Children's National

Pediatric mental health prevalence rates have increased in recent years, while gaps remain in the number of available providers. Ongoing evaluation and understanding of treatment progress and engagement are critical to psychiatric care, and these details are often documented in the electronic health record (EHR). Given the utility of retrospective chart review (RCR) as a tool for psychiatrists, we developed a coding system examining common comorbid conditions (anxiety and attention-deficit/hyperactivity disorder [ADHD]) and adherence and evaluated interrater reliability. We created a coding system with a comprehensive manual and coding instructions that explore both symptom severity domains (anxiety, ADHD, and global) and adherence to medication. Codes were rated using Likert scales, and two independent raters coded all data. RCR was completed for 142 patients with a total of 1139 visits over 2 years. Weighted linear kappa statistics ranged between 0.77 and 0.95, and weighted quadratic kappa statistics ranged between 0.74 and 0.96, suggesting substantial to almost perfect agreement. Interrater agreement was highest for anxiety severity. We created a novel coding system for RCR and found substantial to almost perfect interrater reliability for assessing ADHD severity, anxiety severity, global severity, and medication adherence using psychiatry encounter notes documented in an EHR. Our coding system explores conditions that are often heterogeneous and have waxing and waning presentations, using a continuum that captures the complexity of symptoms. Future directions include utilization of coding systems to explore emotion and behavior change over time to optimize treatment.

Abstract Background Advances in imaging have significantly enhanced the management of lymphoma, particularly through 18 F-FDG PET/CT, which combines metabolic and anatomical assessment. However, reliance on SUVmax alone provides a limited view of total disease burden. Recently, metabolic volumetric metrics such as total metabolic tumor volume (TMTV) and total lesion glycolysis (TLG) have emerged as promising tools, offering more robust prognostic information. This study aims to evaluate the predictive role of TMTV and TLG, as measured on an interim PET-CT scan (after three cycles of chemotherapy), and the change in these values on the end-of-treatment scan (after six cycles), in assessing theraputic response in patients with Hodgkin (HL) and non-Hodgkin lymphoma (NHL), and to correlate these measures with established prognostic indices. Results Ninety-one patients (HL: 38; NHL: 53) were analyzed. Comparison between interim and end-of-treatment scans showed significant reductions in TMTV and TLG in HL group, while reductions in NHL group did not reach statistical significance. For predicting complete metabolic response, ΔTLG demonstrated a diagnostic accuracy of 98.1% (at cut-off of > − 4.2) and ΔTMTV 96.2% (at cut-off of > − 1) in the NHL group, while in the HL group, ΔTLG achieved an accuracy of 92.1% (at cut-off of > − 7) and ΔMTV 89.5% (at cut-off of > − 5). In HL, TLG > 200 and MTV > 20 were associated with progressive disease, while in NHL, thresholds of TLG > 150 and MTV > 15 yielded AUCs of 0.768 and 0.761, respectively. Significant correlations were observed between end-of-treatment TMTV, TLG, and IPS in HL (r = 0.545 and 0.549; p < 0.05), but not with IPI in NHL. Conclusions TMTV and TLG are reliable PET/CT-derived biomarkers for early response prediction in lymphoma. Their dynamic changes outperform conventional SUVmax and correlate significantly with prognostic scores in HL. Incorporating these metabolic volumetric metrics may enhance individualized treatment strategies.

Predicting genetic perturbation responses at a single-cell level is central to building models for cell state and disease. However, existing approaches are limited on predicting phenotypic outcomes beyond expression changes and generalizing predictions across genome-scale perturbations in biologically relevant contexts. Here we introduce pertTF, a transformer-based single-cell genetic perturbation model. pertTF was trained from a unique dataset capturing single cell expressions profiles of 30 full gene knockouts across 14 relevant cell types during human pancreatic development and beta-cell differentiation. pertTF outperforms current methods in predicting expression changes of perturbing unseen genes in unseen cellular contexts. In addition, pertTF infers perturbation-induced shifts in cell identity and population composition, an important phenotypic outcome of perturbation in many physiology and disease settings. Through transfer learning, pertTF operates in physiologically relevant systems, including primary human islets, where large-scale perturbation experiments are challenging. The generalizability of pertTF is further demonstrated by in silico pooled and single-cell CRISPR screens, capturing critical regulators of stem cells and early pancreatic cell development. These results establish pertTF as a framework for integrating large-scale single-cell perturbation data with AI models to predict genetic perturbation effects across cellular systems and disease contexts.

Pregnant women carrying fetuses with congenital heart disease (CHD) report high psychological distress. This study examines associations between prenatal maternal distress and 18-month neurodevelopmental outcomes in toddlers with CHD. We recruited women with pregnancies complicated by fetal CHD. Maternal anxiety, stress, and depression were measured using the Spielberger State-Trait Anxiety Inventory, Perceived Stress Scale, and Edinburgh Postnatal Depression Scale. Neurodevelopment at 18 months was assessed using Bayley Scales of Infant and Toddler Development III and Infant-Toddler Social and Emotional Assessment. Postnatal parenting stress was measured using Parenting Stress Index-Short Form. We studied 123 pregnant women (21-39 gestational weeks). Half screened positive for elevated anxiety, 60% for stress, and 34% for depression. Elevated prenatal maternal psychological distress was associated with lower motor scores and increased externalizing behaviors in CHD toddlers, with stronger associations in single-ventricle than two-ventricle CHD. Elevated prenatal maternal distress was associated with increased postnatal parenting stress at 18-month testing, including higher Total Stress, Parental Distress, and Difficult Child scores. Postnatal parenting stress mediated associations between prenatal psychological distress and toddler externalizing behaviors. Elevated prenatal maternal psychological distress is associated with impaired motor and social-emotional development in toddlers with CHD and persistent postnatal parenting stress. The impact of maternal psychological distress on the long-term neurodevelopmental outcomes of toddlers with congenital heart disease (CHD) is unknown. This study examines the association between prenatal maternal distress and 18-month neurodevelopmental outcomes in toddlers with CHD. Increased prenatal maternal psychological distress is associated with impaired motor and social-emotional development in toddlers with CHD, as well as persistently elevated parenting stress. Associations between elevated prenatal maternal psychological distress and adverse motor and externalizing outcomes were stronger in toddlers with single-ventricle versus two-ventricle CHD. Identifying early risk factors for neurodevelopmental impairment in CHD toddlers is critical for developing early interventions that reduce maternal distress and improve long-term outcomes.

This paper reports on the technical performance of a nonlinear "inside-out" gradient and its feasibility and safety for diffusion encoding in humans. The tested hardware is a solenoidal coil with a 10 cm diameter, positioned at the isocenter of the magnet bore, designed to generate strong gradients within the surrounding volume. Various methods were used to characterize the spatial, temporal, mechanical, and acoustic properties of the gradient coil. ADC mapping studies show the feasibility of diffusion-weighted imaging and ADC mapping in phantoms and humans. The hardware was found to generate high gradient strengths (> 200 mT/m) for expected prostate regions whilst heating, noise, and eddy currents remained within acceptable limits for human scanning. Despite its unconventional geometry and compact form factor, standard calibration methods were effective, enabling accurate ADC mapping in both phantoms and human subjects. No peripheral nerve stimulation (PNS) was reported, even at maximum slew rates and gradient amplitudes. Inside-out nonlinear gradients can be calibrated with conventional techniques, and the hardware can provide a more convenient approach to strong diffusion weighting of targeted organs.

Introduction: Iatrogenic Withdrawal Syndrome affects over 100,000 children annually, particularly following mechanical ventilation termination. Current nursing-based screening is time-consuming and episodic, limited by staffing availability. We developed and validated a computable phenotype using standard electronic health record variables to identify withdrawal through autonomic dysregulation patterns and medication histories. Methods: This single-center, retrospective and prospective multi-cohort study included PICU patients from 2012-2025. Training/testing cohorts comprised 2,100 patients (2012-2023), with prospective validation in 400 patients (2023-2025). Inclusion criteria: extubation from mechanical ventilation and Withdrawal Assessment Tool (WAT-1) evaluation. Data were organized into 4-hour epochs. Variables included vital signs metrics (average, maximum, minimum, measurement frequency) and opioid/sedative exposures (cumulative dose, prior 4-hour dosage, average 4-hour dose). WAT-1 scores ≥3 defined withdrawal. Machine learning utilized XGBoost algorithm. Results: Test cohort performance: AUROC 0.87, AUPRC 0.62, calibration curve R2 0.96 (slope 1.08, intercept -0.02). Prospective validation maintained performance despite lower withdrawal prevalence: AUROC 0.85, AUPRC 0.35, calibration R2 0.93 (slope 1.19, intercept -0.09). Pearson’s correlation between computable phenotype risk and WAT-1 scores: r = 0.95 (p < 0.05). Prediction accuracy exceeded 87% at high/low WAT-1 values and ranged 54.8-73.3% near diagnostic thresholds. Conclusions: We successfully developed and validated a novel machine learning computable phenotype using standard EHR data for pediatric ICU withdrawal risk identification. This objective, personalized approach demonstrated clinically suitable performance metrics with potential for real-time monitoring while reducing nursing administrative burden. The model’s reliance on autonomic dysregulation indicators and medication profiles enables continuous, automated withdrawal surveillance in critically ill children.