Slide Slam Q15
Personal discourse as an index for mental health and neurological impact of COVID-19
Sarah Newman-Norlund1, Lisa Johnson1, Sara Sayers1, Kyler Spell1, Samaneh Nemati1, Nicholas Riccardi1, Roger Newman-Norlund1, Julius Fridriksson1; 1University of South Carolina
Introduction: Emerging evidence suggests that SARS-CoV-2 (COVID-19) can be the cause of long-term neurological impairment, even with disparate levels of infection severity. Neurological symptoms that include; fatigue, impaired cognition, increased anxiety, depression, and loss of taste/smell. Stress and prolonged anxiety can result in acute and chronic changes in neurochemical systems and brain regions (hippocampus, amygdala, and medial prefrontal cortex). Here, we evaluate the utility of self-generated narrative discourse (personal COVID-19 stories) as a potential predictor of persistent neurological and mental health effects of COVID-19 as indexed by questionnaires, assessments, and MRI brain markers. Methods: 110 recovered individuals (ages 20-71) who experienced varying COVID-19 severity completed questionnaires, behavioral, and neuroimaging tasks. Self-generated narrative discourse samples detailing individual COVID-19 experiences were recorded and transcribed using Microsoft 365 software by trained assistants for accuracy. Using the Text Mining package tidytext (R-Studio), we analyzed the frequency of positive/negative words and categories of sentiment as defined in the NRC Word-Emotion Association Lexicon. Emotional valence, sentiment, mental health, viral severity, and brain measures of white matter hyperintensities (WMH) measured by Fazekas scale and gray matter volumes (GMV) measured using Cat12 were evaluated with one-tailed Pearson correlations. Results: Overall emotionality of COVID-19 stories was negatively correlated with mental health and well-being. Negative sentiment was associated with higher anxiety (r(109)=-.203, p = .017), depression (r(109)=-.205, p = .016), sleep disturbance (r(109)=-.211, p = .014,), and PTSD severity (r(109)=-.264, p = .003). More negative content in the COVID-19 story was also correlated with worse MoCA performance (r(109)=.232, p = .007) and MoCA fluency (r(109)=.197, p = .02). Accounting for age, the relationship between the overall emotional valence and MoCA scores remained significant (r(109)=.260, p =.003). Additionally, we evaluated the relationship between neural integrity (GMV and WMH) and emotion scores. Results indicate higher GMV in the left hippocampus with overall negative emotional content (r(105)=-.184, p =.03). GMV in the left hippocampus and left amygdala correlated with sentiment of fear (r(105)=-.172, p =.03), trust with the right and left hippocampus (r(105)=-.190, p =.02). Higher anxiety was correlated with higher GMV in the left hippocampus (r(105)=.228, p =.01). Periventricular WMHs were correlated with fear(r(97)=.183, p =.03), sadness (r(97)=.173, p =.04), and trust (r(97)=.192, p =.03), but not joy (r(97)=.043, p =.329). Deep WMHs were correlated with depression (r(97)=-.171, p =.04). Discussion: Self-generated discourse about prior stressful events can be used as an index of mental health and underlying neurological change in areas known to be related to memory and emotion including the hippocampus and amygdala. Further studies are needed to determine if these effects are a direct result of the COVID-19 virus itself or related to other, co-occurring factors. Analysis of discourse content could help identify individuals facing long-term neurological impacts and most in need of support/intervention.