Predicting prolonged concussion recovery with a decision tree model

By Minhong Kim. This article was initially published in our Concussion Update newsletter; please consider subscribing.

A study by Dr. Michael Robinson et al., published in the American Journal of Physical Medicine & Rehabilitation, used a decision tree to identify patients at higher risk of developing prolonged concussion symptoms. The researchers found that feelings of fogginess and sadness at initial assessment were predictive of prolonged symptoms “lasting 28 days or more postconcussion.”

Robinson et al. recruited 273 patients who had sustained a concussion within the past 14 days for the study. Each patient was assessed using the Sport Concussion Assessment Tool, 5th Edition, Symptom Evaluation. The assessment consists of 22 questions and asks patients to self-rate their concussion symptoms.

The authors processed data from the initial assessment through a decision tree model. The researchers found that “Of the 22 variables, only 2 contributed toward the predictive splits: Feeling like ‘in a fog’ and sadness.” The performance of the decision tree model was statistically significant, accurately predicting concussion outcomes 76% of the time. This study demonstrates that this decision tree model can help assess a patient’s risk of developing prolonged symptoms, helping healthcare providers and concussion patients determine the most optimal treatment. 

However, further research is needed before such models are implemented in actual healthcare settings. First, results need to be validated on larger, more representative sample sizes. In addition, models need to be more accurate. The model used in this study had a positive predictive value of 85.7% and a negative predictive value of 70.6% (predictive values are the likelihood that a positive or negative diagnostic result will actually lead to that outcome, rather than a false positive or negative).