Structural equation modeling of findings in pediatric tele-neuroradiology: a 2-year nationwide study of turnaround time predictors and clinical decision support.

backgroundPediatric neuroradiology faces significant workforce shortages, with teleradiology emerging as a vital solution. However, studies investigating findings and their operational impact from teleradiology centers remain limited.

objectivesTo develop and validate structural equation models for identifying predictors of turnaround time and formulate recommendations for workflow optimization in pediatric tele-neuroradiology services.

designA retrospective cohort study following STROBE guidelines. SETTINGS: 107 hospitals across 17 states in the United States (US) via a teleradiology platform providing interpretation services through US board-certified radiologists. PATIENTS AND

methodsWe analyzed 9985 pediatric neuroradiology scans from 7958 patients (January 2023-December 2024). We utilized confirmatory factor analysis to validate findings structures, followed by structural equation modeling to predict turnaround times. Binary logistic regression models were developed with area-under-the-curve (AUC) estimation for performance assessment. Bootstrap validation with 5000 samples ensured model stability.

main outcome measuresPrimary outcome was turnaround time. Secondary outcomes included requirements for multiple imaging studies, follow-up recommendations, and consultations. SAMPLE SIZE: A total of 9985 studies providing over 99% statistical power for detecting significant relationships.

resultsFactor analysis demonstrated a two-factor structure (trauma: α = 0.742, structural: α = 0.685). The structural model explained 7.8% of turnaround time variance, with computed tomography (CT) modality (β = -0.164), trauma score (β = 0.125), and structural score (β = 0.142) as significant predictors. Among immediate neurosurgical emergencies (n = 180, 1.8%), 89.4% achieved turnaround time within the 60-min benchmark for time-sensitive consultations. Prediction models demonstrated excellent discrimination: traumatic findings (AUC = 0.91), structural findings (AUC = 0.92), critical findings (AUC = 0.95), and a dedicated neurosurgical emergency model (AUC = 0.94, NPV = 0.996). A severity classification system showed strong validation against imaging needs (AUC = 0.76) and consultations (AUC = 0.89).

conclusionsOur study establishes a validated SEM framework for pediatric tele-neuroradiology with excellent predictive performance (AUC = 0.91-0.95). Among immediate neurosurgical emergencies (n = 180, 1.8%), 89.4% met the 60-min benchmark, and a dedicated emergency prediction model achieved AUC = 0.94. However, translation to improved neurosurgical care delivery and patient outcomes remains unvalidated, representing the next investigational priorities. LIMITATIONS: Retrospective design limits causal inference; a single platform may limit generalizability; the CT majority (96.8%) limits magnetic resonance imaging conclusions. Critically, post-diagnostic clinical outcomes, including neurosurgical consultations, interventions performed, and patient outcomes, were not tracked, precluding conclusions about whether documented operational efficiency translated to improved neurosurgical care delivery.