Derivation and Initial Validation of Clinical Phenotypes of Children Presenting with Concussion Acutely in the Emergency Department: Latent Class Analysis of a Multi-Center, Prospective Cohort, Observational Study

The identification of clinical phenotypes may help parse the substantial heterogeneity that characterizes children with concussion. This study used latent class analysis (LCA) to identify discernible phenotypes among children with acute concussion and examine the association between phenotypes and persistent post-concussive symptoms (PPCS) at 4 and 12 weeks post-injury. We conducted LCA of variables representing pre-injury history, clinical presentation, and parent symptom ratings, derived from a prospective cohort, observational study that recruited participants from August 2013 until June 2015 at nine pediatric emergency departments within the Pediatric Emergency Research Canada network. This substudy included 2323 children from the original cohort ages 8.00-17.99 years who had data for at least 80% of all variables included in each LCA. Concussion was defined according to Zurich consensus statement diagnostic criteria. The primary outcome was PPCS at 4 and 12 weeks after enrollment. Participants were 39.5% female and had a mean age of 12.8 years (standard deviation = 2.6). Follow-up was completed by 1980 (85%) at 4 weeks and 1744 (75%) at 12 weeks. LCA identified four groups with discrete pre-injury histories, four groups with discrete clinical presentations, and seven groups with discrete profiles of acute symptoms. Clinical phenotypes based on the profile of group membership across the three LCAs varied significantly in their predicted probability of PPCS at 4 and 12 weeks. The results indicate that children with concussion can be grouped into distinct clinical phenotypes, based on pre-injury history, clinical presentation, and acute symptoms, with markedly different risks of PPCS. With further validation, clinical phenotypes may provide a useful heuristic for clinical assessment and management.