Bayesian mixed models approach to exploring resilience : impact of stress on subjective health and affects over time during the COVID-19 pandemic

Abstract

Background Profound stressors such as the COVID-19 pandemic have highlighted the importance of understanding resilience mechanisms and approaches for quantifying them in longitudinal studies.

Methods We used Bayesian mixed models to analyze resilience dynamics with ordinal dependent variables: subjective physical and mental health, and fear, sadness, and anger. The models included fixed effects for individual stressors and random intercepts for participants, applied to the Gutenberg-COVID-19 cohort study.

Results There were 206,912 responses from 7386 participants (mean age 55.09 years, 51.52% women) over one year (Oct 29, 2020 - Oct 25, 2021). Social stressors, such as loss of social contacts, had stronger negative associations with health and negative affects than work-related stress. Subjective health and emotions declined during lockdowns but quickly recovered afterward.

Conclusion Our longitudinal study design and mixed-model analysis highlight the role of social stress and encourage further research into protective factors like social support and positive reappraisal.