A mental health crisis has emerged during the COVID-19 pandemic. The US Centers for Disease Control and Prevention (CDC) estimates that as of June 2020 nearly one-third of US adults were suffering from anxiety or depression (1). The rates are almost two times higher for young adults, a population that has already seen a significant increase in the prevalence of mental health disorders over the past decade (2). Over 60% of individuals age 18 to 24 y were estimated to be at risk for depression or anxiety and a quarter reported considering suicide in the previous month. These estimates represent a large increase in depression rates compared to about 11% of all adults in 2019 (3) and about 25% of college students prior to the pandemic (4). The rise in depression has occurred at the same time that stay-at-home orders, campus closures, and social distancing measures have caused major disruptions to everyday life, altering the way people live, work, study, and interact.
In this paper we document disruptions in physical activity, sleep, and time use among young adults at the onset of the pandemic and examine the relationship between these disruptions and mental health. We take advantage of a wellness study that has enrolled multiple cohorts of US college students from February 2019 through July 2020. Participants received wearable devices (Fitbits) and answered repeated surveys about their well-being and time use over the course of a semester. Participants in the 2020 cohort began the study in February and continued participating after the university moved all classes online in March and encouraged students not to return to campus.
These data allow us to make two primary contributions. First, we can conduct longitudinal analysis examining how physical activity and mental health have evolved during the pandemic compared both to baseline prepandemic levels as well as to prior cohorts. The use of prepandemic data are critical as the studied behaviour’s exhibit significant seasonal patterns. Second, we can link biometric measures of physical activity and sleep to survey measures of mental well-being and social distancing. This approach allows us to identify risk factors for depression during COVID-19 and compare those factors to predictors of depression prior to the pandemic.
We first document large changes to physical activity and sleep. Over the course of the 3-mo semester, average steps decline by over half from 10,000 to 4,600 steps per day, overall physical activity declines by about a third from 4.4 h to 2.9 h per day, and sleep increases by about 25 to 30 min per night. We also find dramatic shifts in self-reported time use. Time spent socializing with others declines by over half to less than 30 min per day, while screen time more than doubles to over 5 h per day (excluding screen time for classes or work). These lifestyle disruptions stand alongside stark increases in depression during the pandemic. We estimate that at the end of the spring 2020 semester in April an estimated 61% of our participants were at risk for clinical depression. This represents about a 90% increase over rates of 32% in the same population just 2 mo earlier prior to the pandemic.
Using difference-in-differences and individual fixed-effects regressions, we show that the changes in physical activity, sleep, social interactions, screen time, and depression are all statistically significant compared to changes in prior cohorts (). The concurrent decline of both physical activity and mental health is particularly worrisome, as prior work suggests that the coexistence of mental health problems alongside poor physical activity worsens overall health outcomes (5). In line with this work, we find that large declines in physical activity during COVID-19 are associated with 15 to 18 percentage point higher rates of depression compared to small disruptions in baseline habits ().
To link lifestyle and mental health we exploit our rich longitudinal data and use tree-based classification methods to identify risk factors for depression during COVID-19. Taken together, the predictors of depression in the 2020 cohort differ significantly from prior cohorts (). When we examine specific risk factors we find that changes in lifestyle behaviour’s are more closely linked to depression during the pandemic than in prior cohorts. In particular, large disruptions in physical activity emerge as a leading risk factor for depression during COVID-19. In contrast to prepandemic cohorts in which there is little relationship between disruptions and mental health, those participants who sustain their baseline exercise habits during the pandemic are at significantly lower risk of depression.
Building on this analysis, after the spring 2020 semester ended in April we continued to track a subsample of our participants through July 2020. During this period, we find evidence of a partial “bounce back” in physical activity and mental health toward baseline levels. Average daily steps increase to about 6,400 steps per day in May and remain steady through July, closing about a third of the decline from the onset of the pandemic in March and April. There is also some decline in average measures of depression, with estimated rates of depression ranging from 46% to 50% in May through July. This represents an improvement compared to the end of the semester in April but remains flat over this period and is still about 50% higher than prepandemic rates.
In order to examine whether a policy intervention could help counteract some of the adverse impacts of the pandemic, we implemented a randomized intervention halfway through this period. Building on our findings and on prior work on the link between physical activity and mental health (6), in June 2020 we randomized half of our participants to receive incentives for walking at least 10,000 steps per day for 2 wk. Our intervention significantly increased average steps by about 2,300 steps per day and physical activity by almost 40 min per day compared to the control group (), with the treatment group close to their baseline prepandemic levels. However, the impact on exercise did not translate into an improvement in mental health measured at the end of the intervention period.
In a postintervention follow-up we find that average steps in the treatment group declined to the same levels as in the control group about a week after the intervention ended. In July 2020, 1 mo after the intervention ended, we find no differences in average measures of depression between treatment participants who were randomized to the physical activity intervention and participants in the control group.
Our study contributes to the growing literature examining the impact of the coronavirus pandemic on physical activity and mental well-being. Lifestyle disruptions during COVID-19 have been documented in studies focusing on a single type of behavior, such as exercise (7), sleep (8), social distancing (9, 10), or mental health (11⇓⇓⇓⇓⇓⇓⇓⇓⇓–21). While our sample is not nationally representative, our measures of mental health are in line with those from larger and nationally representative samples using various measures of mental health both prior to the pandemic (4) and during the pandemic (1, 22).*
Related work using cross-sectional data finds an association between self-reported changes in physical activity during the pandemic and measures of mental health (25). This paper also relates to the broader research on the determinants of mental health (26⇓⇓⇓–30) as well as work on health behavior change. Prior studies demonstrate how changing circumstances or context can quickly disrupt healthy habits (31, 32). In addition to documenting such disruptions as a consequence of the pandemic, our work investigates the relationships between disruptions in lifestyle habits and well-being.
Taken together, our findings suggest a puzzle: Why are disruptions to physical activity and mental health strongly associated but restoration of physical activity through our intervention does not meaningfully improve mental health? First, the impact of physical activity may require a longer-term intervention. Second, physical activity may have important interactions with other lifestyle behaviors such as social interactions. It may also reflect correlation with other unobserved determinants of mental health. Finally, it could be the case that the relationship between physical activity and depression is driven more by mental health than it is by lifestyle habits. For example, the strong association between maintenance of healthy habits and depression during COVID-19 could partially reflect individuals’ ability to adapt to adversity and sustain their lifestyle despite the pandemic. Such resilience in the face of large disruptions may be critical for well-being during COVID-19.