Risk factors for death caused by early onset sepsis in neonates: a retrospective cohort study

1) EOS was defined as a sepsis occurring within 72 h of birth, which includes culture-positive and clinically proven EOS neonates according to the consensus of Chinese experts (2019 version) [7]: ① Culture-positive EOS neonate is a neonate who has clinical manifestations consistent with sepsis:such as lethargy, hypothermia (< 36.5℃), hyperpyrexia (> 37.5℃), poor feeding, apnea, tachypnea, cyanosis, desaturation, bradycardia (heart rate < 100 bpm for more than 5 s), tachycardia (heart rate > 160 bpm for more than 5 s)and poor perfusion), and also has a positive blood culture in the first 72 h of life② Clinically proven EOS neonates is a neonate with a negative blood culture in the first 72 h, but has manifestations consistent with sepsis and also included any one of the following conditions: nonspecific blood tests (including leukocyte count, platelet count, C-reactive protein, and procalcitonin) were ≥ 2 positive, or the cerebrospinal fluid test was purulent meningitis changes. 2) The diagnosis of septic shock was based on infants with sepsis and hypotension requiring vasopressor therapy and lactate greater than 2 mmol/L despite adequate fluid resuscitation [8]. 3) The diagnosis of bacterial meningitis was based on neonates with positive CSF culture or some cases, although with negative CSF culture, but the initial blood culture grew a pathogen, and the CSF obtained 24 to 36 h after initiation of antibiotic therapy was abnormal [9, 10]. 4) Necrotizing enterocolitis (NEC, Bell stage 2 or 3) was defined according to the classification of Bell [11]. 5) An abnormal procalcitonin (PCT) level was defined as a level with a cut-off point with time according to the age adjustment requirement as in a previous study [12]. 6) Persistent pulmonary hypertension of the newborn (PPHN) was defined according to the clinical presentation of refractory hypoxemia and echocardiographic evidence with an estimated peak systolic pulmonary-artery pressure that was higher than 35 mmHg or more than two-thirds of the systemic systolic pressure as indicated by a tricuspid regurgitant jet, a right-to-left ductus arteriosus shunt, or a right-to-left atrial-level shunt [13] 0.7) Respiratory failure was diagnosed according to the consensus of Jen-Fu Hsu et al. [14] and our institution: ① Clinical manifestations of respiratory distress, and ② requirement respiratory support using a noninvasive or invasive ventilator to maintain a target arterial blood gas: pH value > 7.25, PaO₂ > 50 mmHg, PaCO₂ < 55 mmHg. 9) Diagnostic criteria of heart failure was meeting all the following clinical characteristics [15]: ① shortness of breath > 60 times/min; ② tachycardia > 160 times/min; ③ heart enlargement (X-ray showed cardiothoracic ratio > 0.6 or echocardiography proven); ④ pulmonary edema, and ⑤ liver enlargement > 3 cm or galloping rhythm.

This retrospective single-center cohort study was conducted in the Department of Neonatology at Children’s Hospital of Chongqing Medical University. Infants diagnosed with neonatal EOS between January 2020 and August 2021 were included in the present study.

Human studies were reviewed and approved by the Clinical Research Ethics Committee of Chongqing Medical University (Registration number: 2022/R/181). The Ethics Committee waived the requirement for informed consent due to the anonymized nature of the data and the scientific purpose of the study.

All infants enrolled were divided into two groups based on outcome before hospital discharge: non-survivors (Mortality group) and survivors (Survival group).

We tried to study the specific biomarkers for early recognition, and assessment of the risk factors of death for reducing the mortality related to neonatal EOS. Therefore, neonates were screened for inclusion: definite diagnosis of proven EOS [7] during 24–48 h after admission.

The exclusion criteria were as follows: 1) age on admission > 28 days. 2) reception of antibiotic therapy before admission to our NICU; 3) sepsis that occurred during hospitalization and considered a nosocomial infection; 4) clear genetic disease or chromosomal abnormality; 5) discharged against medical advice without sufficient antibiotic course, and 6) incomplete medical records.

The primary outcome variable was death before hospital discharge, and secondary outcomes were the first laboratory findings to our NICU among infants diagnosed with neonatal EOS.

Clinical data were obtained from the medical data platform of our institutional. Clinical and biological characteristics were retrospectively collected for each patient. The study data included baseline patient demographics, first laboratory data, acute comorbidities, and outcomes (survival or death).

Baseline patient demographics included meconium-stained amniotic fluid, premature rupture of membranes (PROM) > 18 h, mode of delivery, sex, gestational age, birth weight, weight on admission, and age on admission. Laboratory data including conventional inflammation markers (leukocyte count, neutrophil percentage, platelet count, C-reactive protein, and procalcitonin), arterial blood gas (pH, PCO₂, PO₂, HCO₃, base deficit, serum lactic acid, and blood sugar), electrolytes (potassium, calcium, magnesium, and sodium), renal function (urea nitrogen, and serum creatinine), and hepatic function (alanine aminotransferase, aspartate aminotransferase, total protein, and albumin) were measured as early as possible after admission. Blood cultures are done on each patient at the same time. During hospitalization, only the first time of blood samples after admission of laboratory data of each patient was considered and recorded in our study. Acute comorbidities including septic shock, heart failure, respiratory failure, bacterial meningitis, necrotizing enterocolitis, persistent pulmonary hypertension, congenital heart disease, and hypothermia.

We assume the mortality for neonatal sepsis was about 19%, which was based on the previous study reported by Fleischmann-Struzek C et al. [3]. According to the mortality of infants suffering from neonatal sepsis of Mortality group and Survival group in the case–control study, the sample size was estimated by the software PASS 15.0 (2017. NCSS LLC, 329 North 1000 East, Kaysville, Utah, 84,037, USA. www.​ncss.​com.), α = 0·05, and 1-β = 0·80. The estimated sample size was 34 cases per group, and the sample size in this study was larger than the estimated required number of cases.

All data were statistically analysed using SPSS version 26.0 (IBM, Armonk, New York) software. Categorical and continuous variables were expressed as proportions, mean ± standard deviation (M ± SD) and median (interquartile, IQR) respectively. Categorical variables were compared by the chi-square test or Fisher's exact test; Continuous variables were compared by the Mann–Whitney U-test and the t-test depending on the distributions. Moreover, the ROC curve analysis was performed in those first laboratory findings with p < 0.05. Prior to model development, baseline patient demographics, comorbidities (having a univariate result of p-value < 0.05), and the first laboratory findings (having a AUC > 0.5 and p-value < 0.05) were tested for multicollinearity by using variance inflation factor (VIF). A variable was removed having a VIF greater than 2.5 and VIF was repeated until all variables had a VIF less than or equal to 2.5 in Binary regression analysis. A p-value < 0.05 was considered statistically significant.

From January 2020 to August 2021, a total of 743 patients diagnosed proven neonatal EOS were admitted to our NICU. 187 neonates who met the exclusion criteria were excluded. Thus, 556 neonates were analyzed in the present study. There were 38 (6.8%) non-survivors who were divided into the Mortality group, and the other 518 (93.2%) survivors were divided into the Survival group. The inclusion process is shown in Fig. 1.

The Mortality group had significantly lower BW, lower weight on admission, smaller GA, and younger age on admission than the Survival group [1819.5 ± 932.3 g vs 2895.9 ± 847.2 g, p < 0.001; 1828.9 ± 1000.6 g vs 2867.2 ± 872.2 g, p < 0.001; 31.8 (26.7,40.3) weeks vs 38.6 (35.0,39.9)weeks, p < 0.001; 16.5 (11.3,20.0) hours vs 25.5 (11.8,64.5) hours, p < 0.001, respectively]. Moreover, cesarean section occurred more frequently in the Mortality group than the Survival group (76.3% vs 54.6%). In contrast, there were no differences in sex, meconium-stained amniotic fluid, or PROM, as shown in Table 1.

Septic shock, heart failure, respiratory failure, NEC, PPHN and hypothermia all occurred more frequently in the Mortality group than the Survival group (47.4% vs 5.8%, p < 0.001; 10.5% vs 0.4%, p < 0.001; 89.5% vs 46.5%, p < 0.001; 31.6% vs 17.8%, p = 0.035; 55.3% vs 13.5%, p < 0.001; 28.9% vs 3.7%, p < 0.001, respectively). On the other hand, there were no differences in the morbidity of bacterial meningitis, congenital heart disease, or ABO hemolytic disease between the two groups, as shown in Table 1.

Compared with the Survival group, the Mortality group had significantly lower level / percentage of those variables: platelet count (150.5 ± 75.6 *10⁹/L vs 255.7 ± 105.1 *10⁹/L, p < 0.001), neutrophil percentage (59.1 ± 14.4% vs 64.6 ± 16.5%,p = 0.045), CRP > 8 mg/L (10.5% vs 34.4%, p = 0.009), abnormal PCT ( 23.7% vs 34.7%,p < 0.001), pH [ 7.32 (7.21,7.40) vs 7.40 (7.32,7.46), p = 0.002], HCO₃ [20.4 (16.2,22.9) mmol/L vs 22.6 (20.9,24.4) mmol/L, p = 0.006], base deficit [-5.1 (-9.4,-3.2) mmol/L vs -2.9 (-5.2,-0.6) mmol/L, p = 0.002], Na [ 135.9(132.7,138.2) mmol/L vs 137(135.1,139.0) mmol/L, p = 0.045], Ca [2.1 (1.9,2.3) mmol/L vs 2.3 (2.1,2.4) mmol/L, p = 0.019], total protein (44.3 ± 8.7 g/L vs 53.1 ± 8.0 g/L, p < 0.001), and albumin ( 25.9 ± 6.1 g/L vs 31.9 ± 5.3 g/L, p < 0.001). Meanwhile, the Mortality group had significantly higher level of blood lactic acid [4.5 (2.8,9.1) mmol/L vs 2.6 (1.8,3.7) mmol/L, p < 0.001] and AST [ 64.4(45.9,107.9).

U/L vs 51.0 (36.2,73.3) U/L, p = 0.028]. On the contrary, other laboratory findings (including leukocyte count, PCO₂, PO₂, blood sugar, K, Mg, ALT, urea nitrogen, and serum creatinine) were not significantly different between the two groups (all p > 0.05) (Table 2).

ROC curve analysis was performed to further understand the relationship between those variables (having a p < 0.05) and death. Only serum lactic acid (having an AUC of 0.742 and at the cut-off value of 4.4 mmol/L) and AST (having an AUC of 0.632 and at the cut-off value of 45.6 U/L) were found obvious correlation with death, which had an AUC > 0.5 and p < 0.05. However, other values (including platelet count, neutrophil percentage, CRP > 8 mg/L, abnormal PCT, pH, HCO3, base deficit, Na, Ca, total protein, and albumin) did not prove this correlation, because they all can't be satisfied with an AUC > 0.5 and p < 0.05 at the same time (Table 3, Fig. 2).

Blood culture tests were performed in all infants, and positive cultures were obtained in 47 (8.5%) infants. There was no significant difference in positive rate of blood culture and multiple drug-resistant bacteria between the Mortality group and Survival group (10.5% vs 8.3%, p = 0.551; 25.0% vs 30.2%, p = 1.000, respectively). Gram-negative bacilli were the main pathogens in positive blood cultures, followed by CoNS, as shown in Table 4.

First, important covariates were derived from the univariate analysis and ROC curve analysis, which included demographic characteristics, and comorbidities (having a univariate result of p-value < 0.05) and the first traditional laboratory findings (having a AUC > 0.5 and p-value < 0.05). Hence, a total of 13 values, including BW, weight on admission, GA, age on admission, mode of delivery, septic shock, heart failure, respiratory failure, pulmonary hypertension, hypothermia, serum lactic acid, and AST were further analysed in the multivariable analysis by using variance inflation factor (VIF). After variance inflation factor analysis, two values (including weight on admission, and GA with a VIF > 2.5) were excluded. Therefore, only 11 values left were included in the final multivariate logistic regression.

Then, taking death as a reference, the binary logistic regression analysis showed that BW (OR = 1.001, 95% CI[ 1.001, 1.002], p < 0.001), PPHN (OR = 2.604, 95% CI[1.041, 6.515], p = 0.041), septic shock (OR = 6.151, 95% CI [2.523, 14.997], p < 0.001), heart failure (OR = 6.217, 95% CI [0.898, 43.049], p = 0.064), serum lactic acid (OR = 0.818, 95% CI [0.745, 0.898], p < 0.001), and AST (OR = 0.999, 95% CI [0.998, 1.000], p = 0.049) were independent risk factors for death. The best logistic regression model was administered: Logistic (Death) = -3.527 + 0.001 X1(BW) + 0.957 X2 (PPHN) + 1.817 X3 (septic shock) + 1.827 X4 (heart failure)-0.201 X5 (serum lactic acid)-0.001 X6 (AST) (X² = 113.096, p < 0.001), with 94.0% correct predictions, as shown in Table 5.