Artificial Intelligence Predicts Long-Term Effects of Concussions in Student Athletes

Researchers at the University of Michigan are pioneering the use of artificial intelligence (AI) to anticipate the long-term health effects of sports-related concussions on student athletes throughout their college careers. The findings, published in the Annals of Biomedical Engineering, indicate that five distinct AI models demonstrated superior accuracy compared to a basic benchmark model in forecasting the potential health consequences stemming from concussions.

According to the lead author of the study, this research represents a significant advancement in understanding how concussions could affect athletes long after their collegiate sports participation has ended. The study addresses a growing concern regarding the delayed onset of symptoms associated with sport-related concussions.

The primary objective of the research was to evaluate whether AI could effectively predict three critical clinical outcomes: the overall quality of life and symptom burden, cognitive status, and the risk of psychological distress among athletes. Additionally, the study sought to identify which AI models were most effective in predicting these outcomes and to explore how various factors, including the frequency and intensity of concussions, influence their progression over an athlete's career.

To conduct this analysis, the researchers utilized baseline and exit data from approximately 3,200 NCAA athletes enrolled in the Concussion, Assessment, Research, and Education (CARE) Consortium, a national research network that includes NCAA athletes and U.S. military cadets from 30 institutions. The data incorporated both demographic information and health records collected throughout the athletes' college careers.

Several key insights emerged from the research:

• The AI models significantly reduced prediction errors compared to a simplistic model that assumed no changes would occur.
• More straightforward, interpretable AI models surpassed their more complex counterparts in predictive accuracy.
• The most crucial factor in making accurate predictions was the athlete's baseline evaluation conducted at the onset of their college careers.
• Contrary to initial expectations, sport exposure had minimal impact on the progression of symptoms.
• The frequency and intensity of concussions also appeared to exert little influence on symptom progression.

The researchers expressed surprise at the latter two findings, particularly that the frequency and intensity of concussions did not significantly affect the athletes' symptom progression. Initial hypotheses suggested these factors would be critical indicators of health outcomes.

Concussions are recognized as one of the most intricate injuries within sports medicine, with many high school and collegiate athletes experiencing one or multiple concussions during their athletic careers. It is estimated that around 10% of NCAA athletes have suffered one or more sport-related concussions.

One of the study's co-authors emphasized the potential role of AI in enhancing the diagnosis and management of concussions. As interest grows concerning the long-term implications of concussions and head impacts, especially among athletes and military personnel, these findings provide valuable insights that could inform clinical care practices aimed at safeguarding long-term health.

Looking ahead, future research endeavors may involve more extensive and diverse populations. With sufficient data, there is a possibility of integrating AI into applications designed to analyze an athlete's or patient's clinical records, enabling proactive measures and monitoring for potential health issues.

The primary researcher was motivated to delve into concussion research following personal experiences with concussions, which highlighted the need for improved understanding and management of this complex area.

Co-authors of the study include researchers from various institutions, underscoring a collaborative effort to advance concussion research.