base excess values (p = 0.3).

Table 4.

Laboratory results of neonates with sepsis, Hanoi Obstetrics and Gynecology hospital, Vietnam (2022–2024)

Characteristic	Overall N = 1061	Late sepsis N = 821	Early sepsis N = 241	p-value2
White blood cell (n/mm3)	10,950 (5,300, 16,600)	11,700 (5,900, 17,400)	7,800 (4,250, 12,250)	0.058
Neutrophil (n/mm3)	5,769 (2,415, 11,571)	6,527 (2,848, 11,965)	2,990 (1,978, 6,121)	0.022
Platelet (n/mm3)	150,500 (87,000, 237,000)	128,500 (79,000, 188,000)	230,000 (165,000, 267,500)	0.001
C-reactive protein (mg/L)	16 (2, 63)	29 (3, 69)	3 (0, 12)	< 0.001
Blood glucose (mmol/L)	6.3 (3.8, 11.4)	7.5 (4.5, 12.6)	3.8 (3.3, 6.8)	0.001
Blood pH	7.26 (7.17, 7.37)	7.26 (7.17, 7.37)	7.25 (7.17, 7.38)	0.8
Blood lactate (mmol/L)	2.40 (1.40, 3.90)	2.40 (1.35, 3.70)	2.70 (1.70, 3.90)	0.3
Base excess (mEq/L)	−8 (−11, −4)	−8 (−11, −5)	−6 (−10, −3)	0.3
Lumbar puncture performed	81 (76%)	64 (78%)	17 (71%)	0.5
Cerebrospinal fluid white blood cell (n/mm3)	15 (7, 40)	17 (8, 42)	11 (7, 38)	0.8
Cerebrospinal fluid protein (g/L)	1.17 (0.95, 1.58)	1.14 (0.91, 1.55)	1.41 (1.04, 2.01)	0.046
Cerebrospinal fluid glucose (mmol/L)	3.11 (2.21, 4.23)	3.15 (2.25, 4.92)	2.67 (1.92, 3.36)	0.1

¹Median (Q1, Q3); n (%)

²Wilcoxon rank sum test; Fisher's exact test

Bacterial profiles differed between EOS and LOS among neonates (Table 5). Klebsiella pneumoniae was significantly more common in LOS compared to EOS (38% vs. 8.3%, p = 0.006), while Proteus mirabilis was isolated exclusively in EOS cases (8.3%), although this difference did not reach statistical significance (p = 0.050). Escherichia coli was the most frequently isolated pathogen overall (36%), with a higher proportion in EOS (46%) than in LOS (33%), though this difference was not statistically significant (p = 0.2). Enterobacter cloacae were found only in LOS cases (4.9%), and Group B Streptococcus was also detected exclusively in LOS cases (3.7%), although these differences were not statistically significant. Several less common organisms were identified only in EOS cases, including Sphingomonas paucimobilis, Achromobacter species, Acinetobacter baumannii, Listeria monocytogenes, Streptococcus mitis, and Streptococcus gallolyticus. However, their low frequencies precluded the determination of statistical significance. Staphylococcus aureus was more common in LOS (12%) compared to EOS (4.2%), though the difference was not significant (p = 0.4). Candida albicans was isolated in 4.7% of cases, with no meaningful difference between groups.

Table 5.

Causus pathogens of neonates with sepsis, Hanoi Obstetrics and Gynecology hospital, Vietnam (2022–2024)

Characteristic	Overall N = 1061	Late sepsis N = 821	Early sepsis N = 241	p-value2
K. pneumoniae	33 (31%)	31 (38%)	2 (8.3%)	0.006
E. coli	38 (36%)	27 (33%)	11 (46%)	0.2
Enterobacter cloacae	4 (3.8%)	4 (4.9%)	0 (0%)	0.6
Proteus mirabilis	2 (1.9%)	0 (0%)	2 (8.3%)	0.050
Serratia marcescens	1 (0.9%)	1 (1.2%)	0 (0%)	1.0
Stenotrophomonas maltophilia	2 (1.9%)	2 (2.4%)	0 (0%)	1.0
Sphingomonas paucimobilis	1 (0.9%)	0 (0%)	1 (4.2%)	0.2
Achromobacter denitrificans	1 (0.9%)	0 (0%)	1 (4.2%)	0.2
Achromobacter xylosoxidans	1 (0.9%)	0 (0%)	1 (4.2%)	0.2
Acinetobacter baumannii	1 (0.9%)	0 (0%)	1 (4.2%)	0.2
Group B Streptococcus	3 (2.8%)	3 (3.7%)	0 (0%)	1.0
S. aureus	11 (10%)	10 (12%)	1 (4.2%)	0.4
Streptococcus mitis	1 (0.9%)	0 (0%)	1 (4.2%)	0.2
Streptococcus gallolyticus subspecies gallolyticus	1 (0.9%)	0 (0%)	1 (4.2%)	0.2
Listeria monocytogenes	1 (0.9%)	0 (0%)	1 (4.2%)	0.2
Candida albicans	5 (4.7%)	4 (4.9%)	1 (4.2%)	1.0

¹n (%)

²Pearson's Chi-squared test; Fisher's exact test

Discussion

Clinical characteristics and risk factors

This study analyzed the clinical characteristics, risk factors, and laboratory parameters of neonatal sepsis cases at a tertiary obstetrics and gynecology hospital in Vietnam, highlighting the distinct differences between early-onset sepsis (EOS) and late-onset sepsis (LOS).

Significant differences emerged in gestational age and birth weight. Neonates with gestational age of ≥ 32 weeks and birth weight ≥ 1500 g were more prevalent with EOS compared to LOS. Univariate analysis confirmed higher odds of EOS in these categories. Conversely, LOS was more common in extremely preterm and very low birth weight infants. These findings suggest that prematurity and low birth weight exert a greater influence on the development of LOS, likely due to the underdeveloped immune systems of preterm neonates, compounded by limitations in postnatal care within resource-constrained settings. This mirrors studies by Kruse³ et al. (2013) and Tran⁴ et al. (2015), where LOS was associated with prematurity and invasive procedures. Pathophysiologically, preterm infants have immature immune systems, thin skin barriers, and dependencies on mechanical ventilation and central lines, heightening nosocomial infection risks [8, 9].

Maternal infections and complications during delivery, including fever, gynecological inflammation, prolonged amniotic fluid leakage, and contaminated amniotic fluid, were associated with early-onset sepsis (EOS), likely enabling vertical pathogen transmission from mother to neonate, as similarly reported in a multi-country study of neonatal sepsis in low-income settings [2]. Although vaginal delivery may increase the likelihood of neonatal exposure to maternal genital flora, as suggested by a recent study from Amare [10] et al. (2024), our data showed no significant difference in delivery mode between EOS and LOS (58% vs. 65% vaginal delivery, p = 0.6). This finding implies that maternal factors such as intrapartum fever, genital tract infection, and prolonged amniotic fluid leakage (> 18 h) may have a greater influence on the occurrence of EOS in this population. Treatment outcomes revealed a 22% overall mortality rate, higher in LOS (26%) than EOS (8.3%), though not statistically significant. This trend suggests worse prognosis in LOS, possibly due to underlying prematurity and comorbidities, aligning with global data where LOS mortality can reach 20–40% in low-resource settings [1, 2, 11].

These observations emphasize the need for targeted interventions: improved antenatal screening and intrapartum antibiotic prophylaxis to mitigate EOS risks, and enhanced NICU hygiene, reduced invasive device use, and antibiotic stewardship to curb LOS. The hospital’s high-risk patient profile necessitates a dual approach to address both maternal and neonatal factors.

Clinical manifestations

The clinical differences between EOS and LOS provide valuable clues for early recognition and management in resource-constrained settings. The variation in temperature abnormalities, with fever more common in LOS and hypothermia exclusive to it, suggests that temperature instability is a less reliable diagnostic marker for EOS, particularly in preterm infants who may not mount robust febrile responses early in life. This aligns with the understanding that preterm neonates often present with subtle or atypical signs, complicating early diagnosis [9].

Cardiovascular manifestations, particularly tachycardia in LOS, likely reflect the systemic inflammatory response associated with nosocomial infections, which often develop in the context of prolonged illness and comorbidities. The absence of significant differences in shock prevalence suggests that both EOS and LOS can progress to severe systemic compromise, necessitating vigilant monitoring regardless of onset type.

Respiratory symptoms further distinguish between the two conditions. The higher prevalence of chest retraction in EOS may indicate acute respiratory distress related to congenital pneumonia or maternal pathogen transmission, a common feature in early infections. Conversely, the predominance of apnea in LOS aligns with the chronic respiratory vulnerabilities of extremely preterm infants, who often have underlying lung immaturity exacerbated by infection [9]. The significantly higher rate of poor sucking reflex in LOS likely reflects greater illness severity, neurological compromise, and nutritional challenges in these infants, who face prolonged NICU stays.

The nonspecific nature of other symptoms, such as abdominal distension and jaundice, underscores the diagnostic challenge in neonatal sepsis, particularly in preterm infants where clinical signs overlap with other conditions. These findings suggest that EOS may present with acute respiratory symptoms, while LOS manifests with more severe, multisystem involvement due to prematurity and nosocomial exposure.

Laboratory findings

Laboratory parameters offer critical insights into the pathophysiology and severity of neonatal sepsis, with distinct profiles for EOS and LOS. The trend toward lower white blood cell and neutrophil counts in EOS suggests early immunosuppression or neutrophil consumption, likely due to rapid pathogen dissemination from maternal sources. In contrast, the higher prevalence of thrombocytopenia and elevated C-reactive protein (CRP) in LOS reflects chronic inflammation and hematologic derangements associated with nosocomial infections, which often involve prolonged exposure to pathogens and invasive procedures [12].

The lower CRP in EOS is consistent with its delayed rise in early infections, limiting its diagnostic utility in the first 72 hours [13]. This highlights the need for alternative biomarkers or combined diagnostic approaches in EOS. The higher blood glucose in LOS, potentially indicating hyperglycemia from stress and inflammation, contrasts with the risk of hypoglycemia in EOS, possibly due to acute metabolic stress. The lack of differences in acid–base parameters suggests that metabolic acidosis is a common feature of severe sepsis, regardless of onset, reflecting impaired tissue perfusion and cellular hypoxia [12].

Higher cerebrospinal fluid protein levels in EOS may suggest early meningeal inflammation, potentially due to vertical pathogen transmission, though similar cerebrospinal fluid white blood cell and glucose levels in EOS and LOS limit the ability of cerebrospinal fluid analysis to differentiate between them [9]. No differences in acid–base parameters indicate similar metabolic acidosis severity in both groups [12].

Microbiological etiology

The predominance of Gram-negative bacteria, particularly Escherichia coli in EOS and Klebsiella pneumoniae in LOS, reflects the distinct transmission pathways: maternal–fetal for EOS and nosocomial for LOS. This pattern is consistent with the epidemiology of neonatal sepsis in low- and middle-income countries, where Gram-negative organisms dominate due to environmental and hospital-related factors [14–16]. The exclusive presence of Proteus mirabilis in EOS (8.3%, p = 0.050) suggests transmission from maternal gastrointestinal or vaginal flora, while the higher prevalence of K. pneumoniae in LOS indicates hospital-acquired infections, likely due to suboptimal infection control [1, 17].

The low prevalence of Group B Streptococcus in EOS, despite its prominence in high-income settings, likely reflects the impact of Group B Streptococcus screening and intrapartum prophylaxis at the hospital where the study was conducted. This contrasts with regional studies where Gram-positive organisms, like coagulase-negative staphylococci, dominate in some Asian centers [15]. The variability in pathogen profiles across Vietnamese hospitals suggests local factors, such as climate, patient density, and antibiotic use, shape the microbiological landscape. Vietnam’s humid tropical environment may favor Gram-negative proliferation, while high antibiotic use in tertiary centers could drive resistance [17, 18].

These microbiological distinctions have critical implications for empirical antibiotic therapy. For EOS, regimens targeting E. coli (e.g., ampicillin and gentamicin) are appropriate, while LOS requires coverage for K. pneumoniae and other nosocomial pathogens (e.g., vancomycin and meropenem). The high Gram-negative burden underscores the urgent need for hospital-specific surveillance to monitor resistance patterns and guide stewardship, particularly in high-volume NICUs where multidrug-resistant strains may emerge [18].

Study limitations

Several caveats merit mention when interpreting the findings. Firstly, this is a study with a small sample size of 106 and skewed toward LOS, with absence of multivariate analysis to adjust for confounders. Secondly, this was a single-center study conducted in a tertiary hospital, so the findings may not be generalizable to primary care or non-referral settings. Thirdly, the reported pathogen data thus pertain only to culture-confirmed infections. Finally, as an obstetric hospital, there may have been a higher proportion of complicated deliveries and premature infants, which could influence the observed risk factor profile for neonatal sepsis in this study population.

Recommendations for future research

Future studies need to expand this finding on a bigger dataset from a variety of hospitals and explore the influence of these factors on mortality outcomes.

Conclusion

This study confirms clear differences in causes and risk factors between early-onset and late-onset sepsis in preterm babies. In these infants, the pathogens involved include a mix of Gram-negative bacteria, Gram-positive bacteria, and the yeast Candida albicans, with Gram-negative bacteria, especially Escherichia coli and Klebsiella pneumonia, being the most common. Early-onset sepsis is related to maternal fever during labor, infections in the birth canal, amniotic fluid leakage for more than 18 h before birth, and contaminated amniotic fluid, whereas late-onset sepsis is mainly seen in premature babies (under 32 weeks) and those weighing less than 1,500 g at birth. These findings stress the need to understand local infection patterns so that hospitals can choose the right antibiotics and improve infection-control measures in busy tertiary centers.

Abbreviations

EOS

Early on-set sepsis

LOS

Late on-set sepsis

NICU

Neonatal Intensive Care Unit

Authors’ contributions

Hung Mai Trong and Huy Do Duc designed the study and revised the manuscript draft. Lam Nguyen Duc and Nhan Tran Luong supervised data collection. Lam Nguyen Duc, Nhan Tran Luong, Phuong Pham Thi Thu, Anh Tran Tuan, Duc Do Thanh collected data. Huy Do Duc analyzed the data. Nga Nguyen Thi Quynh and Duc Do Thanh drafted the manuscript. All authors approved the final manuscript.

Funding

This study did not receive any funding.

Data availability

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

Declarations

Ethics approval and consent to participate

Our study adhered to the Declaration of Helsinki.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.