Cognitive Screening Tests in Concussion: Classification Agreement and Differential Symptom Sensitivity

The King-Devick (KD) test, Montreal Cognitive Assessment (MoCA), and computerized batteries are widely used to evaluate concussion, but whether different tools produce concordant results for clinical decision-making is unknown. We examined classification agreement among these tests and how symptom burden and time from injury affect concordance.

Retrospective analysis of 253 concussion patients (359 visits with all three tests and concurrent Neurobehavioral Symptom Inventory [NSI] scores). We assessed classification agreement (Cohen's kappa) and Spearman rank correlations, stratified by symptom burden (NSI ≤10, 11-25, >25) and time from injury, and used multivariable regression to test symptom-cognition associations while adjusting for age, sex, prior concussions, loss of consciousness, and time from injury.

Three-way agreement was virtually absent (Fleiss' κ = 0.002), with any agreement driven by concordance on normal results. Pairwise agreement was also slight (all κ < 0.20). The tests agreed on impairment in only 3 of 359 visits, and patients classified as normal by all three still had a mean symptom score of 19.6, with 30% reporting high symptom burden. MoCA did not significantly correlate with symptoms (r = -0.02), while KD (r = + 0.40) and Creyos (r = -0.33) did. Agreement declined with higher symptom burden. After adjustment, symptom burden independently predicted KD and Creyos performance but not MoCA. Test scores did not differ by injury mechanism.

Cognitive tests in concussion frequently disagree, regardless of mechanism. Although KD and Creyos significantly reflect symptom burden, MoCA does not. Concordance varies with symptom level, time from injury, and population. The clinical impression may shift with the choice of instrument. Cognitive tests are one input alongside physical examination and clinical judgment, not a stand-alone measure.