There is a wide variability in the neuropsychiatric presentation of mild traumatic brain injury (mTBI), and accurate diagnosis and treatment is complicated by within-condition heterogeneity and overlapping symptoms of common comorbidities (e.g., PTSD). Such diagnostic complexities can obfuscate clinical decision-making and lead to suboptimal treatment response. In contrast to traditional diagnostic categories, person-centered analysis methods create data-derived groupings wherein individuals within a cluster are similar and individuals across clusters are different. The current study sought to apply clustering to dimensional emotional and neuropsychological features in treatment-seeking Veterans with mTBI, with the goal of identifying more precise, homogeneous clinical profiles.

Study participants were 190 Veterans with mTBI history participating in a clinical neuropsychological assessment of cognitive complaints (Mean age: 34.38, 89.6% male, average years of education: 13.14). Participants completed a diagnostic interview, neuropsychological tests, and symptom questionnaires (NBSI, PCL, BDI, BAI, AUDIT, PSQI). To identify clusters of similar neuropsychiatric presentations, we first conducted dimension reduction on data from the cognitive tests and self-report measures using principal components analysis. Second, cluster analysis and cluster validation was performed on the resultant principal components (R: kmeans, clusterboot, clusterValid) to find homogeneous subgroups of participants.

The clinical data was best represented by principal components reflecting anxious arousal, depressive cognitions, somatic post-concussive symptoms, reexperiencing and avoidance symptoms, and objective cognitive deficits. Cluster analysis using bootstrapping and cluster validity indices (e.g., Silhouette width, Dunn index) indicated that a 6-subgroup solution was optimal (subgroups were labeled Group A-Group F). Group A was characterized by moderate levels across all dimension scores. Group B was characterized by elevated somatic post-concussive symptoms and cognitive deficits. Group C was characterized by intact cognitive performance and low somatic post concussive symptoms. Group D was characterized by elevated depressive cognitions. Group E was characterized by high anxious arousal but low depressive cognitions and reexperiencing and avoidance. Group F was characterized by elevated reexperiencing and avoidance. The subgroups did not differ statistically on any demographic items, such as years of education, age, or gender. However, there were statistically significant differences across groups in performance validity failure (x2(10) = 27.17, p=.002); Group B showed the highest rate of failure.

Results demonstrate that phenotypically similar subgroups of individuals can be identified within treatment-seeking Veterans with mTBI. Data suggest that somatic post-concussive symptoms may be linked to cognitive deficits, however the rate of validity failure indicates that neuropsychological test scores may not reflect true cognitive ability. In contrast to prior studies that treat mTBI as a unitary construct that accounts for symptoms, our data suggest that a nuanced evaluation yields vastly diverse clinical presentations. Cluster analytic frameworks hold promise for better assessment and treatment planning for Veterans, as both patients and their treating clinicians would be greatly served by the ability to use common clinical assessment tools to better identify a given individual’s clinical needs. A critical next step is to validate subgroups using novel samples and data sources (e.g., neurobiology, genetics) and to determine if these subgroupings can be effectively utilized to personalize treatment assignment.