Recorded sessions on demand will be available 24 hours after the session ends

Abstract Body

Background and aims: Between 2012-2015, the PEPaNIC-RCT (Leuven-Rotterdam-Edmonton), including critically ill infants and children, showed that withholding parenteral nutrition early in PICU(late-PN), as compared with initiating supplemental PN early(early-PN), improved PICU outcomes and neurocognitive development assessed 2 years later. The latter was explained by altered DNA-methylation of 37 CpG-sites. As assessment of most neurocognitive domains is only possible from 4 years of age onwards, we performed a 4-year follow-up to determine the impact of late-PN versus early-PN on physical, neurocognitive, and psychosocial development.

Methods: This is a preplanned, 4-year follow-up study of the 1440 PEPaNIC patients and of 396 matched healthy children, who could be tested neurocognitively, blinded for treatment allocation(ClinicalTrials.gov-NCT01536275). Clinical outcomes included anthropometrics, health status, parent/caregiver-reported executive functions and emotional/behavioural problems, and clinical tests for intelligence, visual-motor integration, alertness, motor coordination and memory. Univariable and multivariable linear and logistic regression analyses adjusted for risk factors assessed the impact of early-PN versus late-PN on the outcomes.

Results: Overall, at 4 years follow-up, patients(356 late-PN, 328 early-PN) had worse anthropometric, health status, neurocognitive and psychosocial developmental outcomes than the healthy controls. Outcomes of late-PN patients were never worse than those of early-PN patients. In contrast, late-PN patients had fewer internalising(P=0.042), externalising(P=0.046), and total emotional and behavioural problems(P=0.007) than early-PN patients, which were normalised by late-PN.

Conclusions: Four-years after late-PN or early-PN in PICU, late-PN did not show harm, and protected against emotional and behavioural problems. These data further support de-implementation of PN-use early during critical illness in infants and children.

Abstract Body

Background: Literature on the utility of urinary Neutrophil Gelatinase-Associated Lipocalcin (uNGAL) in predicting acute kidney injury (AKI) resolution in pediatric Diabetic Ketoacidosis (DKA) is limited.

Methods: This was a secondary analysis of prospectively collected data during the SPinK trial. Sixty-six children (>1month-12years) with DKA as defined by ISPAD were included. Children with cerebral edema, chronic kidney disease and who received pre-referral fluids and/or insulin were excluded. uNGAL and urine NGAL-creatinine ratio (uNCR) at 0 and 24 hours were measured to predict renal recovery.

Results: Among 66 children enrolled, 35(53%) had AKI at admission; 29(83%) and 32(91.4%) resolved at 24 and 48 hours of admission respectively. Elevation of uNGAL at admission and 24 hours correlated well with KDIGO stages at 0 and 24 hours respectively. The uNGAL at admission and 24 hours was significantly higher in AKI group [79.8±27.2 vs 54.6±22.0,p=0.0002] and [61.4±28.3 vs. 20.2±14.5,p=0.0003]. Similar trend was observed with uNCR at admission [6.7±3.7 vs.4.1±2.6,p-0.002] and 24 hours [6.3±2.5 vs.1.2±1.0,p=0.01]. uNGAL<84.7 ng/ml and uNCR< 6.5ng/mg individually had 100% sensitivity and 77% specificity in predicting AKI resolution at 48 hours of admission. Percentage decline in uNGAL and uNCR from baseline performed best with AUC of 0.97(0.93,1.0) and 0.98(0.94,1.0) respectively.

Conclusions: uNGAL and uNCR correlated well with serum creatinine and AKI stage. Serial NGAL corrected for creatinine was better than NGAL alone as predictor for AKI resolution.

Abstract Body

Background and aims

To assess the relationship of brain development and -injury after neonatal cardiac surgery with motor outcomes in critical congenital heart disease (CHD) children.

Methods

CHD neonates underwent brain MRI after cardiac surgery between 2016-2018. Brain development was assessed by total/regional brain volumes or white matter (WM) fractional anisotropy (FA) and mean diffusivity (MD). Brain injury was categorized by WM injury (WMI, including volume) or posterior limb internal capsule (PLIC) involvement. General Movements (GMs) were assessed at 3 months, Alberta Infant Motor Scale at 3/9 months and Bayley-III-NL motor development at 3/9/18 months. Repeated cardiac surgery was included in the analysis.

Results

Fifty-one children with postoperative MRI and motor outcome at 3 (n=47), 9 (n=49) and/or 18 months (n=48) were included. Smaller regional brain volumes were associated with abnormal GMs (grey matter) and impaired total and fine motor skills at 9/18 months (all regions). FA-values of association WM were lower in children with 9 month motor score <-1SD compared to scores >-1SD (p<0.01). MD-values of WM structures were negatively correlated with total and gross motor scores at all time points. Injury with higher WMI-volume or PLIC-involvement was associated with worse composite (-0.6/-0.9SD) and gross (-0.8/-1.3SD) motor outcomes at 9 months.

Conclusions

Delayed brain development or brain injury after neonatal cardiac surgery is associated with less favorable motor outcome in CHD children. Impaired fine motor skills are strongly related to delayed volumetric brain development, with gross motor scores being affected by delayed microstructural white matter development or brain injury.

Abstract Body

Background and aims

Northern-Italy has been the first European area affected by the spread of Covid-19-pandemic. Since the beginning of Covid-19-outbreak, a 73-88% reduction in pediatric Emergency-Department admissions has been described. We sought to evaluate the impact of Covid-19-outbreak on PICU-admissions in 4 centers in Northern-Italy.

Methods

Data were extracted from the Italian-Network-of-Pediatric-Intensive-Care-Units (TIPNet) Registry. We included all patients admitted from 8-weeks-before to 8-weeks-after February 24, 2020 (lockdown administrative-decree), and those admitted in the same period in 2019. Centers included were not substantially re-organized for Covid-19-outbreak. Incidence rate ratios (IRR) and 95% Confidence-Intervals (CI) were computed using zero-inflated negative-Poisson regression modeling.

Results

Overall, 1001 PICU-admissions were included. Compared to 2019, the number of PICU-admissions significantly decreased during Covid-19-outbreak by 37% (IRR 0.629 [CI 0.502-0.789], Figure 1), especially among patients <1year (IRR 0.645 [CI 0.471-0.910]). Medical admissions significantly decreased by 50% (CI 33-60%), unplanned admissions by 40%, extra-hospital admissions by 50% (CI 35-63%). Intra-hospital, planned, and surgical admissions did not significantly change. Patients admitted for respiratory failure significantly decreased by 48% (CI 23-64%). Patients discharged with a lower-respiratory-tract diagnosis decreased by 46% (CI 6-70%). Other reasons for admissions and discharge diagnoses did not significantly differ. Patients with comorbid-conditions were admitted significantly less frequently during Covid-19-outbreak compared to 2019 (IRR 0.728 [0.557-0.956]).

Conclusions

Unplanned medical PICU-admissions, rather than elective ones, significantly decreased during Covid-19-outbreak in Northern-Italy, especially those for respiratory failure and among children <1year. A reflection on these data should be made to identify factors associated with the decrease of PICU-admissions.

Abstract Body

Background and Aims

Mortality prediction models support clinical decision making and parental discussions regarding the care of preterm infants. Of these models, >87% were developed using Logistic Regression (LR) with an arbitrary model specification and limited ability to address multiple interactions and non-linear relationships. Alternative machine learning approaches, such as Adaptive Neuro Fuzzy Inference System (ANFIS), an artificial neural network approach, can address these limitations and continually self-update with new data. We applied ANFIS to preterm infant mortality prediction.

Methods

Data from infants <32 weeks gestation admitted to 185 neonatal units in England and Wales from 2010 to 2017 were extracted using a national database. Infants were randomly divided into three cohorts, stratified by mortality and gestation: training (60%), validation (20%), and testing (20%). Prediction models for hospital mortality were derived using LR and ANFIS. Nine antenatal predictors were selected following literature review. Discrimination and calibration were assessed in the testing cohort using area under the curve (AUC) and calibration plots.

Results

63,132 infants were included with 3,588 (5.7%) excluded due to missing data. 5,727 (9.1%) infants died. Both LR and ANFIS models demonstrate similar discrimination with AUC (95% CI) of 0.84 (0.83–0.85) and 0.83 (0.82-0.85) respectively. Both models have good calibration with intercept and slope close to 0 and 1 respectively in the calibration plot (Figure 1).

Conclusions

Using limited predictors, the ANFIS derived neonatal prediction model performs similar to that of LR. By including additional predictors, ANFIS could be used to develop complex, personalised neonatal prediction tools.