Head Impact Scores: Metrics for Concussion

Updated 5 December 2025

Head Impact (HI) Scores are composite metrics that quantify head acceleration events by integrating kinematic data and tissue-level strain.
They combine measures such as peak linear acceleration, angular acceleration, and power, derived from instrumented mouthguards and biomechanical models.
HI scoring systems support real-time monitoring and clinical triage with sex-specific thresholds and standardized error margins for accurate injury risk assessment.

Head Impact (HI) Scores are quantitative metrics aggregating kinematic and tissue-level responses of the brain to head acceleration events (HAEs), serving as indices for the risk of acute neurological consequences such as concussion. HI scores integrate single or multiple severity measures—derived from linear and angular head motion, biomechanical modeling, and statistical thresholding—to enable the identification, monitoring, and clinical triage of potentially injurious impacts in sports, particularly rugby union and football (Tierney et al., 2 Oct 2024, Liu et al., 2020).

1. Definition and Conceptual Basis

HI scores are scalar quantities, often composite, that synthesize biomechanically relevant features of the head’s motion and brain tissue deformation during an HAE. They function as operationalizations of “impact severity,” with the intention of discriminating between HAEs causative of clinical events (such as HIA1 removal in rugby) and benign occurrences. The HI score may be built from directly measured kinematic peaks, injury risk criteria, or output from data-driven systems mapping head motion to tissue-level strain metrics (Tierney et al., 2 Oct 2024).

2. Severity Metrics: Components and Mathematical Formulations

A number of foundational severity metrics form the core elements of current HI scoring schemes:

Peak Linear Acceleration (PLA):

$\mathrm{PLA} = \max_t \|\mathbf{a}(t)\|$ , measured in g, where $\mathbf{a}(t)$ is the three-axis linear acceleration vector at the head center of gravity (CG).

Peak Angular Acceleration (PAA):

$\mathrm{PAA} = \max_t \|\boldsymbol{\alpha}(t)\|$ , in krad/s², using the CG’s angular acceleration vector.

Peak Power (P):

The maximal rate of kinetic energy transfer, accounting for both translation and rotation:

$P(t) = m [a_x(t)v_x(t) + a_y(t)v_y(t) + a_z(t)v_z(t)] + I_{xx} \alpha_x(t)\omega_x(t) + I_{yy} \alpha_y(t)\omega_y(t) + I_{zz} \alpha_z(t)\omega_z(t)$

with $P = \max_t P(t)$ . Here, $m$ is the head mass, $I_{ii}$ the moments of inertia (sex-specific), and the respective kinematic and dynamic variables referenced to the head CG.

Head Acceleration Response Metric (HARM):

$\mathrm{HARM} = C_1 \cdot \mathrm{HIC} + C_2 \cdot \mathrm{DAMAGE}$

with $C_1 = 0.0148$ , $C_2 = 15.6$ ; HARM combines linear head injury criterion (HIC) and a rotational “DAMAGE” metric.

Maximum Principal Strain (MPS):

The 95th percentile of the instantaneous maximum principal strain in a finite element brain model, estimated using data-driven algorithms.

Additional criteria: BrIC, PCS, and various CNN-derived brain deformation metrics are also used to augment HI scoring frameworks, particularly in research settings (Liu et al., 2020).

3. Instrumented Measurement and Data Processing

Acquisition of input data for HI scores relies on instrumented mouthguards (iMGs) equipped with triaxial accelerometers (range up to ±200–400 g, sampling at 1–3.2 kHz) and gyroscopes (up to ±70 rad/s, sampling at 1–8 kHz) (Tierney et al., 2 Oct 2024, Liu et al., 2020). Essential aspects include:

Preprocessing: Low-pass (Butterworth) filtering and coordinate transformations (SAE J211; transformation to head CG).
Triggering and windowing: Accelerometer-based event trigger (e.g., ≥8g), typically capturing a 50-ms window.
Quality control: Event validation via Positive Predictive Value (~0.99), signal-to-event matching using timestamp/video.
Measurement error: Most kinematic and strain metrics can be quantified with mean relative error (MRE) <13% (and <9% for brain-strain metrics).
Standardization: Sex-specific head mass and inertia parameters, and adjustments for device-specific accuracy.

4. Statistical Modeling and Threshold Derivation

Binary logistic regression is employed to relate each severity metric $X$ to the binary response of clinical interest (e.g., HIA1 removal):

$\operatorname{logit}[P(\mathrm{HIA1})] = \beta_0 + \beta_1 X$

Thresholds for actionable HI scores are optimized by maximizing the Youden Index $J = \mathrm{sensitivity} + \mathrm{specificity} - 1$ on ROC curves, providing cutoffs with high discriminative value for case (injurious) versus control (non-injurious) impacts. Area Under the Curve (AUC), sensitivity, and specificity quantitatively describe model performance (Tierney et al., 2 Oct 2024).

Metric	Threshold (Men)	AUC (Men)	Sens (%)	Spec (%)	Threshold (Women)	AUC (Women)	Sens (%)	Spec (%)
Power	1508 W	0.961	90.0	91.3	1193.8 W	0.923	82.1	93.7
MPS	0.17	0.948	86.7	94.5	0.17	0.849	82.1	76.2
HARM	2.87	0.954	86.7	95.0	2.67	0.883	71.4	94.3
PLA	30.64 g	0.947	86.7	93.9	25.05 g	0.947	85.7	92.8
PAA	1.96 krad/s²	0.937	86.7	89.2	1.68 krad/s²	0.917	92.9	86.5
dPAV	14.75 rad/s	0.927	86.7	88.6	11.16 rad/s	0.821	82.1	72.5

Performance analyses reveal that Peak Power provides the highest or near-highest discriminative capacity in both sexes, with specificity and sensitivity suitable for operational deployment (Tierney et al., 2 Oct 2024).

5. Composite HI Scoring Schemes

Composite HI scores incorporate multiple severity metrics into a single scalar designed to flag dangerous impacts. A generic structure is:

$\mathrm{HI\!-\!Score} = W_1 \frac{a_{\mathrm{peak}}}{a_0} + W_2 \frac{\omega_{\mathrm{peak}}}{\omega_0} + W_3 \mathrm{BrIC} + W_4 \frac{\mathrm{95\%MPS}}{\varepsilon_0} + \ldots$

where $a_0, \omega_0, \varepsilon_0$ are normalization constants representing established biomechanical thresholds, and $W_i$ are weights selected through clinical correlation. Measurement errors propagate through these schemes, necessitating that operational cutpoints are increased by the error margin (typically ≥10%) to offset sensor uncertainty (Liu et al., 2020). A common approach is to supplement peak kinematic measures with tissue-level indices such as MPS or BrIC.

Recommended action thresholds (device and sport-specific) include BrIC > 1.0 (≈50% concussion risk) and 95% MPS > 0.30 (elevated axonal strain) (Liu et al., 2020).

6. Implementation and Practical Considerations

Device calibration: Accurate HI scores require precise calibration of iMGs, sex-specific biomechanical parameters, and robust preprocessing pipelines.
Event burden: Threshold crossings must be balanced to minimize false positives, avoiding clinical overload and erosion of staff confidence.
Real-time application: iMG systems allow for pitch-side computation of all input metrics, facilitating near-instantaneous risk triage.
Sex-specific tuning: Lower discriminative performance of HARM and MPS in women implicates the necessity for sex-specific thresholding and/or further model adaptation.
Technical accuracy: Kinematic and brain deformation metrics can be measured reliably within ≤13% error, with sex-dependent and location/velocity factors considered (Liu et al., 2020).

7. Impact and Research Directions

HI scores operationalize the biomechanical correlates of brain injury risk, shaping strategies in concussion monitoring, risk management, and device development in collision sports. Current evidence supports Peak Power as a robust, mechanistically-grounded severity index that outperforms traditional single-parameter kinematic metrics in both men’s and women’s rugby union for discriminating HAE events resulting in HIA1 removal (Tierney et al., 2 Oct 2024). Broader frameworks that integrate kinematic, model-based, and data-driven features are being validated to inform real-time injury prevention systems and underpin regulatory standards.

This suggests future research should address sex-specific score calibration, the integration of more advanced tissue strain metrics, and evaluation of long-term clinical outcomes beyond acute event detection.