Identifying and predicting headache trajectories among those with acute post-traumatic headache.
Lingchao Mao1, Jing Li1, Todd J Schwedt2,3
1School of Industrial and Systems Engineering, Georgia Tech, Atlanta, Georgia, USA.
Headache
|May 30, 2025
Summary
This study identified distinct headache improvement patterns in patients with post-traumatic headache (PTH) after mild traumatic brain injury (mTBI). Machine learning models accurately predict these trajectories early, aiding personalized treatment and clinical trial eligibility.
Area of Science:
- Neurology
- Data Science
- Medical Informatics
Background:
- Post-traumatic headache (PTH) is a frequent consequence of mild traumatic brain injury (mTBI), lacking predictable recovery timelines.
- Current methods struggle to forecast individual PTH symptom evolution, necessitating novel predictive approaches.
Purpose of the Study:
- To identify distinct patient subgroups based on headache symptom trajectories following mTBI.
- To develop machine learning (ML) models for early prediction of these individual headache evolution patterns.
Main Methods:
- Utilized tensor decomposition and clustering on daily electronic headache diary data from 73 individuals with acute PTH over 3 months.
- Developed an ML model to classify individuals into identified subgroups using early symptom data.
Main Results:
- Identified four distinct PTH trajectory subgroups: no improvement, mild improvement, substantial improvement, and mild improvement.
- The ML model achieved high accuracy (0.80-0.84) in subgroup classification, requiring minimal early data (as little as 2 weeks for some subgroups).
Conclusions:
- Distinct PTH recovery trajectories exist, identifiable through ML analysis of early symptom data.
- This predictive approach can forecast headache burden, guiding clinical decisions and optimizing eligibility for PTH clinical trials.
