Machine Learning Used to Compare the Diagnostic Accuracy of Risk Factors, Clinical Signs and Biomarkers and to Develop a New Prediction Model for Neonatal Early-onset Sepsis.

BACKGROUND: Current strategies for risk stratification and prediction of neonatal early-onset sepsis (EOS) are inefficient and lack diagnostic performance. The aim of this study was to use machine learning to analyze the diagnostic accuracy of risk factors (RFs), clinical signs and biomarkers and to develop a prediction model for culture-proven EOS. We hypothesized that the contribution to diagnostic accuracy of biomarkers is higher than of RFs or clinical signs. STUDY DESIGN: Secondary analysis of the prospective international multicenter NeoPInS study. Neonates born after completed 34 weeks of gestation with antibiotic therapy due to suspected EOS within the first 72 hours of life participated. Primary outcome was defined as predictive performance for culture-proven EOS with variables known at the start of antibiotic therapy. Machine learning was used in form of a random forest classifier. RESULTS: One thousand six hundred eighty-five neonates treated for suspected infection were analyzed. Biomarkers were superior to clinical signs and RFs for prediction of culture-proven EOS. C-reactive protein and white blood cells were most important for the prediction of the culture result. Our full model achieved an area-under-the-receiver-operating-characteristic-curve of 83.41% (±8.8%) and an area-under-the-precision-recall-curve of 28.42% (±11.5%). The predictive performance of the model with RFs alone was comparable with random. CONCLUSIONS: Biomarkers have to be considered in algorithms for the management of neonates suspected of EOS. A 2-step approach with a screening tool for all neonates in combination with our model in the preselected population with an increased risk for EOS may have the potential to reduce the start of unnecessary antibiotics.

Cost impact of procalcitonin-guided decision making on duration of antibiotic therapy for suspected early-onset sepsis in neonates.

BACKGROUNDS: The large, international, randomized controlled NeoPInS trial showed that procalcitonin (PCT)-guided decision making was superior to standard care in reducing the duration of antibiotic therapy and hospitalization in neonates suspected of early-onset sepsis (EOS), without increased adverse events. This study aimed to perform a cost-minimization study of the NeoPInS trial, comparing health care costs of standard care and PCT-guided decision making based on the NeoPInS algorithm, and to analyze subgroups based on country, risk category and gestational age. METHODS: Data from the NeoPInS trial in neonates born after 34 weeks of gestational age with suspected EOS in the first 72 h of life requiring antibiotic therapy were used. We performed a cost-minimization study of health care costs, comparing standard care to PCT-guided decision making. RESULTS: In total, 1489 neonates were included in the study, of which 754 were treated according to PCT-guided decision making and 735 received standard care. Mean health care costs of PCT-guided decision making were not significantly different from costs of standard care (€3649 vs. €3616). Considering subgroups, we found a significant reduction in health care costs of PCT-guided decision making for risk category 'infection unlikely' and for gestational age ≥ 37 weeks in the Netherlands, Switzerland and the Czech Republic, and for gestational age < 37 weeks in the Czech Republic. CONCLUSIONS: Health care costs of PCT-guided decision making of term and late-preterm neonates with suspected EOS are not significantly different from costs of standard care. Significant cost reduction was found for risk category 'infection unlikely,' and is affected by both the price of PCT-testing and (prolonged) hospitalization due to SAEs.

Risk Factors for Retinopathy of Prematurity in the Netherlands: A Comparison of Two Cohorts.

INTRODUCTION: Retinopathy of prematurity (ROP) remains an important cause for preventable blindness. Aside from gestational age (GA) and birth weight, risk factor assessment can be important for determination of infants at risk of (severe) ROP. METHODS: Prospective, multivariable risk-analysis study (NEDROP-2) was conducted, including all infants born in 2017 in the Netherlands considered eligible for ROP screening by pediatricians. Ophthalmologists provided data of screened infants, which were combined with risk factors from the national perinatal database (Perined). Clinical data and potential risk factors were compared to the first national ROP inventory (NEDROP-1, 2009). During the second period, more strict risk factor-based screening inclusion criteria were applied. RESULTS: Of 1,287 eligible infants, 933 (72.5%) were screened for ROP and matched with the Perined data. Any ROP was found in 264 infants (28.3% of screened population, 2009: 21.9%) and severe ROP (sROP) (stage ≥3) in 41 infants (4.4%, 2009: 2.1%). The risk for any ROP is decreased with a higher GA (odds ratio [OR] 0.59 and 95% confidence interval [CI] 0.54-0.66) and increased for small for GA (SGA) (1.73, 1.11-2.62), mechanical ventilation >7 days (2.13, 1.35-3.37) and postnatal corticosteroids (2.57, 1.44-4.66). For sROP, significant factors were GA (OR 0.37 and CI 0.27-0.50), SGA (OR 5.65 and CI 2.17-14.92), postnatal corticosteroids (OR 3.81 and CI 1.72-8.40), and perforated necrotizing enterocolitis (OR 7.55 and CI 2.29-24.48). CONCLUSION: In the Netherlands, sROP was diagnosed more frequently since 2009. No new risk factors for ROP were determined in the present study, apart from those already included in the current screening guideline.

Increase in treatment of retinopathy of prematurity in the Netherlands from 2010 to 2017.

PURPOSE: Compare patients treated for Retinopathy of Prematurity (ROP) in two consecutive periods. METHODS: Retrospective inventory of anonymized neonatal and ophthalmological data of all patients treated for ROP from 2010 to 2017 in the Netherlands, subdivided in period (P)1: 1-1-2010 to 31-3-2013 and P2: 1-4-2013 to 31-12-2016. Treatment characteristics, adherence to early treatment for ROP (ETROP) criteria, outcome of treatment and changes in neonatal parameters and policy of care were compared. RESULTS: Overall 196 infants were included, 57 infants (113 eyes) in P1 and 139 (275 eyes) in P2, indicating a 2.1-fold increase in ROP treatment. No differences were found in mean gestational age (GA) (25.9 ± 1.7 versus 26.0 ± 1.7 weeks, p = 0.711), mean birth weight (791 ± 311 versus 764 ± 204 grams, p = 0.967) and other neonatal risk factors for ROP. In P2, the number of premature infants born <25 weeks increased by factor 1.23 and higher oxygen saturation levels were aimed at in most centres. At treatment decision, 59.6% (P1) versus 83.5% (P2) (p = 0.263) infants were classified as Type 1 ROP (ETROP classification). Infants were treated with laser photocoagulation (98 versus 96%) and intravitreal bevacizumab (2 versus 4%). Retreatment was necessary in 10 versus 21 (p = 0.160). Retinal detachment developed in 6 versus 13 infants (p = 0.791) of which 2 versus 6 bilateral (p = 0.599). CONCLUSION: In period 2, the number of infants treated according to the ETROP criteria (Type 1) increased, the number of ROP treatments, retinal detachments and retreatments doubled and the absolute number of retinal detachments increased. Neonatal data did not provide a decisive explanation, although changes in neonatal policy were reported.

C-Reactive Protein, Procalcitonin, and White Blood Count to Rule Out Neonatal Early-onset Sepsis Within 36 Hours: A Secondary Analysis of the Neonatal Procalcitonin Intervention Study.

BACKGROUND: Neonatal early-onset sepsis (EOS) is one of the main causes of global neonatal mortality and morbidity, and initiation of early antibiotic treatment is key. However, antibiotics may be harmful. METHODS: We performed a secondary analysis of results from the Neonatal Procalcitonin Intervention Study, a prospective, multicenter, randomized, controlled intervention study. The primary outcome was the diagnostic accuracy of serial measurements of C-reactive protein (CRP), procalcitonin (PCT), and white blood count (WBC) within different time windows to rule out culture-positive EOS (proven sepsis). RESULTS: We analyzed 1678 neonates with 10 899 biomarker measurements (4654 CRP, 2047 PCT, and 4198 WBC) obtained within the first 48 hours after the start of antibiotic therapy due to suspected EOS. The areas under the curve (AUC) comparing no sepsis vs proven sepsis for maximum values of CRP, PCT, and WBC within 36 hours were 0.986, 0.921, and 0.360, respectively. The AUCs for CRP and PCT increased with extended time frames up to 36 hours, but there was no further difference between start to 36 hours vs start to 48 hours. Cutoff values at 16 mg/L for CRP and 2.8 ng/L for PCT provided a sensitivity of 100% for discriminating no sepsis vs proven sepsis. CONCLUSIONS: Normal serial CRP and PCT measurements within 36 hours after the start of empiric antibiotic therapy can exclude the presence of neonatal EOS with a high probability. The negative predictive values of CRP and PCT do not increase after 36 hours.

Survival and causes of death in extremely preterm infants in the Netherlands.

OBJECTIVE: In the Netherlands, the threshold for offering active treatment for spontaneous birth was lowered from 25+0 to 24+0 weeks' gestation in 2010. This study aimed to evaluate the impact of guideline implementation on survival and causes and timing of death in the years following implementation. DESIGN: National cohort study, using data from the Netherlands Perinatal Registry. PATIENTS: The study population included all 3312 stillborn and live born infants with a gestational age (GA) between 240/7 and 266/7 weeks born between January 2011 and December 2017. Infants with the same GA born between January 2007 and December 2009 (N=1400) were used as the reference group. MAIN OUTCOME MEASURES: Survival to discharge, as well as cause and timing of death. RESULTS: After guideline implementation, there was a significant increase in neonatal intensive care unit (NICU) admission rate for live born infants born at 24 weeks' GA (27%-69%, p<0.001), resulting in increased survival to discharge in 24-week live born infants (13%-34%, p<0.001). Top three causes of in-hospital mortality were necrotising enterocolitis (28%), respiratory distress syndrome (19%) and intraventricular haemorrhage (17%). A significant decrease in cause of death either complicated or caused by respiratory insufficiency was seen over time (34% in 2011-2014 to 23% in 2015-2017, p=0.006). CONCLUSIONS: Implementation of the 2010 guideline resulted as expected in increased NICU admissions rate and postnatal survival of infants born at 24 weeks' GA. In the years after implementation, a shift in cause of death was seen from respiratory insufficiency towards necrotising enterocolitis and sepsis.

Establishing the Cause of Anemia in a Premature Newborn Infant.

The three major causes of anemia in neonates are blood loss, decreased red blood cell production, and increased degradation of erythrocytes. Establishing the cause of anemia in a neonate born prematurely can be challenging. Clinically, fetomaternal hemorrhage (FMH) can be difficult to diagnose-the condition often presents only after the manifestation of severe fetal anemia. FMH can be confirmed by determining the fetal hemoglobin F fraction in the mother, which is traditionally performed using the Kleihauer-Betke test (KBT). Herein, we present a case study of a newborn baby boy of Dutch ethnicity with massive FMH and negative KBT result. The KBT result appeared to be false-negative due to AO antagonism. However, the results of an additional marker alpha-fetoprotein (AFP) test confirmed the diagnosis of massive FMH. Therefore, measuring AFP in maternal blood can be helpful in confirming FMH in unexplained anemia of the neonate.

Procalcitonin-guided decision making for duration of antibiotic therapy in neonates with suspected early-onset sepsis: a multicentre, randomised controlled trial (NeoPIns).

BACKGROUND: Up to 7% of term and late-preterm neonates in high-income countries receive antibiotics during the first 3 days of life because of suspected early-onset sepsis. The prevalence of culture-proven early-onset sepsis is 0·1% or less in high-income countries, suggesting substantial overtreatment. We assess whether procalcitonin-guided decision making for suspected early-onset sepsis can safely reduce the duration of antibiotic treatment. METHODS: We did this randomised controlled intervention trial in Dutch (n=11), Swiss (n=4), Canadian (n=2), and Czech (n=1) hospitals. Neonates of gestational age 34 weeks or older, with suspected early-onset sepsis requiring antibiotic treatment were stratified into four risk categories by their treating physicians and randomly assigned [1:1] using a computer-generated list stratified per centre to procalcitonin-guided decision making or standard care-based antibiotic treatment. Neonates who underwent surgery within the first week of life or had major congenital malformations that would have required hospital admission were excluded. Only principal investigators were masked for group assignment. Co-primary outcomes were non-inferiority for re-infection or death in the first month of life (margin 2·0%) and superiority for duration of antibiotic therapy. Intention-to-treat and per-protocol analyses were done. This trial was registered with ClinicalTrials.gov, number NCT00854932. FINDINGS: Between May 21, 2009, and Feb 14, 2015, we screened 2440 neonates with suspected early-onset sepsis. 622 infants were excluded due to lack of parental consent, 93 were ineligible for reasons unknown (68), congenital malformation (22), or surgery in the first week of life (3). 14 neonates were excluded as 100% data monitoring or retrieval was not feasible, and one neonate was excluded because their procalcitonin measurements could not be taken. 1710 neonates were enrolled and randomly assigned to either procalcitonin-guided therapy (n=866) or standard therapy (n=844). 1408 neonates underwent per-protocol analysis (745 in the procalcitonin group and 663 standard group). For the procalcitonin group, the duration of antibiotic therapy was reduced (intention to treat: 55·1 vs 65·0 h, p<0·0001; per protocol: 51·8 vs 64·0 h; p<0·0001). No sepsis-related deaths occurred, and 9 (<1%) of 1710 neonates had possible re-infection. The risk difference for non-inferiority was 0·1% (95% CI -4·6 to 4·8) in the intention-to-treat analysis (5 [0·6%] of 866 neonates in the procalcitonin group vs 4 [0·5%] of 844 neonates in the standard group) and 0·1% (-5·2 to 5·3) in the per-protocol analysis (5 [0·7%] of 745 neonates in the procalcitonin group vs 4 [0·6%] of 663 neonates in the standard group). INTERPRETATION: Procalcitonin-guided decision making was superior to standard care in reducing antibiotic therapy in neonates with suspected early-onset sepsis. Non-inferiority for re-infection or death could not be shown due to the low occurrence of re-infections and absence of study-related death. FUNDING: The Thrasher Foundation, the NutsOhra Foundation, the Sophia Foundation for Scientific research.