Ability of Head Impact Measurements to Predict Sports Concussions: A Review

Young Joon Kim; Nasir M Stovall; Blake Bacevich; Christopher Stapleton; Ganesh Shankar; Craig A Rovito; Daniel H Daneshvar; Ron Hirschberg; Ross D Zafonte; Brian V Nahed

doi:10.1227/neu.0000000000003524

Ability of Head Impact Measurements to Predict Sports Concussions: A Review

Neurosurgery. 2025 May 27. doi: 10.1227/neu.0000000000003524. Online ahead of print.

Authors

Young Joon Kim^{1

2}, Nasir M Stovall^{1

2}, Blake Bacevich^{1

2}, Christopher Stapleton³, Ganesh Shankar³, Craig A Rovito⁴, Daniel H Daneshvar⁴, Ron Hirschberg⁴, Ross D Zafonte⁴, Brian V Nahed³

Affiliations

¹ Harvard Medical School, Boston, Massachusetts, USA.
² Athlex AI, Boston, Massachusetts, USA.
³ Department of Neurosurgery, Massachusetts General Hospital, Boston, Massachusetts, USA.
⁴ Department of Physical Medicine and Rehabilitation, Massachusetts General Hospital, Boston, Massachusetts, USA.

PMID: 40422552
DOI: 10.1227/neu.0000000000003524

Abstract

Background and objectives: Provide a review of the current landscape of motion sensor-based analyses of mild traumatic brain injury (concussion) and shed light on avenues for further investigation.

Methods: A review of the literature on motion sensor-based concussion studies was conducted using search terms "concussion prediction sensor," "concussion prediction motion," "concussion diagnosis sensor," and "concussion diagnosis motion" in PubMed (between January 2000 and March 2024). In total, 207 publications were initially identified. However, only 14 studies were ultimately included, due to lack of requisite measurement variables, focus on different outcomes, or participant overlap with included studies. Consolidation of mean and standard deviation of measurement variables was performed using the application of Cochrane formula.

Results: Across 14 studies, most used head impact data from football (85.7%) and the Head Impact Telemetry system for data acquisition (92.9%). Most of the studies used data sets from collegiate athletes (71.4%). A minority of studies included female athletes (14.3%). On average, male athletes experienced higher linear and rotational accelerations during concussive vs nonconcussive impacts (97.6 ± 33.8 g and 4614.9 ± 2568.7 rad/s2 vs 24.4 ± 16.2 g and 1641.9 ± 1216.6 rad/s2). Moreover, male athletes experienced higher linear accelerations, but similar rotational accelerations, compared with female athletes specifically during concussive impacts (97.6 ± 33.8 g and 4614.9 ± 2568.7 rad/s2 vs 43.0 ± 11.5 g and 4030 ± 1435 rad/s2). Notably, studies that predicted concussion probability using multivariate regression methods (26.7%) demonstrated challenges with accuracy due to low positive predictive values (ranging from 0.3%-0.9%) and high false-positive rates (up to 39.4%).

Conclusion: Despite the statistical differences in head impact measurements between concussive and nonconcussive impacts, they have limited clinical utility as a stand-alone concussion identification tool. Head impact measurements may be most useful when used as an adjunct to other clinical and physiological markers.

Keywords: Concussion; Impact; Prediction; Sensor; Trauma.