This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in the Journal of Medical Internet Research, is properly cited. The complete bibliographic information, a link to the original publication on http://www.jmir.org/, as well as this copyright and license information must be included.
The vast majority of people worldwide have been impacted by coronavirus disease (COVID-19). In addition to the millions of individuals who have been infected with the disease, billions of individuals have been asked or required by local and national governments to change their behavioral patterns. Previous research on epidemics or traumatic events suggests that this can lead to profound behavioral and mental health changes; however, researchers are rarely able to track these changes with frequent, near-real-time sampling or compare their findings to previous years of data for the same individuals.
By combining mobile phone sensing and self-reported mental health data among college students who have been participating in a longitudinal study for the past 2 years, we sought to answer two overarching questions. First, have the behaviors and mental health of the participants changed in response to the COVID-19 pandemic compared to previous time periods? Second, are these behavior and mental health changes associated with the relative news coverage of COVID-19 in the US media?
Behaviors such as the number of locations visited, distance traveled, duration of phone usage, number of phone unlocks, sleep duration, and sedentary time were measured using the StudentLife smartphone sensing app. Depression and anxiety were assessed using weekly self-reported ecological momentary assessments of the Patient Health Questionnaire-4. The participants were 217 undergraduate students, with 178 (82.0%) students providing data during the Winter 2020 term. Differences in behaviors and self-reported mental health collected during the Winter 2020 term compared to previous terms in the same cohort were modeled using mixed linear models.
During the first academic term impacted by COVID-19 (Winter 2020), individuals were more sedentary and reported increased anxiety and depression symptoms (
Compared with prior academic terms, individuals in the Winter 2020 term were more sedentary, anxious, and depressed. A wide variety of behaviors, including increased phone usage, decreased physical activity, and fewer locations visited, were associated with fluctuations in COVID-19 news reporting. While this large-scale shift in mental health and behavior is unsurprising, its characterization is particularly important to help guide the development of methods to reduce the impact of future catastrophic events on the mental health of the population.
An outbreak of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), the virus that causes coronavirus disease (COVID-19, also known as 2019-nCov), was first reported in Wuhan, China in December 2019, and SARS-CoV-2 was identified as a novel coronavirus in January 2020. On March 11, 2020, the World Health Organization (WHO) declared COVID-19 a global pandemic; as of April 27, 2020, COVID-19 was responsible for over 200,000 deaths and 3,000,000 confirmed cases worldwide [
Initial survey-based research on the psychological impact of the COVID-19 outbreak in China suggested that the mental health impact was moderate to severe for the majority of respondents in the general population, with increased anxiety, depression, and stress attributed to the outbreak by participants [
A handful of studies have employed ecological momentary assessments (EMAs) to assess depression and anxiety more frequently and in near-real-time [
Our current work combines longitudinal smartphone sensing and EMAs collected from a cohort of Dartmouth College undergraduates to determine the impact on mental health and behaviors during the COVID-19 pandemic. Dartmouth College has four academic terms per year, which are roughly split into 10-week terms followed by 2-week (or longer) breaks. Self-reported mental health has been observed to vary across typical academic terms [
Key Dartmouth College academic dates for the Winter 2020 term and relevant COVID-19 events.
Date | Event | Term week |
January 6 | First day of classes | 1 |
January 20 | First confirmed COVID-19a case in the United States | 3 |
February 4 | Travelers from China asked to self-quarantine | 5 |
March 2 | First COVID-19 case near campus | 9 |
March 6 | Last day of classes | 9 |
March 9 | First day of final examinations | 10 |
March 10 | Students asked to leave campus as soon as possible (March 16 at the latest) | 10 |
March 11 | WHOb labels COVID-19 a pandemic; all Dartmouth College athletics cancelled | 10 |
March 12 | Travel between the United States and Europe restricted | 10 |
March 13 | COVID-19 declared a national emergency; last day of final examinations | 10 |
March 14 | Start of spring break | 11 (Break) |
March 16 | Gatherings of more than 50 individuals cancelled | 11 (Break) |
March 17 | Online-only off-campus learning for Spring 2020 term announced | 11 (Break) |
aCOVID-19: coronavirus disease.
bWHO: World Health Organization.
Proportion of US news and media stories containing the term
The current work seeks to answer the following questions. First, how and when was the mental health of college students impacted by the COVID-19 epidemic, and were changes in depression and anxiety statistically different from those in previous terms in the same student cohort? Second, how are daily behaviors (as measured by smartphone sensing) and changes in depression and anxiety impacted by COVID-19 media coverage?
All data in the current study were obtained from the second iteration of the StudentLife study [
Data were collected from 219 participants who agreed to provide mobile sensing data via the StudentLife app [
At Dartmouth College, the academic calendar consists of a flexible, year-round calendar that is approximately divided into four academic terms, or quarters. Each term consists of 10 weeks, typically followed by 2 (or more) weeks of break. The Winter 2020 academic term (January 6 start date) includes the progression of the COVID-19 pandemic, ranging from the first confirmed case in the United States (
Smartphone sensing data and EMA surveys were administered using the StudentLife application (iOS and Android) [
Sedentary time, or stationary duration, is computed to measure students’ activity or, more precisely, their lack of activity. The app continuously infers physical activities using the Android activity recognition application programming interface [
Sleep was inferred through a combination of passive sensing features (ambient light, movement activity, screen on/off). In this way, 3 features were computed: sleep onset, wake time, and sleep duration. These measures of sleep have been shown to be accurate within 30 minutes for total sleep duration [
Density-based spatial clustering of applications with noise (DBSCAN) [
Unlock duration is a measurement of time during which a mobile phone is unlocked and the screen is on; it is calculated from the time the user unlocks the phone until they either manually relock the phone or it autolocks due to disuse (the iOS default is 30 seconds, while Android defaults vary by manufacturer; users can also alter this by changing their phone settings). Notification and system services do not influence the measurement of unlock duration. While unlock duration is not an absolute measurement of phone usage, it is the closest approximation implemented in StudentLife. From the start of the study in September 2017 until September 2018, the unlock duration was measured by remotely triggering the mobile phones every 10 minutes, sampling 1 minute every 10 minute period (minimum 10% temporal coverage). If conversation was detected during the 1-minute sampling period, the sampling was extended to 3 min for a maximum of 30% temporal coverage. After September 2018, the mobile phones were remotely triggered every 3 minutes, with subsequent sampling for 1 minute. Lock/unlock behaviors within that minute were recorded in real time, while lock/unlocks for the remaining 2 minutes were logged during the next remote trigger.
To obtain an unbiased measurement of media exposure to COVID-19, the number of news articles published with the term
Data processing was performed in R [
Each variable of interest (sedentary time, depression, and anxiety) was individually modeled with COVIDTerm, term week (linear), and random intercepts per subject. The next model added an interaction term between COVIDTerm and term week (Linear). The third model added the term week (quadratic) variable, and the fourth model added an interaction between COVIDTerm and term week (quadratic). For each of the variables of interest, these four models were compared using the analysis of variance function from the base stats package in R. For each variable, the model with the lowest deviance was selected.
To obtain a daily variable mirroring possible exposure to COVID-19–related news content, a
To determine if mental health was associated with the COVID-19 news ratio, self-reported depression and anxiety variables were added to the daily model above. Given that individual participants only answered mental health EMAs on a weekly basis, the subsequent models include far fewer time points than the smartphone-sensing only model above; again, this limits the statistical comparability between the models.
Sedentary time (top), depression (middle), and anxiety (bottom) scores across an academic term and the first 2 weeks of break, with the term influenced by the outbreak of the COVID-19 pandemic as a separate line. The shaded ribbons represent the standard error for each week. Weeks influenced by policy changes related to COVID-19 are represented with a shaded box from weeks 9 to 12. Sedentary time was calculated using data from the StudentLife app. Depression and anxiety were measured with the PHQ-2 and GAD-2 scales through the StudentLife app. Control terms include data from the same group of individuals across previous academic terms. COVID-19: coronavirus disease; Finals: final examinations; GAD-2: Generalized Anxiety Disorder-2; PHQ-2: Patient Health Questionnaire-2.
Compared with all other academic terms, we observed differences in both behavior and mental health between weeks 9 and 12 of the Winter 2020 term, which correspond to the last week of classes, the week of final examinations, and the 2-week spring break. Sedentary time appeared to be very similar across the Winter 2020 and previous terms until week 11 (first week of break), when individuals spent over one hour more per day sedentary compared with a typical term; the sedentary time increased further through the second week of break (
Multiple models for sedentary time, anxiety, and depression were tested (see the Methods section for specific details). For each of these variables, superior fit (measured by lowest deviance) was observed with the most complex model. This model included the COVID-19 term, linear term, and quadratic term week trends as well as the interaction of the COVID-19 term with each of the term week trends; the model also allowed for random intercepts for each participant’s data (random effects). Modeling of the academic term affected by COVID-19 compared with academic terms prior to the COVID-19 pandemic identified significantly increased sedentary time, depression, and anxiety (
Models of sedentary time, depression, and anxiety by week and by presence of COVID-19 during the academic term.
Variable | Dependent variables | |||||
|
Sedentary time (observations=113,864) | Depression (observations=20,323) | Anxiety (observations=113,864) | |||
|
Parameter (SD) | Parameter (SD) | Parameter (SD) | |||
COVID-19a term | 0.150 (0.008) | <.001 | 0.176 (0.016) | <.001 | 0.111 (0.017) | <.001 |
Term week (linear) | –0.046 (0.003) | <.001 | 0.016 (0.005) | .004 | 0.004 (0.006) | .51 |
Term week (quadratic) | –0.045 (0.003) | <.001 | –0.078 (0.006) | <.001 | –0.108 (0.006) | <.001 |
COVID-19 term: term week (linear) | 0.138 (0.008) | <.001 | 0.058 (0.016) | .001 | 0.079 (0.016) | <.001 |
COVID-19 term: term week (quadratic) | 0.160 (0.008) | <.001 | 0.064 (0.016) | <.001 | 0.123 (0.016) | <.001 |
Constant | –0.040 (0.032) | .21 | 0.012 (0.046) | .80 | 0.040 (0.045) | .37 |
aCOVID-19: coronavirus disease.
After establishing broad differences in sedentary time, depression, and anxiety between the Winter 2020 term and previous terms, the next goal was to determine if these behaviors changed in a finer-grained fashion, particularly mirroring the relative news coverage of COVID-19. The proportion of new stories including the term
In the second model of the COVID-19 news ratio, we again attempted to make inferences with mobile sensing features plus the addition of self-reported anxiety and depression scores. When anxiety and depression were added to the previously used sensing model, we observed that increased anxiety but not depression was significantly associated with a higher COVID-19 news ratio (
Inferences of the proportion of COVID-19 news reports with smartphone features and self-reported mental health variables.
Variable | Dependent variables | |||
|
COVID-19a news model 1 (observations=100,300) | COVID-19 news model 2 (observations=18,432) | ||
|
Parameter (SD) | Parameter (SD) | ||
Depression | N/Ab | N/A | 0.003 (0.002) | .03 |
Anxiety | N/A | N/A | 0.009 (0.002) | <.001 |
Unlock duration | 0.023 (0.001) | <.001 | 0.017 (0.001) | <.001 |
Unlock number | –0.007 (0.001) | <.001 | –0.002 (0.001) | .13 |
Sedentary time | 0.011 (0.001) | <.001 | 0.011 (0.001) | <.001 |
Sleep duration | –0.00002 (0.001) | .97 | –0.0002 (0.001) | .87 |
Number of locations visited | –0.018 (0.001) | <.001 | –0.020 (0.001) | <.001 |
Distance traveled | 0.001 (0.0004) | .14 | 0.001 (0.001) | .44 |
Term week (linear) | 0.022 (0.0004) | <.001 | 0.024 (0.001) | <.001 |
Term week (quadratic) | 0.015 (0.0004) | <.001 | 0.016 (0.001) | <.001 |
Constant | 0.029 (0.002) | <.001 | 0.028 (0.002) | <.001 |
aCOVID-19: coronavirus disease.
bNot applicable.
Coefficient plots from mixed linear models of COVID-19 news inferred with mobile smartphone features (top) or mobile smartphone features and self-reported mental health (bottom). Intercept and random intercepts per subject are not plotted. *, **, and ***:
During the outbreak of a pandemic, the primary focus is on the pathogen and its influences on physical health. Mental health and behavioral changes are often considered secondary concerns. In the present study, we leveraged longitudinal data in a cohort of 217 college students to demonstrate the impact of the COVID-19 pandemic on mental health and behavior compared to previous academic terms. During the Winter 2020 academic term, sedentary time increased along with symptoms of anxiety and depression. Subsequent analyses examined the association between increased COVID-19 news coverage and behavior (inferred from mobile sensing data) and mental health. We found that with increasing COVID-19–related news, individuals were more sedentary, visited fewer locations (as inferred from GPS tracking), and showed increases in anxiety and depression. These analyses identified behavioral changes from smartphone sensing that are consistent with individuals adhering to the “Stay Safe, Stay Home” policies implemented by local and national governments.
During the Winter 2020 academic term, increased depression, anxiety, and sedentary time were observed, suggesting a large mental health and behavioral impact beyond the actual reach of the SARS-CoV-2 pathogen. The days and weeks leading up to and including the final examination period are a particularly stressful time for students [
Individuals are generally more active over break periods; however, we observed a large increase in sedentary behavior, which can be attributed to the COVID-19 pandemic. Studies show that physical activity, particularly aerobic activity, reduces self-reported depressive symptoms with a similar efficacy to that of low-dose antidepressants [
As COVID-19 news reporting intensified, we observed increased sedentary behavior and duration of phone usage along with decreases in the number of locations visited and a decreased number of phone unlocks. Initially, the number of phone unlocks was somewhat surprisingly inversely associated with COVID-19 news; however, this is likely due to increased phone unlock duration (ie, screen time), which would otherwise span multiple phone unlocks. The decreased number of locations visited is consistent with the “Stay Home, Stay Safe” polices many governments have implemented; it also stands in contrast to initial work on social distancing among college students in the Netherlands, where social distancing policies were not implemented during the examined time period [
Primary takeaways from models inferring the proportion of COVID-19 news stories suggest that during the start of the COVID-19 pandemic, students were more depressed and anxious, used their phones more, visited fewer locations, and spent more time sedentary. At this critical time of increased depression and anxiety, we issue a call to public health officials and individual citizens to raise public awareness about the benefits of aerobic exercise and unplugging from technology (moderating phone usage), as each of these has previously been shown to have positive effects on alleviating anxiety and depression [
The current study has a variety of limitations, although most provide incentives for future research. First, our participants are approximately the same age, are undergraduate college students with smartphones compatible with the StudentLife app, and are willing to participate in a multiyear longitudinal research study; this limits the generalizability of the current findings to the general population. Simultaneously, it provides distinct advantages, such as longitudinal measurement across behavioral cycles, wherein previous academic terms can be compared to the term affected by COVID-19. A second limitation is the moderate number of individuals included in the study, with 217 total participants across all terms and 178 across the COVID-19 term (83%). Despite the moderate sample size, strong significant effects on mental health and behavior were observed, suggesting robust effects.
While smartphone sensing is quite robust, there are some limitations to the interpretation of the available data. When mobility is decreased, such as during a stay-at-home order, individuals may not have their mobile phones with them at all times, which could lead to overestimation of sedentary time. Additionally, participants may be preferentially accessing larger screens (eg, tablets or laptops); therefore, phone usage (as measured by screen unlock duration or number of unlocks) may underestimate the total amount of screen time. Even so, we observed increased phone usage despite the possibility of underestimated changes in the total amount of screen time. Future work should also identify the types of screens used during this increased consumption period and quantify the relative amounts of news, social media, and other content consumed. In further work, we could also use smartwatches to improve the measurement of behaviors such as sedentary time and allow for more frequent sampling of phone usage, location, and other measures.
While this work primarily focuses on the initial days of the COVID-19 pandemic, future work would be well suited to investigating differences in mental health and behaviors between typical residential academic terms and terms that have been shifted from residential to online coursework due to COVID-19. Furthermore, identifying causal patterns between the pandemic, policy changes on national, local, and college levels, mental health, and behavior could provide further insight into the optimal development of interventions designed to mitigate mental health crises in the face of global crises. These findings suggest a greatly expanded scope of the impact of the COVID-19 pandemic beyond that directly associated with the SARS-CoV-2 pathogen.
Understanding behaviors during typical academic cycles can be fruitful for determining how everyday events impact students’ mental health and behaviors. Fringe events such as the COVID-19 pandemic provide opportunities to examine how mental health and behaviors deviate from baseline.
This study provides preliminary insight into mental health and related behaviors during the initial phases of the COVID-19 pandemic. Depression, anxiety, and sedentary time increased as the COVID-19 pandemic encroached on a college campus in parallel with large-scale policy changes. Using a mixed linear model of smartphone mobile sensing and self-reported mental health questions, we were able to infer the proportion of COVID-19–related news stories; moreover, we could validate that participants’ mental health and related behaviors changed in lockstep with increased media coverage and proximity of the pandemic. For these college students, the early days of the pandemic coincided with what is typically a time of increased time and depression, and we observed altered behavioral patterns and a decrease in mental health above and beyond typical academic terms. Much more work remains to be done to understand how behaviors and mental health change and interact in the face of monumental adverse global events.
coronavirus disease
density-based spatial clustering of applications with noise
ecological momentary assessment
Generalized Anxiety Disorder-2
Patient Health Questionnaire-2
Patient Health Questionnaire-4
severe acute respiratory syndrome coronavirus 2
World Health Organization
This work was supported by NIMH 5R01MH059282-14.
None declared.