Data analysis

All statistical analyses were conducted using SPSS (Version 24; IBM, Armonk, NY). We used linear regression models for the main analyses. Continuous covariates were centered prior to use in statistical models with z-scores. Early life trauma, as measured by the Risky Family Questionnaire, was modeled as a continuous predictor. The primary hypothesis was tested using a linear regression, with early life trauma predicting changes in sleep quality from Time 1 to Time 2, while controlling for age, sex, income, and sleep quality at Time 1. In a separate hierarchical linear regression model, the relationship between early life trauma and COVID-19-specific psychological stress was investigated. In each of these hierarchical regression models, the covariates were entered into Step 1, and early life trauma was entered in Step 2. Next, to test for indirect effects of early life trauma on changes in sleep quality through psychological stress related to the pandemic, a bootstrapping approach was utilized.⁴¹ A point estimate of the indirect effect was derived from the mean of 5000 estimates of the indirect pathways, and 95% confidence intervals (CIs) were computed using the cutoffs for the 2.5% highest and lowest scores of the distribution. Indirect effects were considered statistically significant when the CI did not include 0.