The study was conducted in accordance with the Declaration of Helsinki (as revised in 2013). The study was approved by the institutional review board (IRB) of the University of Michigan Medical School and was registered on ClinicalTrials.gov (NCT04766788). Study participants completed an IRB-approved informed consent through the SignNow platform [44]. The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. There were no deviations from the registered study on ClinialTrials.gov NCT04766788. In addition, there were no deviations from the published protocol manuscript [41]. Of note, the study was originally intended to last 1 year (Year I); however, it was expanded to follow the study participants over several years, continuing the following 2 years, designated as Year II and Year III. This was not originally described in the published protocol manuscript [41]. Please refer to the study schema in Figure 1.

All participant data were anonymized or deidentified to safeguard participant information. Study participants were compensated for their participation. In Year I, participants who completed the baseline and monthly surveys received a US $10 gift card for each survey completed. Each study participant who enrolled received a Fitbit device (valued at US $150) to use during the study and to keep afterward. Year II participants were compensated based on the surveys they completed, which varied in length: US $35 for the baseline survey, US $10 for each monthly survey (up to three surveys), and US $25 for the exit survey. Participants could earn an additional US $10 if they completed all Year II surveys, potentially receiving up to US $100. In addition, those who chose to share their Fitbit data were compensated up to US $5 per week for a maximum of 17 weeks (approximately 4 months), totaling up to US $85. Year III participants received US $25 for completing a one-time survey.

The study took place at a large public academic institution in the Midwestern United States. For the safety of participants and staff during the pandemic, study activities were conducted entirely remotely, with study materials shipped to participants’ residences. This was a 3-phase longitudinal and observational study conducted in the fall semesters of 2020, 2021, and 2022, referred to as Year I, Year II, and Year III, respectively. Notably, the same cohort of students was invited to participate each year, beginning from Year I, aside from any participants who had formally dropped out.

The research protocol was previously published [41], and additional information regarding the study design for Year I can be found in our previous manuscript [42]. Briefly, eligible participants were required to be college students of at least 18 years of age. Additional eligibility requirements included having access to the necessary resources for participating in a mHealth technology-based intervention (ie, smartphone or tablet and internet access) and being willing to use their personal equipment and internet for the study.

We recruited the Year I cohort of students in the fall semester of 2020 through an email distribution channel managed by the institution and approved by the IRB; this email list initially reached all currently registered and enrolled college students in 2020. In addition, we distributed flyers throughout campus buildings and buses. Interested participants were provided with additional study information, including an overview of study procedures, risks, and benefits. If they met the eligibility criteria, participants were given a digital IRB-approved informed consent form. In Year I, there were 2164 participants enrolled in the study. For Year II and Year III, these students were recontacted in the respective fall semester to complete HRQOL surveys.

Roadmap and Fitbit apps: In Year I, participants were instructed to download the Roadmap and Fitbit apps to their smartphones (both free of charge and available on Apple and Google app stores). Each individual participant was provided with a unique Roadmap ID in order to access the app. The Roadmap app offered 8 resilience-based positive activities, previously described in Rozwadowski et al [28] and provided in Multimedia Appendix 1. Participants could also engage with each other anonymously through a Chat Forum (ie, peer-to-peer support) on each of these topics. Participants were instructed to use these activities as much or as little as they found helpful. In addition, participants were asked to record their daily mood on a scale of 1 to 10 (1 being the worst and 10 being the best), with a reminder sent through push notification to their smartphones at 8:00 PM. In Year II, participants had the option to continue using the Roadmap activities to monitor their physical activity and sleep with Fitbits. In Year III, participants did not have access to the Roadmap activities.

Participants self-reported HRQOL assessments through the Roadmap app. The Roadmap app provides a link to Qualtrics (Qualtrics), which hosts study-specific surveys and the secure storage of participant data. We connected participant surveys together using their unique ID associated with the Roadmap app rather than any personally identifiable information. The survey data collected during Year I, Year II, and Year III were used in this study’s analysis. In Year I and Year II, the survey intervals included baseline, Month 1, Month 2, and Month 3; in Year III, the survey was collected only at baseline. At the appropriate intervals, leveraging both Roadmap app reminders and study team emails, we prompted participants to complete each survey and provided compensation for each completed survey. Surveys were scored in accordance with their standardized scoring and as previously published [42] (Textbox 1).

We analyzed data from the 3 phases, encompassing 2164 students in Year I, with 1128 students returning in Year II and 1033 students returning in Year III. All analyses were performed using the statistical software R (version 4.2.3; R Core Team) and organized in the following manner (Textbox 2).

Of the 2164 students who participated in Year I, 1128 returned in Year II, and 1033 returned in Year III. Notably, 755 students participated in all 3 years, as shown in the CONSORT (Consolidated Standards of Reporting Trials) flow diagram [57] (Figure 2). The majority of participants in Year I were White (1243/2164, 57%), female (1467/2164, 68%), and undergraduates (1429/2164, 66%). The demographic characteristics remained largely consistent across time, with the exception of an increase in those who reported no longer being students and a higher proportion of females (Table 1).

We examined student coping skills over the 3-year study period using pairwise comparisons between each pair of years, restricted to students who responded in both years. While problem-focused coping decreased from Year I to Year II (P<.001), it remained stable between Year I and Year III (P=.23) but increased from Year II to Year III (P=.02; Figure 3A). Emotion-focused coping decreased from Year I to Year II (P<.001) and from Year I to Year III (P<.001) but remained stable between Year II and Year III (P=.65; Figure 3B). Avoidant coping decreased across all time points, including from Year I to Year II (P<.001), Year I to Year III (P<.001), and Year II to Year III (P=.02; Figure 3C).

Next, we examined self-reported outcomes of depression (PHQ-9), anxiety (GAD-7), loneliness, and flourishing. Although there was no change in average depression (Figure 4A) and anxiety symptoms (Figure 4B) between Year I and Year II, both significantly declined from Year I to Year III (P=.005 and P=.046, respectively) and from Year II to Year III (P<.001 and P<.001, respectively).

The average reported loneliness scores decreased significantly over the 3 study years (P<.001 across all pairwise comparisons; Figure 5A). This correlated with increased average flourishing from Year I to Year III (P<.001) and from Year II to Year III (P<.001; Figure 5B).

We explored potential pathways to better understand the factors that may be contributing to changes in students’ flourishing over time using a series of multilevel regression analyses guided by a directed acyclic graph (Figure 6). We found that students’ flourishing improved significantly with time from the peak of the pandemic (β=.13). This improvement was partially influenced by loneliness and depression. That is, in any given year, students with more loneliness or more depression had lower flourishing (β=–.169 and –.071, respectively), while average loneliness, anxiety, and depression also decreased with the passage of time (β=–.127, –.229, and –.145, respectively). Although a formal mediation analysis confirmed the mediating role of loneliness on flourishing, anxiety and depression were not significant mediators of the relationship between pandemic year or severity and flourishing (Multimedia Appendix 2).

Next, we examined student app usage patterns over the study period. As shown in the UpSet plot (Figure 7), the 2 most frequently used activities were positive piggy bank and gratitude journal, followed by pleasant activity scheduling, savoring, and engaging with beauty. The least favored activities were love letters, signature strengths, and random acts of kindness.

Single activity usage was the most common (ie, only used that activity and none of the others within a “single” session of using the app), but notable combinations included students who used both positive piggy bank and gratitude journaling. Smaller intersections included combinations, such as (1) pleasant activity scheduling and gratitude journal; (2) engaging with beauty and gratitude journal and positive piggy bank; and (3) savoring gratitude journal and positive piggy bank.

This UpSet plot visualizes the intersection of the positive activities engaged by students using the app in Year I and Year II. The x-axis represents unique combinations of activities, while the y-axis indicates the number of users for each combination. Of note, single app usage is represented by the set size on the left: “Positive piggy bank” (431 unique users) and “Gratitude journal” (402 unique users). Other activities included “Pleasant activity scheduling” (227 users), “Savoring” (205 users), and “Engaging with beauty” (184 users). “Love letter” (78 users) and “Signature strengths” (74 users) had moderate engagement, while “Random acts of kindness” (66 users) were the least used.

Based on the above findings, we were interested in determining whether app use correlated with improved flourishing in the subsequent month. After adjusting for demographic and sociobehavioral characteristics in a longitudinal analysis, more frequent app use was associated with greater flourishing (P=.03). For example, implied by Table 2, 30 instances of app use was associated with a nearly 1-unit increase on the flourishing scale. This table presents the estimates and P values for the fixed effects in the mixed effects model.

The main finding of this study was the reporting of sustained yet evolving mental health challenges among college students over the 3 years. Depression and anxiety levels of our student population were higher than previously reported levels [12,13,58]; they remained consistent between Year I and Year II (2020-2021) but showed significant improvements by Year III (2022). This trend suggests a potential period of initial adjustments to the pandemic’s disruptions, followed by gradual mental health recovery as students adapted to the new norms and reduced pandemic-related stressors. Before vaccination, students experienced a multitude of changes related to pandemic restrictions, including lockdowns and stay-at-home orders, remote learning mandates, travel bans, social distancing measures, mask mandates, and closure of nonessential businesses [59].

Building on our previous work [41,42], the initial cohort of college students from the 2020 mHealth Roadmap Study were invited to participate in 2 subsequent semesters in the Fall of 2021 and 2022. Leveraging the use of mHealth technology, we longitudinally followed their mental health and well-being outcomes during a historically unprecedented period marked by ongoing global challenges [13]. In this study, we observed changes in student coping strategies over time. Problem-focused coping initially decreased but then increased between Year II and Year III. Emotion-focused coping consistently decreased from Year I, while avoidant coping showed a steady decline across all time points. These adaptive shifts suggest that students may have developed new strategies to manage themselves. It is possible that the type of coping strategy employed may influence mental health outcomes, as previously discussed [60]. A notable and encouraging finding in this study was the significant decrease in reported feelings of loneliness, which was associated with increases in flourishing across the study period. This suggests that students may have reestablished social connections, possibly through remote platforms initially, mHealth technology, and later in-person interactions as restrictions ease.

Students engaged with the Roadmap app with a strong preference for activities focusing on personal reflection, such as a Positive piggy bank and gratitude journal. Our findings suggest that app use may have positively impacted flourishing, which correlated with increased problem-focused coping and reduced avoidant coping. Reduced levels of depression, anxiety, and loneliness may have also played key roles.

Counter to our initial hypothesis that mental health outcomes and well-being would worsen over the 3-year study period, our study revealed that, even amid the challenges posed by the pandemic, students reported significant improvements in flourishing over the 3-year period. Flourishing, a term that encapsulates functioning in a manner conducive to growth, resilience, and overall goodness, provides key insights into mental health viewed in a positive manner [61]. This concept aligns with aspects of well-being, including elements such as purpose in life, meaningful relationships, and optimism. Thus, we sought to understand possible mechanisms explaining the link between time from pandemic severity, depression, anxiety, loneliness, and flourishing.

Our data showed that the passage of time from the peak of the pandemic corresponded with higher levels of flourishing. Furthermore, this increased duration was also associated with decreased feelings of loneliness, anxiety, and depression. An increase in flourishing was partially influenced by lower levels of loneliness and depression in the regression model. Interestingly, the mediation analysis only supported the role of loneliness partially mediating increased flourishing, whereas anxiety and depression did not significantly impact the relationship between time since the pandemic’s peak and flourishing. While anxiety had a direct negative effect on depression, anxiety did not influence loneliness. These findings align with ongoing debates about the causal relationships and comorbidity between anxiety and depression [62], as well as their interplay with loneliness [63]. Accordingly, there may be additional factors between loneliness and the constructs of anxiety and depression that were not captured by this study.

Our findings suggest that the time elapsed since the early days of the pandemic may have directly influenced students’ flourishing, possibly due to reasons such as increased rates of vaccination, increased opportunities for social interaction with lifts on social restrictions and mask mandates, and return to in-person classroom settings. It is also possible that app use may have contributed to some of our observations when students had access to the Roadmap’s positive activities in Year I and Year II of the study. During the early period of social restrictions, mHealth technology may have offered an avenue for students to re-establish social connections and practice simple strategies through well-studied constructs, such as positive daily reflection, gratitude, and savoring.

Previous studies have shown that emotion-focused coping (eg, acceptance, religion, and social support) is associated with improved psychological adjustment, while avoidant coping is linked with increased distress. Problem-focused coping is less used in unpredictable events [64,65] but correlates with positive outcomes [66]. Early in the pandemic, studies indicate that students primarily used emotion-focused coping strategies [65,67,68]. Contrary to these findings, our study participants reported decreased use of problem-focused, emotion-focused, and avoidant coping strategies in the pandemic year (2020). This decline may reflect the widespread use of disengagement strategies at the pandemic’s peak. Over time, students in our study reported a decline in avoidant coping with changes in problem-focused and emotion-focused coping, aligning with other recent pandemic-related studies [65,67,68].

The coping patterns observed herein correlated with an overall decline in depression and anxiety symptoms across each subsequent year of the study. Studies have shown varied findings regarding student psychological distress during the pandemic, with some reporting positive self-efficacy in online learning [69]. However, during the same period (2020-2022), a large national survey of college student mental health (Healthy Minds Network) reported increases in the percentages of students with elevated depression and anxiety scores (PHQ-9: 18%-23%; GAD-7: 31%-37%) [58]. Although we acknowledge differences in participant pools, our study followed the same cohort across the 3 years, while the Healthy Minds Network recruits new students annually; a notable difference is that our students had access to a positive psychology–based mHealth app (Roadmap). In contrast, to our knowledge, access to mHealth app support in the Healthy Minds Network population is not known.

Students who engaged with the Roadmap app favored activities that focused on personal reflection and mindfulness (ie, inward-facing), such as the positive piggy bank and gratitude journal. A recent study showed that mindfulness meditation delivered through mHealth was an approach used by college students to reduce stress and improve mindfulness and self-compassion [70]. Interestingly, we observed lower engagement in activities, such as random acts of kindness or love letters, which involved interaction with others or had an external focus (ie, “outward-facing”). It is possible that during periods of strict lockdown and social distancing, activities requiring physical interaction or outward expressions of kindness, like random acts of kindness or writing love letters, may have been perceived as difficult or impractical. The pandemic created barriers, particularly at colleges, to social interactions, making it harder for students to engage in activities that relied on direct, personal contact or social exchange [71]. During significant life transitions, of both university life as well as the COVID-19 pandemic, it is possible students favored more inward-facing strategies to support their well-being rather than prosocial, outward-facing strategies [72].

The literature suggests that the pandemic may have exacerbated mental health challenges among college students, a population already experiencing increased reports of anxiety, depression, and stress. The experience of social isolation, particularly during a crucial period of social interactions and development growth as emerging adults, may have led to poorer management of other aspects of well-being. Understanding these trends and identifying groups with increased prevalence are crucial for developing novel approaches to mental health care aimed at effectively addressing these issues. To mitigate these challenges, the adoption of novel mHealth technologies through smartphone apps has significantly accelerated over the past several years [73]. Such technologies, as leveraged in this study, hold great potential to fundamentally address mental health challenges, particularly in college students [74].

This study had several limitations. Students were self-selected to be in this study, which was meant to focus on positive psychology-based strategies to enhance mental health. It relied on self-report surveys to collect information about mental health and well-being, and the majority of participants were female. Approximately one-third of the initial participants did not take part in subsequent years of the study (ie, lost to follow-up). In addition, as we followed students over the course of 3 years, some graduated and were no longer on campus, potentially reducing their access to college resources. It is important to note that our findings are specific to students from a single Midwestern location and may not be generalizable to other regions or student populations. For example, certain demographic groups might be underrepresented or overrepresented due to the college’s location in the Midwest. We did not collect data on participants’ ongoing or concurrent mental health treatments or medications. Finally, we acknowledge the lack of a fully powered, randomized controlled trial design. Despite these limitations, this study constitutes a large-scale, longitudinal study conducted during an unprecedented time, providing valuable insights into the mental health challenges of college students.

This longitudinal study of college students observed a decline in average loneliness, depression, and anxiety with increasing time from the pandemic peak that coincided with increased flourishing. This increased flourishing was partially but not fully explained by improved loneliness, anxiety, and depression. Students had access to a positive psychology-based app and used variable coping strategies during the study. Future research should examine the efficacy of the Roadmap app through a randomized controlled trial, using a robust study design to assess its impact on mental health outcomes. To our knowledge, this study did not produce any unintended consequences through student participation. There were no privacy breaches or technical problems during the study that may have impacted the outcomes.