Background: preterm and critically ill neonates often experience clinically suspected sepsis during their prolonged hospitalization in the neonatal intensive care unit (NICU), which can be the initial sign of final adverse outcomes. Therefore, we aimed to utilize machine learning approaches to predict neonatal in-hospital mortality through data-driven learning. Methods: a total of 1095 neonates who experienced clinically suspected sepsis in a tertiary-level NICU in Taiwan between August 2017 and July 2020 were enrolled. Clinically suspected sepsis was defined based on clinical features and laboratory criteria and the administration of empiric antibiotics by clinicians. The variables used for analysis included patient demographics, clinical features, laboratory data, and medications. The machine learning methods used included deep neural network (DNN), k-nearest neighbors, support vector machine, random forest, and extreme gradient boost. The performance of these models was evaluated using the area under the receiver operating characteristic curve (AUC). Results: the final in-hospital mortality of this cohort was 8.2% (90 neonates died). A total of 765 (69.8%) and 330 (30.2%) patients were randomly assigned to the training and test sets, respectively. Regarding the efficacy of the single model that most accurately predicted the outcome, DNN exhibited the greatest AUC (0.923, 95% confidence interval [CI] 0.953–0.893) and the best accuracy (95.64%, 95% CI 96.76–94.52%), Cohen’s kappa coefficient value (0.74, 95% CI 0.79–0.69) and Matthews correlation coefficient value (0.75, 95% CI 0.80–0.70). The top three most influential variables in the DNN importance matrix plot were the requirement of ventilator support at the onset of suspected sepsis, the feeding conditions, and intravascular volume expansion. The model performance was indistinguishable between the training and test sets. Conclusions: the DNN model was successfully established to predict in-hospital mortality in neonates with clinically suspected sepsis, and the machine learning algorithm is applicable for clinicians to gain insights and have better communication with families in advance.

Serum lactate was used to predict the severity and outcome of neonatal hypoxic ischemic encephalopathy (HIE) before the era of therapeutic hypothermia (TH). There is no report on neurodevelopment (ND) outcome of neonates with HIE treated with TH in Taiwan. Between April 2011 and December 2012, newborn infants admitted to Chang Gung Memorial Hospital (CGMH), with gestational age > 35 weeks and birth weight ≥ 1800 g, who had acute perinatal events, evidence of significant fetal compromise, and ongoing clinical encephalopathy were prospectively enrolled for TH. Whole body cooling method was used to maintain the affected neonate's esophageal temperature at 33.5 ± 0.5 °C for 72 hours. Demographic data were recorded and hemogram, biochemical parameters, serum lactate, and creatine kinase (CK) were measured as well. Brain magnetic resonance imaging (MRI) was performed between 7 and 14 days of life. ND outcome of infants was evaluated by Bayley Scales of Infant Development, third edition (BSID-III) at 24 months of corrected age. Poor ND (PND) outcome was defined as infants surviving with either disability or ND delay. Seventeen patients were enrolled. Fifty-nine percent of babies (10/17) were born through cesarean section and 77% of babies (13/17) were transferred from outside hospitals. Six babies were moderate HIE and 11 babies were severe HIE. Among the 14 surviving patients, eight infants had PND outcome. There was no difference in demographic data between infants with and without PND. Serum level of lactate (mg/dL) after 72 hours of TH was higher (35.6 vs. 13.8, p = 0.042) in infants with PND. Neonates with abnormal brain MRI findings were also associated with PND (p = 0.01). This is the first report on ND outcome of neonates with HIE treated with TH in Taiwan. Higher serum level of lactate following TH and abnormal results of brain MRI are associated with poor ND outcome.

Group B Streptococcus (GBS) is an important pathogen that causes high mortality and morbidity in young infants. However, data on clinical manifestations between different GBS serotypes and correlation with molecular epidemiology are largely incomplete. The aim of this study was to determine the serotype distribution, antimicrobial resistance, clinical features and molecular characteristics of invasive GBS isolates recovered from Taiwanese infants. From 2003 to 2017, 182 non-duplicate GBS isolates that caused invasive disease in infants less than one year of age underwent serotyping, multilocus sequence typing (MLST) and antibiotic susceptibility testing. The clinical features of these infants with GBS disease were also reviewed. Of the 182 patients with invasive GBS disease, 41 (22.5%) were early-onset disease, 121 (66.5%) were late-onset disease and 20 (11.0%) were late late-onset disease (> 90 days of age). All these patients were treated with effective antibiotics on time. Among them, 51 (28.0%) had meningitis, 29 (16.0%) had neurological complications, 12 (6.6%) died during hospitalization, and 15 (8.8%) out of 170 patients who survived had long-term neurological sequelae at discharge. Serotype III GBS strains accounted for 64.8%, followed by serotype Ia (18.1%) and Ib (8.2%). MLST analysis revealed 11 different sequence types among the 182 isolates and ST-17 was the most dominant sequence type (56.6%). The correlation between serotype III and ST17 was evident, as ST17 accounted for 87.3% of all serotype III isolates. There was an obvious increasing trend of type III/ST-17 GBS that caused invasive disease in infants. All isolates were susceptible to penicillin, cefotaxime, and vancomycin, while 68.1 and 65.9% were resistant to erythromycin and clindamycin, respectively. Despite timely and appropriate antibiotic treatment, a significant proportion of invasive GBS disease still inevitably causes adverse outcomes. Further study to explore preventive strategies and development of serotype-based vaccines will be necessary in the future.

Retinopathy of prematurity (ROP), a vasoproliferative vitreoretinal disorder, is the leading cause of childhood blindness worldwide. Although angiogenic pathways have been the main focus, cytokine-mediated inflammation is also involved in ROP etiology. Herein, we illustrate the characteristics and actions of all cytokines involved in ROP pathogenesis. The two-phase (vaso-obliteration followed by vasoproliferation) theory outlines the evaluation of cytokines in a time-dependent manner. Levels of cytokines may even differ between the blood and the vitreous. Data from animal models of oxygen-induced retinopathy are also valuable. Although conventional cryotherapy and laser photocoagulation are well established and anti-vascular endothelial growth factor agents are available, less destructive novel therapeutics that can precisely target the signaling pathways are required. Linking the cytokines involved in ROP to other maternal and neonatal diseases and conditions provides insights into the management of ROP. Suppressing disordered retinal angiogenesis via the modulation of hypoxia-inducible factor, supplementation of insulin-like growth factor (IGF)-1/IGF-binding protein 3 complex, erythropoietin, and its derivatives, polyunsaturated fatty acids, and inhibition of secretogranin III have attracted the attention of researchers. Recently, gut microbiota modulation, non-coding RNAs, and gene therapies have shown promise in regulating ROP. These emerging therapeutics can be used to treat preterm infants with ROP.

Elevated C-reactive protein (CRP) level is widely used in clinical practice as a marker to distinguish between neonates with or without sepsis. However, some neonates with bacteremia have a CRP level within the normal range and they are not well characterized. All episodes of neonatal culture-proven bloodstream infections (BSIs) between July 2004 and June 2012 were enrolled. Patients characteristics were compared for three CRP groups (low, ≤ 10 mg/L; intermediate, 11–100 mg/L; and high, > 100 mg/L) using the Chi-square test and one-way ANOVA. The sepsis-attributable mortality rates were compared using logistic regression analyses. Of 986 episodes of neonatal BSI, 247 (25.1 %) had CRP ≤10 mg/L at the onset of clinical sepsis. In the low CRP group, patients had lower gestational age and birth weight, and an earlier occurrence of BSI. Patients with underlying gastrointestinal pathology, renal disorders, cholestasis, and pulmonary hypertension had a non-significant elevated CRP level at the onset of sepsis. In the blood culture of the low CRP group, coagulase-negative staphylococci (CoNS) were relatively more common (55.9 %, p < 0.001) than the other two groups, although one-fourth were infected with gram-negative bacilli (19.0 %), fungi (2.8 %), or polymicrobial pathogens (3.6 %). Of the BSIs with initial low CRP, 29.1 % were treated with inadequate antibiotics, 13.0 % progressed to septic shock, and 5.3 % had infectious complications. The sepsis-attributable mortality rate was lower in the low CRP group (4.9 %) than in the high CRP group (13.6 %). A considerable proportion of neonatal BSIs had a normal or low initial CRP level (≤10 mg/L), which was more likely to occur in low birth weight or extremely preterm infants, those with earlier onset of sepsis, and those infected with CoNS. Plasma CRP level should not be used to rule out severe culture-proven sepsis or guide the empirical choice of antibiotics.