Data were collected between April 9 and May 3, 2020 as a part of a larger study on COVID-19 risk perception . This is a cross-sectional study conducted through a web-based survey on LimeSurvey® in Italian which participants accessed via a link distributed via e-mail and social network messaging during the early stages of COVID-19 outbreak in Italy. Participants were a convenience sample selected based on their accessibility to the online survey.
A total of 1,765 respondents completed the questionnaire under the restrictions that they had to be at least 18 years old and living in Italy during the compilation of the survey. The study was approved by the Ethical Committee of the Department of Brain and Behavioral Sciences of the University of Pavia (no. 46/2020), and informed consent was obtained from each participant.
Emotional states and engagement in daily activities during the pandemic were the key variables considered for this study. For this reason, we excluded individuals who could present alterations or limitations in the variables examined, such as healthcare workers (n = 108; 6.1%) and participants reporting being diagnosed with COVID-19 (n = 8; 0.5%) or experiencing symptoms attributable to it (n = 192; 11%). Applying these exclusion criteria, the final sample was composed of 1,457 participants (Table 1).
We assessed socio-demographic variables by asking participants’ age, gender (male = 0; female = 1), education (not having university degree = 0; having university degree = 1), marital status (not married = 0; married = 1), employment (not working = 0; working = 1), living alone (0 = no; 1 = yes).
Positive and negative emotional experiences
We assessed participants’ positive and negative emotional experiences using the 37-item shortened version of the Profile of Mood States (POMS) . Participants rated, on a 5-point Likert scale (0 = not at all; 4 = extremely), the extent to which they experienced each emotional state during the last week. POMS yields separate subscales for depression-dejection (e.g., unhappy, sad), tension-anxiety (e.g., anxious, nervous), anger-hostility (e.g., angry, annoyed), fatigue-inertia (e.g., exhausted, weary), confusion-bewilderment (e.g., confused, bewildered), and vigor-activity (e.g., energetic, active). The items of the vigor-activity subscale were summed to create a score of positive emotional experienceFootnote 1 (Cronbach’s alpha = 0.87). The scores of the depression-dejection, tension-anxiety, anger-hostility, fatigue-inertia, and confusion-bewilderment were summed to create a negative emotional experience score (Cronbach’s alpha = 0.96). Higher scores indicate high levels of either positive or negative emotional experience.
Everyday activities during pandemic
Information on everyday activities that participants engaged during the COVID-19 lockdown was obtained using the open question: “What did you start or re-start doing during COVID-19 lockdown?” Participants’ responses were coded offline to identify the activities started or re-started during the lockdown. Then, activities were grouped into 8 categories following the classification adopted in previous studies [12, 13]: physical activity (e.g., sport, gym, yoga, pilates), cognitive activity (e.g., reading, writing, playing musical instruments, studying something new, playing games, working crossword puzzles), productive activity (e.g., sewing, painting, drawing, model-making, photography), recreational activity (e.g., watching television, listening to the radio, browsing social networks), domestic activity (e.g., clean house, gardening), social activity (e.g., spending time with family and friends, participating in video or phone calls), self-care activity (e.g., relaxing, resting, sleeping, time for oneself), and religious/spiritual activity (e.g., praying, meditating). The dependent variable for each category was coded as 1 when participants engaged in at least one activity in the category or 0 otherwise.
Perception of more closeness to relatives and friends during the pandemic
We assessed participants’ perception of more closeness to family (i.e., wife/husband, son, parents) and friends during the pandemic asking participants “Compared to usual, in this period of emergency, I feel closer to family/friends”. For each question (family and friends, respectively), responses were provided on a 5-point Likert scale (0 = not at all; 4 = extremely).
Skewness of frequency distribution was used to judge the normality of data. Data are normally distributed when skewness is equal to zero , with values between − 2 and + 2 considered acceptable cut-off . Absolute values of skewness for all our continuous variables fall below 1, indicating that adoption of parametric tests was appropriate. Therefore, we summarized continuous variables through means and standard deviations, and categorical variables through frequencies.
First, correlation analyses were conducted to examine the relationships between age, positive/negative emotional experiences and the other key variables included in the study. We computed Pearson correlations between age, positive/negative emotional experiences, and closeness to family and friends. The correlations of age, positive/negative emotional experiences and other variables are point-biserial correlations reflecting relationships between dichotomous variables and continuous variables. Subsequently, we ran hierarchical regression analyses to test whether age differences in positive and negative emotional experiences were explained by the engagement in daily activities and/or by the perception of more closeness to family and friends. In these regressions, we entered age in the first step, followed by socio-demographic variables in the second step, activities in the third step, and perception of more closeness to family and friends in the final step. Only variables that were significantly associated with age were entered. The assumption of normal distribution of the residuals was checked by inspecting skewness values with acceptable limits of ± 2 . We checked absence of multicollinearity among predictors using tolerance statistic greater than 0.2 . The assumption of independent errors was checked using the Durbin–Watson statistic, with values close to 2 meaning that the residuals are uncorrelated . All analyses were conducted using SPSS .