Research · Injury & Recovery

Baseline Concussion Testing Feels Like Rigour. Check Its Measurement Properties.

By Muhammad Dzulhisham · MSc Sports Coaching, NSCA-CSCS

The argument for preseason baseline testing is intuitive enough that it rarely gets examined. You cannot know what a concussion has changed unless you know where the athlete started, so test everyone in August, and after a head knock, retest and compare. Schools buy it. Clubs advertise it. Parents find it reassuring.

The logic depends entirely on an assumption that the test measures the same thing twice. That assumption is testable, and it has been tested repeatedly.

How stable is a baseline score

The most widely used computerised battery in concussion management is ImPACT. A study reviewing 300 consecutive score reports from secondary-school students, of which 92 met inclusion criteria, computed test-retest reliabilities for its composites and subscales.¹

The authors note these figures are consistent with a large segment of the existing literature, and that the ceiling effects they found likely suppress reliability further — when a lot of healthy teenagers score at or near the maximum, the test cannot detect a decline from an inflated starting point.¹ Their practical warning is about false negatives: a concussed athlete whose baseline was artificially capped can look unchanged.

A second study of 129 collegiate athletes across two seasons found intraclass correlations that were moderate for most measures — verbal memory 0.46, visual memory 0.64, reaction time 0.61, impulse control 0.52 — and good for only one, visual motor composite at 0.77.² A third study comparing ImPACT against another tool across two consecutive seasons found that of the ImPACT outcomes, only the visual motor speed composite achieved a significant intraclass correlation between seasons.³

Three independent samples, the same picture: a baseline taken a year ago is a moderately good predictor of the same healthy athlete's score today, and for several measures it is worse than that.

How big a change has to be before it means anything

This is where the abstraction becomes concrete. Minimal detectable change is the amount a score must move before you can say it has genuinely changed rather than wobbled. A study of Division I athletes tested at baseline before two consecutive seasons calculated MDC at the 95% confidence interval for each composite.⁴

A visual memory composite has to fall by more than 24 points before the drop can be distinguished from ordinary test-to-test noise. In practice, decisions get discussed around differences far smaller than that. The same study found no difference in composite scores between the two annual timepoints, and the authors position the MDC as the clinician's guide for judging whether a change is real.⁴

None of this makes the test worthless. It makes the comparison much coarser than the precision of the printout suggests.

What the assessment literature says the diagnosis rests on

A systematic review conducted for the 2022 Amsterdam consensus process asked which tests and measures accurately diagnose persisting post-concussive symptoms. It screened 3,298 records and included 26 articles covering 1,016 participants with concussion and 531 in comparison groups — studies that compared concussed individuals either with a comparison group or with their own pre-injury baseline.⁵

The finding that matters here: no study focused on diagnostic accuracy. Some studies found differences, but definitive conclusions were not possible, because most had small convenience samples, cross-sectional designs and a high risk of bias. The reviewers concluded that diagnosis continues to rely on symptom report, preferably via standardised symptom rating scales, and that the existing research does not indicate that any other specific tool or measure has satisfactory accuracy for clinical diagnosis.⁵

There is a second timing constraint worth knowing. The systematic review that informed the Sport Concussion Office Assessment Tool screened 9,913 studies and included 127, and reported that current sport-related concussion tools have limited utility beyond 72 hours.⁶ Most of the decisions that worry a parent — is it safe to go back to training in week two, week three — sit outside the window where these tools perform.

So what is baseline testing actually for

A defensible answer is narrow. Where an athlete has a condition that makes normative comparison misleading — a learning difficulty, ADHD, a history of prior concussions — an individual baseline may be more informative than a population norm. That is a real use case, and it is the one that survives the reliability data best, because it does not depend on detecting small changes.

What the evidence does not support is the version most commonly sold: that a preseason test creates a precise personal benchmark against which a post-injury score can be read, and that the comparison can clear a child to play. Reliability in the 0.4 to 0.7 range and a 24-point detection threshold do not sustain that claim.

It is worth being clear about the failure mode, because it runs one direction. The concern raised in the reliability literature is false negatives — an athlete who is genuinely impaired but whose score sits within the noise band, and who is therefore cleared. A test that produces reassurance it has not earned is worse than no test, because the coach who has it stops asking.

What this changes in practice

Do not let a score overrule a symptom. The assessment reviews put symptom report at the centre of diagnosis. A normal-looking retest in a child who still says their head hurts is a reason to doubt the test, not the child.

Know the detection threshold before quoting a drop. If verbal memory has to move 18.6 points to register as real, a six-point fall is not evidence of anything and should not be described to a parent as if it were.

Refresh baselines, and expect them to move. Reliability degrades over a year, and adolescents are developing quickly. A baseline from two seasons ago is a weak reference.

Watch the ceiling. Where memory composites cluster near maximum, a genuine post-injury decline can hide inside the score range. This is the specific mechanism by which the test misses people.

Do not treat purchase as policy. Buying a testing platform is not a concussion protocol. Recognition, removal from play, and graded return are the parts with evidence behind them — covered separately in what the consensus statements actually recommend for recovery.

Where this evidence stops

The reliability studies here are modest in size — 92 usable reports in one, 129 athletes in another, 30 in a third — and two of the three drew on collegiate rather than school-age populations, which matters because adolescents are precisely the group whose cognition is changing fastest between tests. Reliability figures from college athletes may not transfer cleanly down.

Nor does any of this establish that baseline testing causes harm. No trial has randomised athletes to baseline testing or not and followed their outcomes, which means the honest position is that a widespread practice has never been shown to improve anything, not that it has been shown to make things worse.

And the boundary line, stated plainly: concussion diagnosis and return-to-play clearance are medical decisions. What the measurement literature contributes is narrower and still useful — a sense of how much confidence a preseason number deserves when it is placed in front of you, which is considerably less than its two decimal places imply.