2.4. Analytic Methods
In Section 3.1, scores of the COVID-19 anxiety variable were categorized into quintiles, and the quintiles were dummy-coded with the lowest one used as the reference category [20] Regression models were used to estimate the relationship between the COVID-19 anxiety quintiles and the PHQ-15 subscales (pain, gastrointestinal, cardiopulmonary, and fatigue). Model 1 included the four dummy-coded COVID-19 anxiety variables as predictors of the four PHQ-15 subscales. The regression coefficient for each dummy-coded variable is interpreted as the mean difference between each quintile and the lowest one. Model 1 was also run separately with the total PHQ-15 summed scale score replacing the subscales. In Model 2 the covariates (age, gender, income, pre-existing health problems, and GAD) were included as predictors, with the addition of Italian region mourning for COVID-19 losses factors.
In Section 3.2, measures of depression (PHQ-9), generalized anxiety (GAD-7), trauma symptoms relating to the pandemic (ITQ) and COVID-19 anxiety were considered as dependent variables for three binary logistic regression models [19]. The predictor variables were age, gender, living location, living alone, presence of children in the household, income, pre-existing health conditions in self and someone close, and perceived risk of infection over the next month. Italian region and mourning for COVID-19 losses factors were included.