A Data-Driven Approach to Unlikely, Possible, Probable, and Definite Acute Concussion Assessment

Area of Science:

• Sports Medicine
• Neurology
• Data Science

Background:

• Current concussion diagnosis relies on clinical experience, lacking objective data for certainty.
• Previous guidelines suggested incorporating certainty levels but were not data-driven.

Purpose of the Study:

• To develop and validate a data-driven predictive model for concussion risk stratification.
• To classify athletes into unlikely, possible, probable, or definite concussion categories with diagnostic certainty.
• To enhance evidence-based concussion assessment through predictive modeling.

Main Methods:

• Utilized data from the Concussion Assessment, Research, and Education (CARE) Consortium.
• Developed a predictive framework combining data-driven optimization and machine learning.
• Validated the model using acute concussion and normal performance assessment data.

Main Results:

• The algorithm achieved high sensitivity (91.07-97.40%) in classifying probable or definite concussions.
• Significant differences in Standard Assessment of Concussion (SAC), Sport Concussion Assessment Tool (SCAT) symptoms, and Balance Error Scoring System (BESS) scores were observed across risk categories.
• Baseline to post-injury changes in SAC, SCAT, and BESS scores differentiated possible/probable concussions from normal performances.

Conclusions:

• A data-driven approach to concussion risk stratification shows promise for evidence-based assessment.
• The developed framework offers a more objective method for classifying concussion diagnostic certainty.
• Further clinical interpretation is necessary, but the model represents a significant advancement in concussion diagnostics.