Ventilator Weaning and Terminal Extubation: Withdrawal of Life-Sustaining Therapy in Children. Secondary Analysis of the Death One Hour After Terminal Extubation Study.

OBJECTIVE: Terminal extubation (TE) and terminal weaning (TW) during withdrawal of life-sustaining therapies (WLSTs) have been described and defined in adults. The recent Death One Hour After Terminal Extubation study aimed to validate a model developed to predict whether a child would die within 1 hour after discontinuation of mechanical ventilation for WLST. Although TW has not been described in children, pre-extubation weaning has been known to occur before WLST, though to what extent is unknown. In this preplanned secondary analysis, we aim to describe/define TE and pre-extubation weaning (PW) in children and compare characteristics of patients who had ventilatory support decreased before WLST with those who did not. DESIGN: Secondary analysis of multicenter retrospective cohort study. SETTING: Ten PICUs in the United States between 2009 and 2021. PATIENTS: Nine hundred thirteen patients 0-21 years old who died after WLST. INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: 71.4% ( n = 652) had TE without decrease in ventilatory support in the 6 hours prior. TE without decrease in ventilatory support in the 6 hours prior = 71.4% ( n = 652) of our sample. Clinically relevant decrease in ventilatory support before WLST = 11% ( n = 100), and 17.6% ( n = 161) had likely incidental decrease in ventilatory support before WLST. Relevant ventilator parameters decreased were F io2 and/or ventilator set rates. There were no significant differences in any of the other evaluated patient characteristics between groups (weight, body mass index, unit type, primary diagnostic category, presence of coma, time to death after WLST, analgosedative requirements, postextubation respiratory support modality). CONCLUSIONS: Decreasing ventilatory support before WLST with extubation in children does occur. This practice was not associated with significant differences in palliative analgosedation doses or time to death after extubation.

Prehospitalization Trauma and Physiologic Factors Associated with the Presence of Post-traumatic Stress 3 Months After PICU Discharge.

BACKGROUND: Children admitted to the pediatric intensive care unit (PICU) have post-traumatic stress (PTS) rates up to 64%, and up to 28% of them meet criteria for PTS disorder (PTSD). We aim to examine whether a prior trauma history and increased physiologic parameters due to a heightened sympathetic response are associated with later PTS. Our hypothesis was children with history of prehospitalization trauma, higher heart rates, blood pressures, cortisol, and extrinsic catecholamine administration during PICU admission are more likely to have PTS after discharge. METHODS: This is a prospective, observational study of children admitted to the PICU at an urban, quaternary, academic children's hospital. Children aged 8 to 17 years old without developmental delay, severe psychiatric disorder, or traumatic brain injury were included. Children's prehospitalization trauma history was assessed with a semistructured interview. All in-hospital variables were from the electronic medical record. PTS was present if children had 4 of the Diagnostic and Statistical Manual of Mental Disorders IV criteria for PTSD. Student's t- and chi-squared tests were used to compare the presence or absence of prior trauma and all of the PICU-associated variables. RESULTS: Of the 110 children at baseline, 67 had 3-month follow-up. In the latter group, 46% met the criteria for PTS, mean age of 13 years (SD 3), 57% male, a mean PRISM III score of 4.9 (SD 4.3), and intensive care unit length of stay 6.5 days (SD 7.8). There were no statistically significant differences in the demographics of the children with and without PTS. The only variable to show significance was trauma history; children with prehospitalization trauma were more likely to have PTS at 3-month follow-up (P = .02). CONCLUSIONS: Prehospitalization trauma history was associated with the presence of PTS after admission to the PICU. This study suggests future studies should shift to the potential predictive benefit of screening children for trauma history upon PICU admission.

A Novel Application of Spectrograms with Machine Learning Can Detect Patient Ventilator Dyssynchrony.

Patients in intensive care units are frequently supported by mechanical ventilation. There is increasing awareness of patient-ventilator dyssynchrony (PVD), a mismatch between patient respiratory effort and assistance provided by the ventilator, as a risk factor for infection, narcotic exposure, lung injury, and adverse neurocognitive effects. One of the most injurious consequences of PVD are double cycled (DC) breaths when two breaths are delivered by the ventilator instead of one. Prior efforts to identify PVD have limited efficacy. An automated method to identify PVD, independent of clinician expertise, acumen, or time, would potentially permit early, targeted treatment to avoid further harm. We performed secondary analyses of data from a clinical trial of children with acute respiratory distress syndrome. Waveforms of ventilator flow, airway pressure and esophageal manometry were annotated to identify DC breaths and underlying PVD subtypes. Spectrograms were generated from those waveforms to train Convolutional Neural Network (CNN) models in detecting DC and underlying PVD subtypes: Reverse Trigger (RT) and Inadequate Support (IS). The DC breath detection model yielded AUROC of 0.980, while the multi-target detection model for underlying dyssynchrony yielded AUROC of 0.980 (RT) and 0.976 (IS). When operating at 75% sensitivity, DC breath detection had a number needed to alert (NNA) 1.3 (99% specificity), while underlying PVD had a NNA 1.6 (98.5% specificity) for RT and NNA 4.0 (98.2% specificity) for IS. CNNs using spectrograms of ventilator waveforms can identify DC breaths and detect the underlying PVD for targeted clinical interventions.

Narrowing the gap: expected versus deployment performance.

OBJECTIVES: Successful model development requires both an accurate a priori understanding of future performance and high performance on deployment. Optimistic estimations of model performance that are unrealized in real-world clinical settings can contribute to nonuse of predictive models. This study used 2 tasks, predicting ICU mortality and Bi-Level Positive Airway Pressure failure, to quantify: (1) how well internal test performances derived from different methods of partitioning data into development and test sets estimate future deployment performance of Recurrent Neural Network models and (2) the effects of including older data in the training set on models' performance. MATERIALS AND METHODS: The cohort consisted of patients admitted between 2010 and 2020 to the Pediatric Intensive Care Unit of a large quaternary children's hospital. 2010-2018 data were partitioned into different development and test sets to measure internal test performance. Deployable models were trained on 2010-2018 data and assessed on 2019-2020 data, which was conceptualized to represent a real-world deployment scenario. Optimism, defined as the overestimation of the deployed performance by internal test performance, was measured. Performances of deployable models were also compared with each other to quantify the effect of including older data during training. RESULTS, DISCUSSION, AND CONCLUSION: Longitudinal partitioning methods, where models are tested on newer data than the development set, yielded the least optimism. Including older years in the training dataset did not degrade deployable model performance. Using all available data for model development fully leveraged longitudinal partitioning by measuring year-to-year performance.

Development of a deep learning model that predicts Bi-level positive airway pressure failure.

Delaying intubation for patients failing Bi-Level Positive Airway Pressure (BIPAP) may be associated with harm. The objective of this study was to develop a deep learning model capable of aiding clinical decision making by predicting Bi-Level Positive Airway Pressure (BIPAP) failure. This was a retrospective cohort study in a tertiary pediatric intensive care unit (PICU) between 2010 and 2020. Three machine learning models were developed to predict BIPAP failure: two logistic regression models and one deep learning model, a recurrent neural network with a Long Short-Term Memory (LSTM-RNN) architecture. Model performance was evaluated in a holdout test set. 175 (27.7%) of 630 total BIPAP sessions were BIPAP failures. Patients in the BIPAP failure group were on BIPAP for a median of 32.8 (9.2-91.3) hours prior to intubation. Late BIPAP failure (intubation after using BIPAP > 24 h) patients had fewer 28-day Ventilator Free Days (13.40 [0.68-20.96]), longer ICU length of stay and more post-extubation BIPAP days compared to those who were intubated ≤ 24 h from BIPAP initiation. An AUROC above 0.5 indicates that a model has extracted new information, potentially valuable to the clinical team, about BIPAP failure. Within 6 h of BIPAP initiation, the LSTM-RNN model predicted which patients were likely to fail BIPAP with an AUROC of 0.81 (0.80, 0.82), superior to all other models. Within 6 h of BIPAP initiation, the LSTM-RNN model would identify nearly 80% of BIPAP failures with a 50% false alarm rate, equal to an NNA of 2. In conclusion, a deep learning method using readily available data from the electronic health record can identify which patients on BIPAP are likely to fail with good discrimination, oftentimes days before they are intubated in usual practice.

Repertoires of SARS-CoV-2 epitopes targeted by antibodies vary according to severity of COVID-19.

Antibodies to SARS-CoV-2 are central to recovery and immunity from COVID-19. However, the relationship between disease severity and the repertoire of antibodies against specific SARS-CoV-2 epitopes an individual develops following exposure remains incompletely understood. Here, we studied seroprevalence of antibodies to specific SARS-CoV-2 and other betacoronavirus antigens in a well-annotated, community sample of convalescent and never-infected individuals obtained in August 2020. One hundred and twenty-four participants were classified into five groups: previously exposed but without evidence of infection, having no known exposure or evidence of infection, seroconverted without symptoms, previously diagnosed with symptomatic COVID-19, and recovered after hospitalization with COVID-19. Prevalence of IgGs specific to the following antigens was compared between the five groups: recombinant SARS-CoV-2 and betacoronavirus spike and nucleocapsid protein domains, peptides from a tiled array of 22-mers corresponding to the entire spike and nucleocapsid proteins, and peptides corresponding to predicted immunogenic regions from other proteins of SARS-CoV-2. Antibody abundance generally correlated positively with severity of prior illness. A number of specific immunogenic peptides and some that may be associated with milder illness or protection from symptomatic infection were identified. No convincing association was observed between antibodies to Receptor Binding Domain(s) (RBDs) of less pathogenic betacoronaviruses HKU1 or OC43 and COVID-19 severity. However, apparent cross-reaction with SARS-CoV RBD was evident and some predominantly asymptomatic individuals had antibodies to both MERS-CoV and SARS-CoV RBDs. Findings from this pilot study may inform development of diagnostics, vaccines, and therapeutic antibodies, and provide insight into viral pathogenic mechanisms.

Predicting High Flow Nasal Cannula Failure in an Intensive Care Unit Using a Recurrent Neural Network With Transfer Learning and Input Data Perseveration: Retrospective Analysis.

BACKGROUND: High flow nasal cannula (HFNC) provides noninvasive respiratory support for children who are critically ill who may tolerate it more readily than other noninvasive ventilation (NIV) techniques such as bilevel positive airway pressure and continuous positive airway pressure. Moreover, HFNC may preclude the need for mechanical ventilation (intubation). Nevertheless, NIV or intubation may ultimately be necessary for certain patients. Timely prediction of HFNC failure can provide an indication for increasing respiratory support. OBJECTIVE: The aim of this study is to develop and compare machine learning (ML) models to predict HFNC failure. METHODS: A retrospective study was conducted using the Virtual Pediatric Intensive Care Unit database of electronic medical records of patients admitted to a tertiary pediatric intensive care unit between January 2010 and February 2020. Patients aged <19 years, without apnea, and receiving HFNC treatment were included. A long short-term memory (LSTM) model using 517 variables (vital signs, laboratory data, and other clinical parameters) was trained to generate a continuous prediction of HFNC failure, defined as escalation to NIV or intubation within 24 hours of HFNC initiation. For comparison, 7 other models were trained: a logistic regression (LR) using the same 517 variables, another LR using only 14 variables, and 5 additional LSTM-based models using the same 517 variables as the first LSTM model and incorporating additional ML techniques (transfer learning, input perseveration, and ensembling). Performance was assessed using the area under the receiver operating characteristic (AUROC) curve at various times following HFNC initiation. The sensitivity, specificity, and positive and negative predictive values of predictions at 2 hours after HFNC initiation were also evaluated. These metrics were also computed for a cohort with primarily respiratory diagnoses. RESULTS: A total of 834 HFNC trials (455 [54.6%] training, 173 [20.7%] validation, and 206 [24.7%] test) met the inclusion criteria, of which 175 (21%; training: 103/455, 22.6%; validation: 30/173, 17.3%; test: 42/206, 20.4%) escalated to NIV or intubation. The LSTM models trained with transfer learning generally performed better than the LR models, with the best LSTM model achieving an AUROC of 0.78 versus 0.66 for the 14-variable LR and 0.71 for the 517-variable LR 2 hours after initiation. All models except for the 14-variable LR achieved higher AUROCs in the respiratory cohort than in the general intensive care unit population. CONCLUSIONS: ML models trained using electronic medical record data were able to identify children at risk of HFNC failure within 24 hours of initiation. LSTM models that incorporated transfer learning, input data perseveration, and ensembling showed improved performance compared with the LR and standard LSTM models.

Continuous Prediction of Mortality in the PICU: A Recurrent Neural Network Model in a Single-Center Dataset.

OBJECTIVES: Develop, as a proof of concept, a recurrent neural network model using electronic medical records data capable of continuously assessing an individual child's risk of mortality throughout their ICU stay as a proxy measure of severity of illness. DESIGN: Retrospective cohort study. SETTING: PICU in a tertiary care academic children's hospital. PATIENTS/SUBJECTS: Twelve thousand five hundred sixteen episodes (9,070 children) admitted to the PICU between January 2010 and February 2019, partitioned into training (50%), validation (25%), and test (25%) sets. INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: On 2,475 test set episodes lasting greater than or equal to 24 hours in the PICU, the area under the receiver operating characteristic curve of the recurrent neural network's 12th hour predictions was 0.94 (CI, 0.93-0.95), higher than those of Pediatric Index of Mortality 2 (0.88; CI, [0.85-0.91]; p < 0.02), Pediatric Risk of Mortality III (12th hr) (0.89; CI, [0.86-0.92]; p < 0.05), and Pediatric Logistic Organ Dysfunction day 1 (0.85; [0.81-0.89]; p < 0.002). The recurrent neural network's discrimination increased with more acquired data and smaller lead time, achieving a 0.99 area under the receiver operating characteristic curve 24 hours prior to discharge. Despite not having diagnostic information, the recurrent neural network performed well across different primary diagnostic categories, generally achieving higher area under the receiver operating characteristic curve for these groups than the other three scores. On 692 test set episodes lasting greater than or equal to 5 days in the PICU, the recurrent neural network area under the receiver operating characteristic curves significantly outperformed their daily Pediatric Logistic Organ Dysfunction counterparts (p < 0.005). CONCLUSIONS: The recurrent neural network model can process hundreds of input variables contained in a patient's electronic medical record and integrate them dynamically as measurements become available. Its high discrimination suggests the recurrent neural network's potential to provide an accurate, continuous, and real-time assessment of a child in the ICU.

Interpreting a recurrent neural network's predictions of ICU mortality risk.

Deep learning has demonstrated success in many applications; however, their use in healthcare has been limited due to the lack of transparency into how they generate predictions. Algorithms such as Recurrent Neural Networks (RNNs) when applied to Electronic Medical Records (EMR) introduce additional barriers to transparency because of the sequential processing of the RNN and the multi-modal nature of EMR data. This work seeks to improve transparency by: 1) introducing Learned Binary Masks (LBM) as a method for identifying which EMR variables contributed to an RNN model's risk of mortality (ROM) predictions for critically ill children; and 2) applying KernelSHAP for the same purpose. Given an individual patient, LBM and KernelSHAP both generate an attribution matrix that shows the contribution of each input feature to the RNN's sequence of predictions for that patient. Attribution matrices can be aggregated in many ways to facilitate different levels of analysis of the RNN model and its predictions. Presented are three methods of aggregations and analyses: 1) over volatile time periods within individual patient predictions, 2) over populations of ICU patients sharing specific diagnoses, and 3) across the general population of critically ill children.

Machine Learning to Predict Cardiac Death Within 1 Hour After Terminal Extubation.

OBJECTIVES: Accurate prediction of time to death after withdrawal of life-sustaining therapies may improve counseling for families and help identify candidates for organ donation after cardiac death. The study objectives were to: 1) train a long short-term memory model to predict cardiac death within 1 hour after terminal extubation, 2) calculate the positive predictive value of the model and the number needed to alert among potential organ donors, and 3) examine associations between time to cardiac death and the patient's characteristics and physiologic variables using Cox regression. DESIGN: Retrospective cohort study. SETTING: PICU and cardiothoracic ICU in a tertiary-care academic children's hospital. PATIENTS: Patients 0-21 years old who died after terminal extubation from 2011 to 2018 (n = 237). INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: The median time to death for the cohort was 0.3 hours after terminal extubation (interquartile range, 0.16-1.6 hr); 70% of patients died within 1 hour. The long short-term memory model had an area under the receiver operating characteristic curve of 0.85 and a positive predictive value of 0.81 at a sensitivity of 94% when predicting death within 1 hour of terminal extubation. About 39% of patients who died within 1 hour met organ procurement and transplantation network criteria for liver and kidney donors. The long short-term memory identified 93% of potential organ donors with a number needed to alert of 1.08, meaning that 13 of 14 prepared operating rooms would have yielded a viable organ. A Cox proportional hazard model identified independent predictors of shorter time to death including low Glasgow Coma Score, high Pao2-to-Fio2 ratio, low-pulse oximetry, and low serum bicarbonate. CONCLUSIONS: Our long short-term memory model accurately predicted whether a child will die within 1 hour of terminal extubation and may improve counseling for families. Our model can identify potential candidates for donation after cardiac death while minimizing unnecessarily prepared operating rooms.

The Impact of Coronavirus Disease 2019 Pandemic on U.S. and Canadian PICUs.

OBJECTIVES: There are limited reports of the impact of the coronavirus disease 2019 pandemic focused on U.S. and Canadian PICUs. This hypothesis-generating report aims to identify the United States and Canadian trends of coronavirus disease 2019 in PICUs. DESIGN AND SETTING: To better understand how the coronavirus disease 2019 pandemic was affecting U.S. and Canadian PICUs, an open voluntary daily data collection process of Canadian and U.S. PICUs was initiated by Virtual Pediatric Systems, LLC (Los Angeles, CA; http://www.myvps.org) in mid-March 2020. Information was made available online to all PICUs wishing to participate. A secondary data collection was performed to follow-up on patients discharged from those PICUs reporting coronavirus disease 2019 positive patients. MEASUREMENTS AND MAIN RESULTS: To date, over 180 PICUs have responded detailing 530 PICU admissions requiring over 3,467 days of PICU care with 30 deaths. The preponderance of cases was in the eastern regions. Twenty-four percent of the patients admitted to the PICUs were over 18 years old. Fourteen percent of admissions were under 2 years old. Nearly 60% of children had comorbidities at admission with the average length of stay increasing by age and by severity of comorbidity. Advanced respiratory support was necessary during 67% of the current days of care, with 69% being conventional mechanical ventilation. CONCLUSIONS: PICUs have been significantly impacted by the pandemic. They have provided care not only for children but also adults. Patients with coronavirus disease 2019 have a high frequency of comorbidities, require longer stays, more ventilatory support than usual PICU admissions. These data suggest several avenues for further exploration.

The impact of extraneous features on the performance of recurrent neural network models in clinical tasks.

Electronic Medical Records (EMR) are a rich source of patient information, including measurements reflecting physiologic signs and administered therapies. Identifying which variables or features are useful in predicting clinical outcomes can be challenging. Advanced algorithms, such as deep neural networks, were designed to process high-dimensional inputs containing variables in their measured form, thus bypass separate feature selection or engineering steps. We investigated the effect of extraneous input features on the predictive performance of Recurrent Neural Networks (RNN) by including in the input vector extraneous features that were randomly drawn from theoretical and empirical distributions. RNN models using different input vectors (EMR features only; EMR and extraneous features; extraneous features only) were trained to predict three clinical outcomes: in-ICU mortality, 72-h ICU re-admission, and 30-day ICU-free days. The measured degradations of the RNN's predictive performance with the inclusion of extraneous features to EMR variables were negligible.

Predicting individual physiologically acceptable states at discharge from a pediatric intensive care unit.

Objective: Quantify physiologically acceptable PICU-discharge vital signs and develop machine learning models to predict these values for individual patients throughout their PICU episode. Methods: EMR data from 7256 survivor PICU episodes (5632 patients) collected between 2009 and 2017 at Children's Hospital Los Angeles was analyzed. Each episode contained 375 variables representing physiology, labs, interventions, and drugs. Between medical and physical discharge, when clinicians determined the patient was ready for ICU discharge, they were assumed to be in a physiologically acceptable state space (PASS) for discharge. Each patient's heart rate, systolic blood pressure, diastolic blood pressure in the PASS window were measured and compared to age-normal values, regression-quantified PASS predictions, and recurrent neural network (RNN) PASS predictions made 12 hours after PICU admission. Results: Mean absolute errors (MAEs) between individual PASS values and age-normal values (HR: 21.0 bpm; SBP: 10.8 mm Hg; DBP: 10.6 mm Hg) were greater (p < .05) than regression prediction MAEs (HR: 15.4 bpm; SBP: 9.9 mm Hg; DBP: 8.6 mm Hg). The RNN models best approximated individual PASS values (HR: 12.3 bpm; SBP: 7.6 mm Hg; DBP: 7.0 mm Hg). Conclusions: The RNN model predictions better approximate patient-specific PASS values than regression and age-normal values.

Low-dose dexmedetomidine as an adjuvant to propofol infusion for children in MRI: A double-cohort study.

INTRODUCTION: Propofol is an effective sedative for magnetic resonance imaging. Nevertheless, it may cause hemodynamic and respiratory complications in a dose dependent fashion. We investigated the role of low-dose dexmedetomidine (0.5 µg/kg) as an adjuvant to propofol sedation for children undergoing magnetic resonance imaging. We hypothesized that dexmedetomidine would decrease the propofol dose required, airway complications, and hemodynamic instability. METHODS: We performed a retrospective chart review of patients' age of 1 month to 20 years. Children were divided into 2 groups; group P received only propofol; group D + P received intravenous bolus of dexmedetomidine (0.5 µg/kg) and propofol. RESULTS: We reviewed 172 children in P and 129 in D + P (dexmedetomidine dose, median: 0.50 µg/kg (IQR: 0.45-0.62). An additional dexmedetomidine bolus was given to 17 children for sedation lasting longer than 2 hours. Total propofol dose (µg/kg/min) was significantly higher in group P than D + P; 215.0 (182.6-253.8) vs 147.6 (127.5-180.9); Median Diff = -67.8; 95%CI = -80.6, -54.9; P < .0001. There was no difference in recovery time (minutes); P: 28 (17-39) vs D + P: 27 (18-41); Median Diff = -1; 95%CI = -6.0, 4.0; P = .694. The need for airway support was significantly greater in P compared to D + P; 15/172 vs 3/129; OR = 0.25; 95%CI = 0.07 to 0.90; P = .02 (2-sample proportions test). Mean arterial pressure was significantly lower in P compared to D + P across time over 60 minutes after induction (coef = -0.06, 95%CI = -0.11, -0.02, P = .004). DISCUSSION & CONCLUSION: A low-dose bolus of dexmedetomidine (0.5 µg/kg) used as an adjuvant can decrease the propofol requirement for children undergoing sedation for magnetic resonance imaging. This may decrease the need for airway support and contribute to improved hemodynamic stability without prolonging recovery time.

Applying Machine Learning to Pediatric Critical Care Data.

OBJECTIVES: To explore whether machine learning applied to pediatric critical care data could discover medically pertinent information, we analyzed clinically collected electronic medical record data, after data extraction and preparation, using k-means clustering. DESIGN: Retrospective analysis of electronic medical record ICU data. SETTING: Tertiary Children's Hospital PICU. PATIENTS: Anonymized electronic medical record data from PICU admissions over 10 years. INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: Data from 11,384 PICU episodes were cleaned, and specific features were generated. A k-means clustering algorithm was applied, and the stability and medical validity of the resulting 10 clusters were determined. The distribution of mortality, length of stay, use of ventilation and pressors, and diagnostic categories among resulting clusters was analyzed. Clusters had significant prognostic information (p < 0.0001). Cluster membership predicted mortality (area under the curve of the receiver operating characteristic = 0.77). Length of stay, the use of inotropes and intubation, and diagnostic categories were nonrandomly distributed among the clusters (p < 0.0001). CONCLUSIONS: A standard machine learning methodology was able to determine significant medically relevant information from PICU electronic medical record data which included prognosis, diagnosis, and therapy in an unsupervised approach. Further development and application of machine learning to critical care data may provide insights into how critical illness happens to children.

Development and Validation of an Empiric Tool to Predict Favorable Neurologic Outcomes Among PICU Patients.

OBJECTIVES: To create a novel tool to predict favorable neurologic outcomes during ICU stay among children with critical illness. DESIGN: Logistic regression models using adaptive lasso methodology were used to identify independent factors associated with favorable neurologic outcomes. A mixed effects logistic regression model was used to create the final prediction model including all predictors selected from the lasso model. Model validation was performed using a 10-fold internal cross-validation approach. SETTING: Virtual Pediatric Systems (VPS, LLC, Los Angeles, CA) database. PATIENTS: Patients less than 18 years old admitted to one of the participating ICUs in the Virtual Pediatric Systems database were included (2009-2015). INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: A total of 160,570 patients from 90 hospitals qualified for inclusion. Of these, 1,675 patients (1.04%) were associated with a decline in Pediatric Cerebral Performance Category scale by at least 2 between ICU admission and ICU discharge (unfavorable neurologic outcome). The independent factors associated with unfavorable neurologic outcome included higher weight at ICU admission, higher Pediatric Index of Morality-2 score at ICU admission, cardiac arrest, stroke, seizures, head/nonhead trauma, use of conventional mechanical ventilation and high-frequency oscillatory ventilation, prolonged hospital length of ICU stay, and prolonged use of mechanical ventilation. The presence of chromosomal anomaly, cardiac surgery, and utilization of nitric oxide were associated with favorable neurologic outcome. The final online prediction tool can be accessed at https://soipredictiontool.shinyapps.io/GNOScore/. Our model predicted 139,688 patients with favorable neurologic outcomes in an internal validation sample when the observed number of patients with favorable neurologic outcomes was among 139,591 patients. The area under the receiver operating curve for the validation model was 0.90. CONCLUSIONS: This proposed prediction tool encompasses 20 risk factors into one probability to predict favorable neurologic outcome during ICU stay among children with critical illness. Future studies should seek external validation and improved discrimination of this prediction tool.