Abstract
Objective: In Australia, over 180,000 mild traumatic brain injuries (mTBI) occur annually. ∼20-50% of people experience symptoms that persist for months-to-years after injury. mTBI management is variable, with limited capacity to predict delayed recovery. AUS-mTBI is using app-based technology to identify key predictors of mTBI recovery in the Australian population with the objective of delivering improved care advice.
Methods: AUS-mTBI participants aged five and over who sustained mTBI in the previous 14 days, self- or proxy-report demographic, injury, health status, symptom burden, and care management data through the HeadCheck app. Post-concussion symptoms and recovery are assessed using the Rivermead Post-Concussion Questionnaire (adults) and Post-Concussion Symptom Inventory (5–18y), administered at day 1,3,5,7,14,30 and then fortnightly for three months post-injury and monthly thereafter until symptom resolution. Data analysis uses machine learning, including Random Survival Forest models for adults and children to identify recovery predictors.
Results: AUS-TBI has recruited 562 participants to-date (55.4% children, 44.6% adults). 01069340051.2% identify as male, 52.0% sustained mTBI from non-sporting mechanisms, most commonly falls and 60.3% attended hospital post-mTBI. Preliminary models suggest that the strongest predictors of recovery in adults are symptom severity, injury mechanism, history of previous mTBI and healthcare interactions (e.g. hospital care, follow up treatment)(C-index 0.78±0.05). For children, the strongest predictors of recovery are symptom severity, age, socioeconomic status, peri-injury neck pain, resilience and injury mechanism (C-index 0.72±0.06).
Conclusions: Predicting delayed recovery after mTBI is challenging. Using machine learning to identify modifiable factors, AUS-mTBI may help shape future directions of HeadCheck and inform data-driven care pathways.