Time-Domain HRV Metrics as Predictors of Concussion Recovery in Adolescents: A Boosted Tree Approach

Area of Science:

• Neurology
• Biomedical Engineering
• Sports Medicine

Background:

• Adolescent concussions pose significant public health challenges due to prolonged recovery and diagnostic difficulties.
• Current clinical assessments may not fully capture the severity or recovery trajectory of concussions in young individuals.
• Objective physiological biomarkers are needed to improve concussion assessment and management.

Purpose of the Study:

• To evaluate heart rate variability (HRV) as a potential objective biomarker for monitoring concussion recovery in adolescents.
• To compare traditional statistical analyses with machine learning approaches for concussion outcome prediction using HRV.

Main Methods:

• Thirty-seven concussed adolescents and 37 healthy controls underwent 5-minute electrocardiogram recordings and completed the Post-Concussion Symptom Scale (PCSS).
• Time-domain HRV metrics (SDNN, RMSSD) were calculated and severity-adjusted (SDNNidx, RMSSDidx).
• A Boosted Tree machine learning algorithm was employed to predict clinical recovery outcomes based on HRV features.

Main Results:

• Standard HRV analyses did not reveal significant group differences.
• Severity-adjusted HRV indices correlated with reduced symptom severity.
• The machine learning model demonstrated strong predictive performance (AUC=.88), with high sensitivity and specificity in classifying recovery outcomes.

Conclusions:

• Machine learning effectively identified nonlinear HRV patterns indicative of clinical recovery, surpassing traditional statistical methods.
• HRV-based predictive modeling offers a noninvasive approach for personalized autonomic monitoring in adolescents.
• This approach supports evidence-based, individualized concussion management strategies.