Beyond total scores: symptom composition as a predictor of recovery according to concussion history

The aim of this study was to determine whether early symptom cluster composition at presentation predicts recovery after sport-related concussion, and whether these associations differ by concussion history.

The authors conducted a retrospective cohort study of athletes aged 12-22 years who were evaluated at their regional concussion center between November 2017 and April 2022. Cluster ratios for cognitive, somatic, and emotional symptoms were calculated as domain sums divided by the initial Post-Concussion Symptom Scale (PCSS) score. Primary outcomes were the number of days to return to learn (RTL), symptom resolution (SR), and return to play (RTP). Multivariable linear regression models were fit separately for athletes with no prior concussion and for those with ≥ 1 prior concussion, entering all three cluster ratios and adjusting for age, sex, time to presentation, and a history of attention-deficit/hyperactivity disorder, migraine, and psychiatric conditions. Effect modification was assessed using a hierarchical full-cohort model that included a cognitive ratio × number of prior concussions interaction term. Kaplan-Meier curves were used to assess recovery across strata defined by prior concussion (yes/no) and high versus low cognitive burden (median split), with groups compared using log-rank tests.

Of 2086 screened patients, 476 met the inclusion criteria (58.4% with no prior concussion and 41.6% with ≥ 1 prior concussion). Athletes with prior concussion were older (mean age 16.43 ± 1.92 vs 15.67 ± 1.55 years, p < 0.001), more often White (87.8% vs 76.3%, p = 0.011), presented later (mean 3.59 ± 3.34 days vs 2.86 ± 3.13 days, p = 0.016), and had similar initial PCSS scores (mean 29.39 ± 22.07 vs 28.32 ± 22.82, p = 0.607), but longer RTP (mean 33.40 ± 24.46 days vs 27.50 ± 22.34 days, p = 0.017). In multivariable models, symptom cluster ratios were not associated with RTL, SR, or RTP in the no prior concussion group. In the prior concussion group, a higher cognitive cluster ratio was associated with longer RTL (p = 0.020) and independently predicted longer SR (p = 0.003). In whole-cohort models, the interaction between cognitive ratio and prior concussion was significant for SR (p = 0.005), indicating that the adverse impact of cognitive-dominant profiles on recovery increased with the number of prior concussions. Kaplan-Meier analyses showed stepwise delays in RTL, SR, and RTP across cognitive burden-prior concussion strata (SR: log-rank χ2(3) = 10.62, p = 0.014; RTP: χ2(3) = 10.48, p = 0.015; RTL: χ2(3) = 8.10, p = 0.044), with the slowest recovery demonstrated in the group with prior concussion and high cognitive burden.

Symptom composition at presentation is prognostic primarily in athletes with prior concussion. Cognitive-dominant profiles predicted slower SR and RTL in this subgroup, and the adverse effect of cognitive burden on recovery increased with increasing concussion history.