Research > COVID — Jinwoo Lee

Jinwoo Lee

Research

Stress resilience and vulnerability in children facing COVID-19-related discrimination: causal machine learning approach

Summary

While perceived COVID-19-related discrimination generally worsened children’s stress levels during the pandemic, its effects were highly variable.

Individual differences in stress resilience and vulnerability to perceived discrimination were primarily linked to the frontotemporal brain structure.

Both polygenic and environmental factors influenced the neural aspects related to stress resilience and vulnerability to social stressors.

Lee J., Lee E., Kim B., Kim G., Joo Y. Y., & Cha J. (2024). Stress resilience and vulnerability in children facing COVID-19-related discrimination: a quasi-experimental study using polygenic, brain, and sociodemographic data. under review.

Fig 1. Research framework for causal forest analysis

Motivation

During the pandemic, social injustices related to COVID-19 negatively impacted children’s mental health, especially stress levels. However, it remains unclear who is more resilient or vulnerable to these effects. By examining polygenic, neuroanatomical, psychological, and environmental features of children, we addressed the following questions:

Did COVID-19-related discrimination worsen children’s stress levels during the pandemic?
Did its effect show significant individual variability?
Which attributes are mostly associated with stress resilience and vulnerability to COVID-19-related discrimination?

To faciliate robust multivariate analysis, we conducted causal forest analysis using the ABCD longitudinal dataset (N = 1,116; fig 1). We replicated all results in the larger out-of-bag samples (N = 2,503). Following variables were used in the analyses.

covariate: polygenic, neuroanatomical, psychological, and environmental features (09/2016 ~ 11/2020)
treatment: perceived COVID-19-related discrimination (11/2020 ~ 02/2021)
outcome: acute stress response (11/2020 ~ 02/2021)

Fig 2. Heterogeneous treatment effect of the discrimination and its neural correlates

Heterogeneous effects of discrimination

The acute average treatment effect (ATE) of perceived discrimination on children’s stress levels was significant (ATE = .536, P < .001). However, the effect varied notably among individuals. We divided children into five groups based on individualized treatment effect estimates (i.e., Q1 = resilience group with the least effects; Q5 = vulnerable group with the largest effects) and found significant differences in group-level ATE estimates (fig 2a).

Which baseline features are mostly associated with stress resilience and vulnerability? Neuroanatomical features in the frontotemporal regions showed the highest variable importance in causal forest analysis (fig 2b). White matter connectivity and gray matter volume variables displayed partial dependence relationships with individual treatment effects (fig 2c).

Fig 3. Multivariate interaction among gene, brain structure, and environment

Gene, environment, and brain structure

We found a close association between children’s neuronal structure and stress resilience/vulnerability to perceived discrimination. However, this does not negate the importance of genetic and environmental factors, as the brain functions as an ‘endophenotype’ that ingetrates genetic and environmental interactions.

To further explore how genetic and environmental interactions shape the frontotemporal structure involved in stress resilience or vulnerability, we used spaese generalized canonical correlation analysis (sgCCA). This method identified latent interplay among polygenic, environmental, and neuroanatomical features (fig 3a & 3b). School and familial environments emerged as primary environmental protectors for the brain, while genetic liability to mental disorders was positively associated with neural risk factors in stress resilience (fig. 3c, 3d, & 3e).

Related materials

preprint manuscript

causal machine learning workshop video (Kor)