Pre-season computerized baseline batteries (ImPACT-style) versus post-injury results — the reliable change index and its role in guiding concussion management
Before a single practice snap, computerized neurocognitive test batteries such as ImPACT (Immediate Post-Concussion Assessment and Cognitive Testing), CogState, and Axon establish an individualized pre-injury reference point. Rather than comparing an injured athlete to a generic population average, clinicians can compare the athlete to themselves — a substantially more sensitive approach given how widely "normal" cognitive performance varies between individuals.
Modern computerized batteries decompose cognition into discrete, independently scored domains rather than a single global number:
• Verbal memory — word-list and passage recall tasks (immediate and delayed) probing encoding and retrieval of verbal material; sensitive to hippocampal/temporal lobe disruption
• Visual memory — recall and recognition of abstract designs, patterns, or object arrays; taps visuospatial encoding networks distinct from verbal circuits
• Processing speed — timed symbol-matching, number-sequencing, or visual-motor tasks measuring how quickly information is scanned and manipulated, independent of accuracy
• Reaction time — simple and choice reaction-time paradigms measured in milliseconds; among the most consistently sensitive metrics to concussion because it stresses fast neural transmission and attentional networks
• Impulse control — commission-error tasks (e.g., responding to a "no-go" stimulus) that quantify response inhibition; poor impulse control after injury may reflect frontal-executive disruption or reduced test engagement
Each domain is scored and combined into a composite index, but clinicians are trained to examine the domain-level pattern rather than the composite alone — a global drop driven entirely by reaction time has different implications than a drop distributed evenly across all five domains.
ImPACT — the most widely deployed platform in North American school and professional sport, used by the NFL, NHL, NCAA programs, and thousands of high schools; produces a standardized composite plus domain scores and a symptom checklist administered alongside the cognitive battery.
CogState — a computer-adaptive battery originally developed for dementia and clinical trial research, later adapted for sport; uses playing-card-based tasks designed to minimize cultural and language bias and to be repeatable with reduced practice effects.
Axon (formerly XLNTbrain / Sports Concussion Assessment Tool digital companions) and other platforms — increasingly integrate baseline cognitive testing with balance, vestibular-ocular, and symptom-report modules into a single digital workflow.
Across platforms, the shared design goal is the same: capture a reliable, individualized snapshot of cognitive function while the athlete is healthy, stored for comparison if a suspected concussion occurs during the season.
A baseline score is only as useful as the conditions under which it was collected. Best practice calls for baseline testing in a quiet, proctored, individual setting with the athlete rested, not under the influence of stimulants/depressants, and not currently symptomatic from a prior unresolved injury — deviations from this protocol quietly erode the value of every future comparison.
Baseline testing policies vary widely by level of competition and governing body. Professional leagues and many NCAA programs mandate annual baseline testing for all contact/collision-sport athletes. High school programs often follow state athletic association requirements, which range from mandatory annual testing to no requirement at all. Youth and recreational leagues frequently have no formalized baseline testing infrastructure whatsoever.
For athletes who are baselined, re-testing on a roughly annual to biennial cycle is common practice, particularly in adolescents whose cognitive profiles change rapidly with normal development — a baseline collected at age 13 is a poor reference point for evaluating an injury at age 16. Some protocols also require a fresh baseline following a prior concussion, once the athlete has fully recovered, since the pre-concussion baseline may no longer represent the athlete's stable "normal" state.
The baseline itself is stored securely as part of the athlete's medical record and is typically accessible to the treating athletic trainer, team physician, or credentialed clinician managing any subsequent injury — a chain of custody that matters both for privacy and for ensuring the correct reference score is used at the point of care.
A baseline score is not a fixed, perfectly measured constant — it is a single sample from a noisy underlying distribution. Test-retest reliability for computerized batteries is moderate at best, natural day-to-day biological variability shifts scores independent of any injury, and the incentive structure of return-to-play clearance creates a documented motive for athletes to deliberately underperform on baseline testing. Understanding these limitations is essential to interpreting any single baseline number correctly.
Reliability coefficients (intraclass correlation, ICC) for computerized neurocognitive domains typically fall in the 0.6–0.8 range — respectable for a psychometric instrument, but far from perfect. This means a meaningful fraction of the variance between two testing sessions in the same healthy athlete reflects measurement noise rather than any true change in brain function.
Compounding this, repeat administration of the same or similar test items produces a practice effect: familiarity with the format, item content, and response strategy inflates scores on retest, independent of any change in underlying cognition. Athletes who re-baseline annually often show modest composite score gains purely from having taken the test before — a gain that, if not accounted for, could mask a true post-injury decline or be mistaken for genuine improvement.
Natural biological variability — sleep quality, hydration, caffeine intake, time of day, stress, and even menstrual cycle phase — further shifts scores test to test, all before any concussion occurs. This background noise sets a floor below which small score changes cannot be confidently attributed to injury.
Because a lower baseline score makes it statistically easier to "return to normal" after a real concussion (a smaller post-injury drop is needed to fall back within the athlete's own normal range), some athletes — particularly in high-stakes competitive environments — intentionally underperform during baseline testing. This practice, colloquially termed "sandbagging," directly undermines the individualized-comparison model that baseline testing depends on.
Survey-based studies estimate self-reported deliberate poor effort at baseline in roughly 4–10% of tested athletes, with higher rates reported in populations facing greater competitive or scholarship pressure. Because sandbagging artificially lowers the reference point, it can allow a genuinely concussed athlete to test as "back to baseline" while still cognitively impaired — a direct patient-safety concern.
This vulnerability motivated the development of embedded validity/effort indices — algorithms within the test software that flag implausible response patterns (e.g., correct-answer rates below chance, inconsistent reaction times, or scores far outside what the item difficulty would predict). A flagged "invalid" baseline should prompt re-testing rather than being accepted as a true reference point.
Testing conditions materially affect scores. Group-administered baseline testing in a noisy gymnasium or computer lab, with dozens of athletes tested simultaneously and minimal proctoring, introduces distraction, rushed responses, and reduced engagement compared with individually proctored, quiet-room administration. Studies comparing group versus individual administration have found meaningfully different score distributions and higher rates of invalid profiles in group settings.
Device and platform variability (touchscreen vs. mouse vs. trackpad, screen size, browser lag) can also subtly influence reaction-time-dependent domains. Consistency of testing environment between baseline and any future post-injury administration is a recommended — though not always achievable — best practice.
Once a concussion is suspected, the temptation is to re-test cognition immediately. In practice, testing during the acute symptomatic window is confounded by headache, fatigue, dizziness, and reduced concentration that would depress scores in almost anyone, injured or not. Proper methodology delays testing, and — critically — uses statistical methods designed to distinguish a true cognitive decline from the ordinary noise inherent in any repeated measurement.
Testing an athlete within minutes to hours of a suspected concussion, while acute symptoms (headache, dizziness, nausea, photophobia, brain fog) are at their peak, produces artificially depressed scores that reflect the acute symptomatic state as much as any lasting cognitive impairment. Most protocols therefore recommend deferring formal neurocognitive re-testing until the athlete is relatively settled — commonly 24 to 72 hours post-injury or later, and ideally once acute symptom burden has at least partially subsided — so that the resulting score comparison is more interpretable.
This delay is a deliberate methodological choice, not a matter of convenience: testing too early conflates "currently has a headache" with "has lasting cognitive impairment," while testing too late risks missing the window where results could inform early management decisions.
Because any two administrations of the same test will differ somewhat due to measurement error and practice effects even with no true change, a raw score drop cannot, by itself, be interpreted as clinically meaningful decline. The Reliable Change Index (RCI) formalizes this distinction statistically:
RCI = (Post-injury score − Baseline score) ÷ (SEM × √2)
Where SEM (standard error of measurement) is derived from the test's reliability coefficient and standard deviation: SEM = SD × √(1 − reliability). The √2 term accounts for the fact that two separate measurements — baseline and post-injury — each carry their own error.
A resulting RCI value beyond a chosen threshold (commonly ±1.645 for a 90% confidence one-tailed test, or ±1.96 for 95% two-tailed) indicates the observed change is unlikely to have occurred by measurement error and practice-effect variability alone — i.e., a "reliable" change. Some protocols apply a practice-effect correction, subtracting an expected retest gain before computing the RCI, since without correction natural practice gains can mask a real decline.
Critically, RCI answers a narrow statistical question — "is this change larger than expected noise?" — not the broader clinical question of whether the athlete is safe to return to play. It is one quantitative input into that larger judgment.
An athlete whose post-injury composite score drops 6 points from an 85-point baseline might look concerning on the surface. But if the SEM×√2 for that battery is roughly 6.5 points, the resulting RCI (~0.92) falls short of the 1.645 threshold — statistically indistinguishable from normal test-retest variability, even though the raw number "looks worse."
A large fraction of athletes, especially in youth and community sport settings without baseline testing infrastructure, have no individual baseline on file. In these cases, clinicians fall back on population normative data — age-, sex-, and sometimes education-adjusted reference ranges derived from large healthy cohorts.
Normative comparisons are inherently less sensitive than individualized baseline comparisons: an athlete who was already, say, one standard deviation below the population mean on processing speed before any injury (entirely normal individual variation) could appear "impaired" relative to norms even without a concussion, while a high-performing athlete's post-injury decline could still fall within the "normal" population range and be missed. This asymmetry is a core argument in favor of baseline testing where feasible — but is also why testing infrastructure gaps in under-resourced sport settings represent a genuine equity concern in concussion care.
No responsible concussion protocol treats a neurocognitive test score, however statistically rigorous, as a standalone diagnostic tool or an automatic return-to-play gate. Cognition is only one system affected by concussion, testing captures only a narrow slice of function at a single moment, and the correlation between symptom status and test performance is imperfect in both directions.
Two patterns of discordance are well documented and both carry clinical significance:
• Normal neurocognitive scores despite ongoing symptoms: an athlete may report persistent headache, dizziness, or fogginess while scoring within their normal range (or showing an RCI below threshold) on computerized testing. This does NOT rule out concussion or clear the athlete — symptom report remains a primary clinical indicator, and computerized batteries do not capture every affected domain (e.g., they are poor at capturing vestibular, ocular-motor, or exertional symptoms).
• Abnormal cognition despite symptom resolution: conversely, some athletes who report feeling entirely back to normal still show a reliable decline on repeat testing. This scenario is precisely why many protocols require neurocognitive testing to return toward baseline (in addition to symptom resolution) before advancing through return-to-play stages — self-reported feeling well is necessary but should not be treated as sufficient on its own.
Both patterns illustrate the same underlying point: symptom checklists and cognitive testing measure overlapping but distinct aspects of concussion pathophysiology, and neither alone provides a complete picture.
Current best-practice concussion assessment combines neurocognitive testing with several other pillars, typically including:
• Graded symptom checklist — self-reported severity across physical, cognitive, emotional, and sleep-related symptom clusters, tracked serially over the recovery course
• Balance testing — e.g., the Balance Error Scoring System (BESS) or instrumented postural sway assessment, capturing vestibulospinal and cerebellar contributions not reflected in cognitive scores
• Vestibular-ocular-motor screening (VOMS) — smooth pursuit, saccades, convergence, and vestibulo-ocular reflex testing, which frequently reveals deficits and symptom provocation even when cognitive testing is unremarkable
• Clinical judgment and history — a licensed clinician's longitudinal assessment, incorporating injury mechanism, prior concussion history, comorbidities (migraine, ADHD, learning disability, mental health history — all of which can independently affect baseline and post-injury scores), and exertional/cognitive-load testing
Neurocognitive test results are interpreted within this full context, weighted alongside — never substituted for — the other pillars.
Every major sport medicine consensus statement is explicit on this point: neurocognitive testing should never function as the sole gatekeeper for return-to-play decisions. It informs, but does not replace, comprehensive clinical evaluation by a qualified healthcare provider.
A single post-injury test administration provides a snapshot; serial testing across the recovery course provides a trajectory, which is generally far more clinically informative. Athletes are commonly re-tested at intervals — for example at initial post-injury assessment, again after symptoms substantially resolve, and again before final return-to-play clearance — allowing clinicians to visualize whether cognitive scores are trending back toward the individualized baseline in step with symptom resolution, lagging behind it, or in the more concerning scenario, failing to trend upward at all.
A flat or worsening trajectory over multiple sessions, even if any single RCI value is not dramatic, is treated with more clinical concern than an isolated one-time score below threshold. This longitudinal framing also helps buffer against the reliability limitations discussed earlier — a consistent pattern across three sequential tests is considerably more trustworthy than any single data point in isolation, since random measurement noise is less likely to produce the same directional error repeatedly.
Despite widespread adoption, the research literature on baseline and computerized neurocognitive testing is genuinely mixed. Some studies show meaningful added diagnostic value over clinical assessment alone; others find limited incremental benefit once symptom report, balance, and vestibular-ocular findings are already accounted for. International consensus statements have converged on a measured, adjunctive role for these tools — one that continues to be revised as evidence accumulates.
A substantial body of research has evaluated whether computerized neurocognitive testing meaningfully improves concussion diagnosis and return-to-play decision-making beyond what a thorough clinical history, symptom inventory, and physical/vestibular exam already provide. Findings are inconsistent:
Supportive evidence: some studies find that neurocognitive testing detects residual deficits in athletes who report feeling symptom-free, suggesting an added layer of sensitivity; serial testing has demonstrated utility in tracking recovery trajectories over time in structured research cohorts.
Less supportive evidence: other studies find that once a comprehensive clinical exam and symptom tracking are performed, computerized test results change management decisions in only a minority of cases; concerns have also been raised about the moderate test-retest reliability discussed earlier, about the sandbagging vulnerability, and about the tests' relatively narrow sampling of the full range of concussion-related impairment (e.g., limited coverage of higher-order executive function, and no direct assessment of exertional or vestibular symptom provocation).
The net effect in the literature is a picture of "helpful adjunct with real limitations" rather than either "essential diagnostic gold standard" or "no clinical value" — a nuanced conclusion that resists simple headlines in either direction.
The Concussion in Sport Group's consensus statements — most recently the Amsterdam Consensus Statement (2023), building on the Berlin (2016) and earlier Zurich statements — have consistently positioned computerized neurocognitive testing as one tool within a multimodal assessment framework, explicitly cautioning against:
• Using neurocognitive test results as a stand-alone diagnostic instrument • Making baseline testing an absolute prerequisite for return-to-play clearance when no baseline is available • Relying on a single test administration or a single composite score in isolation from clinical trajectory
At the same time, these statements acknowledge neurocognitive testing can add useful objective, quantifiable data points to the clinical picture, particularly for tracking recovery over time and for identifying discordant cases (abnormal cognition with resolved symptoms, or vice versa) that might otherwise be missed by symptom report alone. Guidance in this area has been revised across each successive consensus cycle as new evidence has emerged, and further evolution should be expected.
The trajectory across Zurich (2012) → Berlin (2016) → Amsterdam (2023) consensus statements has been toward increasingly cautious, qualified language about computerized testing's role — a reflection of accumulating evidence that its stand-alone diagnostic value is more modest than the tools' rapid commercial adoption in the 2000s and 2010s initially suggested.
Baseline testing infrastructure is not evenly distributed. Professional and well-resourced collegiate/high-school programs routinely administer annual baseline testing with dedicated athletic training staff to proctor and interpret results. Many youth, community, and under-resourced school programs lack the licensing budget, hardware, trained personnel, or scheduling capacity to do the same.
This creates a two-tiered reality: athletes without access to baseline infrastructure must be evaluated against population normative data (less sensitive, as discussed earlier) or on clinical grounds alone, potentially receiving a different — not necessarily worse, but different — standard of assessment than athletes in well-resourced programs. Equitable concussion care requires that management decisions remain sound and safety-focused even in the complete absence of baseline neurocognitive data, reinforcing why guidelines refuse to make such testing a mandatory prerequisite for safe return-to-play decision-making.