Recent conceptualizations of concussion symptoms have begun to shift from a latent perspective (which suggests a common cause; i.e., head injury), to a network perspective (where symptoms influence and interact with each other throughout injury and recovery). Recent research has examined the network structure of the Post-Concussion Symptom Scale (PCSS) cross-sectionally at pre-and post-concussion, with the most important symptoms including dizziness, sadness, and feeling more emotional. However, within-subject comparisons between network structures at pre-and post-concussion have yet to be made. These analyses can provide invaluable information on whether concussion alters symptom interactions. This study examined within-athlete changes in PCSS network connectivity and centrality (the importance of different symptoms within the networks) from baseline to post-concussion.

Participants were selected from a larger longitudinal database of high school athletes who completed the PCSS in English as part of their standard athletic training protocol (N=1,561). The PCSS is a 22-item self-report measure of common concussion symptoms (i.e., headache, vomiting, dizziness, etc.) in which individuals rate symptom severity on a 7-point Likert scale. Participants were excluded if they endorsed history of brain surgery, neurodevelopmental disorder, or treatment history for epilepsy, migraines, psychiatric disorders, or alcohol/substance use. Network analysis was conducted on PCSS ratings from a baseline and acute post-concussion (within 72-hours post-injury) assessment. In each network, the nodes represented individual symptoms, and the edges connecting them their partial correlations. Estimations of the regularized partial correlation networks were completed using the Gaussian graphical model, and the GLASSO algorithm was used for regularization. Each symptom’s expected influence (the sum of its partial correlations with other symptoms) was calculated to identify the most central symptoms in each network. Recommended techniques from Epskamp et al. (2018) were completed for assessing the accuracy of the estimated symptom importance and relationships. Network Comparison Tests were conducted to observe changes in network connectivity, structure, and node influence.

Both baseline and acute post-concussion networks contained negative and positive relationships. The expected influence of symptoms was stable in both networks, with difficulty concentrating having the greatest expected influence in both. The strongest edges in the networks were between symptoms within similar domains of functioning (e.g., sleeping less was associated with trouble falling asleep). Network connectivity was not significantly different between networks (S=0.43), suggesting the overall degree to which symptoms are related was not different at acute post-concussion. Network structure significantly differed at acute post-concussion (M=0.305), suggesting specific relationships in the acute post-concussion network were different than they were at baseline. In the acute post concussion network, vomiting was less central and sensitivity to noise and mentally foggy more central.

PCSS network structure at acute post-concussion is altered, suggesting concussion may disrupt symptom networks and certain symptoms’ associations with the experience of others after sustaining a concussive injury. Future research should compare PCSS networks later in recovery to examine if similar structural changes remain or return to baseline structure, with the potential that observing PCSS network structure changes post-concussion could inform symptom resolution trajectories.

Previous studies have found differences between monolingual and bilingual athletes on ImPACT, the most widely used sport-related concussion (SRC) assessment measure. Most recently, results suggest that monolingual English-Speaking athletes outperformed bilingual English- and Spanish-speaking athletes on Visual Motor Speed and Reaction Time composites. Before further investigation of these differences can occur, measurement invariance of ImPACT must be established to ensure that differences are not attributable to measurement error. The current study aimed to 1) replicate a recently identified four-factor model using cognitive subtest scores of ImPACT on baseline assessments in monolingual English-Speaking athletes and bilingual English- and Spanish-speaking athletes and 2) to establish measurement invariance across groups.

Participants included high school athletes who were administered the ImPACT as part of their standard pre-season athletic training protocol in English. Participants were excluded if they had a self-reported history of concussion, Autism, ADHD, learning disability or treatment history of epilepsy/seizures, brain surgery, meningitis, psychiatric disorders, or substance/alcohol use. The final sample included 7,948 monolingual English-speaking athletes and 7,938 bilingual English- and Spanish-speaking athletes with valid baseline assessments. Language variables were based on self-report. As the number of monolingual athletes was substantially larger than the number of bilingual athletes, monolingual athletes were randomly selected from a larger sample to match the bilingual athletes on age, sex, and sport. Confirmatory factor analysis (CFA) was used to test competing models, including one-factor, two-factor, and three-factor models to determine if a recently identified four-factor model (Visual Memory, Visual Reaction Time, Verbal Memory, Working Memory) provided the best fit of the data. Eighteen subtest scores from ImPACT were used in the CFAs. Through increasingly restrictive multigroup CFAs (MGCFA), configural, metric, scalar, and residual levels of invariance were assessed by language group.

CFA indicated that the four-factor model provided the best fit in the monolingual and bilingual samples compared to competing models. However, some goodness-of-fit-statistics were below recommended cutoffs, and thus, post-hoc model modifications were made on a theoretical basis and by examination of modification indices. The modified four-factor model had adequate to superior fit and met criteria for all goodness-of-fit indices and was retained as the configural model to test measurement invariance across language groups. MGCFA revealed that residual invariance, the strictest level of invariance, was achieved across groups.

This study provides support for a modified four-factor model as estimating the latent structure of ImPACT cognitive scores in monolingual English-speaking and bilingual English- and Spanish-speaking high school athletes at baseline assessment. Results further suggest that differences between monolingual English-speaking and bilingual English- and Spanish-speaking athletes reported in prior ImPACT studies are not caused by measurement error. The reason for these differences remains unclear but are consistent with other studies suggesting monolingual advantages. Given the increase in bilingual individuals in the United States, and among high school athletics, future research should investigate other sources of error such as item bias and predictive validity to further understand if group differences reflect real differences between these athletes.

Frascati international research criteria for HIV-associated neurocognitive disorders (HAND) are controversial; some investigators have argued that Frascati criteria are too liberal, resulting in a high false positive rate. Meyer et al. recommended more conservative revisions to HAND criteria, including exploring other commonly used methodologies for neurocognitive impairment (NCI) in HIV including the global deficit score (GDS). This study compares NCI classifications by Frascati, Meyer, and GDS methods, in relation to neuroimaging markers of brain integrity in HIV.

Two hundred forty-one people living with HIV (PLWH) without current substance use disorder or severe (confounding) comorbid conditions underwent comprehensive neurocognitive testing and brain structural magnetic resonance imaging and magnetic resonance spectroscopy. Participants were classified using Frascati criteria versus Meyer criteria: concordant unimpaired [Frascati(Un)/Meyer(Un)], concordant impaired [Frascati(Imp)/Meyer(Imp)], or discordant [Frascati(Imp)/Meyer(Un)] which were impaired via Frascati criteria but unimpaired via Meyer criteria. To investigate the GDS versus Meyer criteria, the same groupings were utilized using GDS criteria instead of Frascati criteria.

When examining Frascati versus Meyer criteria, discordant Frascati(Imp)/Meyer(Un) individuals had less cortical gray matter, greater sulcal cerebrospinal fluid volume, and greater evidence of neuroinflammation (i.e., choline) than concordant Frascati(Un)/Meyer(Un) individuals. GDS versus Meyer comparisons indicated that discordant GDS(Imp)/Meyer(Un) individuals had less cortical gray matter and lower levels of energy metabolism (i.e., creatine) than concordant GDS(Un)/Meyer(Un) individuals. In both sets of analyses, the discordant group did not differ from the concordant impaired group on any neuroimaging measure.

The Meyer criteria failed to capture a substantial portion of PLWH with brain abnormalities. These findings support continued use of Frascati or GDS criteria to detect HIV-associated CNS dysfunction.