Development and External Validation of a Clinical Decision Support Tool for Managing Children with Minor Head Trauma and Intracranial Injuries

Abstract

Background: There are no validated tools guiding the appropriate level-of-care for children with minor head trauma (MHT) and radiographic intracranial injuries. We developed and externally validated a clinical decision support tool for risk-stratifying these patients. Methods: The derivation cohort included children enrolled in the PECARN head injury study from 2004-2006. The validation group consisted of a separate cohort of children with MHT treated at six centers from 2006-2019. Children < 18 years were included if they presented within 24 hours of injury, had a Glasgow Coma Scales (GCS) score of 13-15, and had intracranial injury on neuroimaging. A risk model was developed using recursive partitioning to predict serious neurological events, defined as neurosurgical intervention, intubation longer than 24 hours due to head injury, or death from head injury. Risk thresholds were determined for clinical decision-making. This model was compared to a logistic regression model previously developed using the PECARN cohort. Findings: Based on 839 patients meeting inclusion criteria in the PECARN cohort, the tree-based model included seven predictors, with midline shift, depressed skull fracture, epidural hematoma, and GCS score being the most influential. Three risk thresholds were evaluated to classify patients as ‘low’ risk. In the validation cohort consisting of 1,630 patients, the most conservative cutoff (i.e. any predictor present) identified 119/119 children with serious neurological events (sensitivity 100%; 95% CI 96·9-100), but had the lowest specificity (26.3%; 95% CI 24·1-28·6). All three thresholds had negative predictive values above 99% and collectively outperformed the logistic regression model. Using these thresholds, a clinical decision support tool was created that identified 75% of the population as ‘low’ or ‘very low’ risk. Interpretation: We developed and validated a clinical decision support tool that accurately risk-stratifies children with MHT and radiographic intracranial injuries. This tool can potentially improve patient safety and reduce resource use. Funding : Thrasher Research Fund and the Agency for Healthcare Research and Quality. Funding Statement: This study was funded by grants from the Agency for Healthcare Research and Quality (1F32HS027075-01A1) and the Thrasher Research Fund (#15024). Declaration of Interests: MAO reports grant funding from Merck, Sanofi Pasteur, and Pfizer, and consulting fees from Pfizer. EJ reports medical expert testimony and patent holdings (US Patent 9993631). DDL has received research funds and/or research equipment for unrelated projects from Medtronic, Inc., Karl Storz, Inc., and Microbot Medical, Inc, and has received philanthropic equipment contributions for humanitarian relief work from Karl Storz, Inc. and Aesculap, Inc. No other authors disclose any competing interests. Ethics Approval Statement: All data analyzed were from the deidentified, public-use dataset, and consequently the Washington University in St. Louis Human Research Protection Office deemed the analysis not subject to institutional review board (IRB) oversight.