External Validation of a Prognostic Model for Outcome After Mild Traumatic Brain Injury at 6 Months Post Injury

Authors:
Open Access
Publisher website
Google Scholar
Abstract
Background: While most people fully recover after mild traumatic brain injury (mTBI), a substantial minority experience persistent symptoms and incomplete recovery. This may be prevented by early interventions specifically targeting patients at risk of incomplete recovery. The UPFRONT‐model was developed to identify patients at risk of poor functional outcome. This study aimed to externally validate the UPFRONT‐model in an independent sample. Methods: A prospective, longitudinal, multicenter cohort study with 126 mTBI patients recruited from emergency and neurology departments (ED) of six hospitals in the Netherlands was performed. Predictors in the UPFRONT‐model included educational level, Glasgow Coma Scale (GCS), neck pain at injury, alcohol intoxication, Post Traumatic Amnesia (PTA), pre‐injury mental health, anxiety, depression, coping, and the number and severity of post‐traumatic complaints. Functional recovery at 6 months was assessed using the Glasgow Outcome Scale Extended (GOS‐E). The external validity of the UPRFONT model was assessed with measures of calibration and discrimination. Results: The model showed acceptable discriminative ability (AUC = 0.74), comparable to the development sample (AUC = 0.77). However, calibration revealed systematic underestimation of recovery (predicted: 46%; observed: 75%), with a calibration intercept of 1.52 and a slope of 0.70. Despite differences in some predictor effects, psychological variables were robust and consistent across samples. Conclusion: The UPFRONT‐model demonstrated solid discriminative performance in an external cohort, but tended to underestimate the likelihood of complete recovery. Further validation and optimization are needed before clinical implementation. The model holds promise for early identification of at‐risk patients, enabling targeted interventions following mTBI.

This publication has 32 references indexed in Scilit: