Creative leisure activities, mental health and well-being during 5 months of the COVID-19 pandemic: a fixed effects analysis of data from 3725 US adults

Introduction We investigated whether changes in engagement in home-based creative activities were associated with changes in depressive symptoms, anxiety symptoms and life satisfaction during the COVID-19 pandemic, aiming to replicate findings from the UK in a USA sample. Methods 3725 adults were included from the COVID-19 Social Study in the USA, a panel study collecting data weekly during the COVID-19 pandemic. We measured engagement in eight types of creative leisure activities on the previous weekday between April and September 2020. Data were analysed using fixed effects regression models. Results Increased time spent gardening was associated with reductions in depressive and anxiety symptoms and enhanced life satisfaction. Spending more time doing woodwork/DIY and arts/crafts were also associated with enhanced life satisfaction. However, more time watching television, films or other similar media (not for information on COVID-19) was associated with increased depressive symptoms. Other creative activities were not associated with mental health or well-being. Conclusion Some findings differ from evidence obtained in the UK, demonstrating the importance of replicating research across countries. Our findings should also be considered when formulating guidelines for future stay-at-home directives, enabling individuals to stay well despite the closure of public resources.

Participants were asked how long they had spent on the last weekday engaging in 1) reading for pleasure, 2) a home-based arts or crafts activity (e.g., painting, creative writing, sewing, playing music, etc.), 3) digital arts activities (e.g., streaming a concert, virtual tour of a museum, etc.), 4) gardening, 5) watching TV, films, Netflix, or similar (not for information on COVID-19), 6) listening to the radio or music (not for information on COVID-19), 7) doing DIY, woodwork, metal work, model making, or similar, and 8) another hobby not already mentioned. Responses were recorded on a five-point frequency scale, from "did not do" to "did for 6 or more hours". Given the low frequency of engagement in most activities, and consistent with previous research (Bu et al., 2021), we collapsed the time spent on each activity into three categories, none, low (<30 min) or high (≥30 min).

Mental health and wellbeing
Depressive symptoms were measured using the Patient Health Questionnaire (PHQ-9), a nine-item measure with scores ranging between 0 and 27 (Kroenke et al., 2001). Higher scores indicate more depressive symptoms. The standard PHQ-9 was modified in this study to ask about symptoms 'over the last week', instead of 'over the last two weeks', as data were collected weekly.
Anxiety symptoms were measured using the Generalized Anxiety Disorder Assessment (GAD-7), a sevenitem measure with scores ranging between 0 and 21 (Spitzer et al., 2006). Higher scores indicate more anxiety symptoms. As with the PHQ-9, questions asked about symptoms 'over the last week' instead of the last two weeks.
Life satisfaction (evaluative wellbeing) was measured using a single question 'Overall, in the past week, how satisfied have you been with your life?', on a scale of 0 to 10. Higher scores indicate more life satisfaction.

Statistical analysis
We used fixed effects models to test the longitudinal associations of engagement in creative leisure activities with mental health and wellbeing. This approach uses only within-individual variation to examine how the change in leisure activity engagement is related to the change in mental health within individuals over time. As individuals are compared with themselves over time, all time-invariant factors (such as gender, age, income, education, and area of living) are accounted for automatically, even if unobserved. Fixed effects models thus control for individual heterogeneity, eliminating potential biases in the estimates of timevariant variables (Allison, 2009). We tested three fixed effects models, using depressive symptoms, anxiety symptoms, and life satisfaction as separate outcomes. All eight types of leisure activities were included in each model simultaneously.
To balance the sample in relation to the target population demographics, we weighted data to match the characteristics of the non-institutionalised US population aged 18 and over. We weighted the final analytical sample according to age, gender, race/ethnicity, and education, obtained from the US Census Bureau (US Census Bureau, 2021), using the Stata user-written package ebalance (Hainmueller & Xu, 2013). To remove extreme variation, weights were trimmed to a maximum of the median plus five times the interquartile range, and then adjusted so that the total summed to the number of participants (Chowdhury et al., 2007;Potter & Zheng, 2015). For comparison, unweighted and weighted demographic characteristics of the sample are presented in Table 1. All analyses were performed using Stata 16 (StataCorp, 2019).
BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s) Note. Time spent doing activities measured on last weekday. Low = less than 30 mins doing activity during the day. High = 30 mins or more spent on activity during the day. Note. PHQ-9 score could range from 0-27, with higher scores indicating more depressive symptoms. GAD-7 score could range from 0-21, with higher scores indicating more depressive symptoms. Life satisfaction (evaluative wellbeing) could range from 0-10, with higher scores indicating more life satisfaction.  Note. Time spent doing activities measured on the last weekday. Low = less than 30 mins doing activity during the day. High = 30 mins or more spent on activity during the day. Both low and high were compared to doing none of this activity. Models show associations between changes in time spent on leisure activities and changes in mental health and wellbeing across the follow-up period (6th April -6th September 2020).  (-0.22, 0.25)  Note. Leisure activities were treated as continuous to test whether there was overall evidence for an interaction with baseline employment status. Outcomes were standardised, so coefficients represent changes in standard deviation units.  Note. Time spent doing activities measured on the last weekday. Low = less than 30 mins doing activity during the day. High = 30 mins or more spent on activity during the day. Both low and high were compared to doing none of this activity. Models show associations between changes in time spent on leisure activities and changes in mental health and wellbeing across the follow-up period (6th April -6th September 2020). Outcomes were standardised, so coefficients represent changes in standard deviation units.