*3.3. Risk of Bias in the Included Studies*

Eleven studies were assigned good quality scores [19,25,35,38,43,44,47,48,56,58,59] and sixteen studies were identified as of medium or fair quality [36,37,39–42,45,46,49–55,57] (Appendix A, Table A2). The kappa statistic (inter-rater agreement) was 89.5%.

#### *3.4. Meta-Analysis*

#### 3.4.1. Anxiety

The pooled prevalence of anxiety in twenty studies [19,35,36,38–40,42–44,47,50–59] with a sample size 84,097 was 39.4% (95% CI: 28.6,51.3; I<sup>2</sup> = 99.8%; *p*-value < 0.0001; Table 1, Figure 2). Sub-analyses by additional categorical moderators, including gender, quality of study, continent, country, type of survey tool, and anxiety level were also conducted. Results of sub-analyses are given in Table 1.

### 3.4.2. Depression

The pooled prevalence of depression in fourteen studies [19,25,36,38,40,43,44,46–49,57–59] with a sample size 61,392 was 31.2% (95% CI: 19.7,45.6; I2 = 99.8%, *p* < 0.0001, Table 2, Figure 3). Sub-analyses by additional categorical moderators, including gender, quality of study, continent, country, type of survey tool, and level of anxiety was also conducted (Table 2).


**Table 1.** Pooled estimates of anxiety by categorical moderator variables (subgroup analyses).

GAD: Generalized Anxiety Disorder.


**Figure 2.** Forest plot showing pooled estimates of anxiety among students.


**Table 2.** Pooled estimates of depression by categorical moderator variables (subgroup analyses).

PHQ: Patient Health Questionnaire.


3.4.3. Other Psychological Outcomes

The pooled prevalence of stress in three studies [39,41,58] with a sample size of 1799 was 26.0% (95% CI: 7.7,59.5; I2= 98.9%, *p* < 0.0001). Post-traumatic stress disorder (PTSD) in a sample of 4242 students across three studies [25,38,43] was 29.8% (95% CI:3.0, 85.4; I<sup>2</sup> = 99.8%, *p* < 0.001). The overall prevalence of impaired sleep quality among three studies [46,47,58] in a sample size of 698 was 50.5% (95% CI:23.9,76.8; I2 = 97.6%; *p* < 0.001). Suicidal ideation was assessed in only two studies [38,40] with rates of 31.3% and 63.3% respectively.

#### 3.4.4. Publication Bias

Except anxiety (*p* = 0.11), P values of Egger test indicate insignificant publication bias for depression (*p* = 0.17), stress (*p* = 0.68), sleep disturbances (*p* = 0.99), and PTSD (*p* = 0.78).

#### 3.4.5. Certainty of the Evidence

All primary studies were cross-sectional; therefore, the quality of the evidence would be moderate. However, most of the studies included in this analysis were of fair and good quality, which contributes to the certainty of the current meta-analysis evidence.

#### **4. Discussion**

The current metanalysis included 27 studies with a sufficiently large sample of (*N* = 90,879) college students to explore psychological dimensions during the pandemic. Prior studies and a few systematic review protocols [26] investigated the association between psychological health outcomes and COVID-19, but quantitative synthesis was lacking. To our knowledge, the current meta-analysis provides the first collective evidence of the negative psychological burden of COVID-19 on the mental health of college students. This evidence is critical to inform colleges, universities, and other educational institutions in designing interventions and policies to improve college students' mental health. Previous global evidence indicated that psychological morbidities were long-standing issues among college students even before the pandemic, with nearly 50% of mental issues starting at an early age of 14 years [60–62]. Globally, suicide remains among the leading causes of death among adolescents, which warrants the need to develop early interventions to address this population's mental health and emotional needs [62]. The consequences of not addressing these concerns during the early phases of life will be dire. A lack of early intervention may lead to psychological morbidities in later life phases [62]. Regarding the pandemic, it is important to intervene early to promote post-traumatic growth among students in existing and repairing phases of the pandemic. Our findings suggest a higher prevalence of anxiety (39.4%), depression (31.2%), and stress (26.0%) than those reported in the pre-pandemic period with 22.1% anxiety, 19.7% depression, and 13.4% stress [60–62]. Corollaries associated with COVID-19, including uncertainty and fear, exert an additional driving force to explain these rising trends [24]. The timeline to graduation, sudden transition to virtual learning, quality and logistics of internships, and post-graduation plans are all in uncertainty, causing significant distress among college students [24,52]. Association of other contributing factors, such as compliance to the new rules, propagation of ambiguous messages through media, and lack of scientific understanding, need to be explored fully to design a holistic public health approach to address mental health challenges among college students [60,62].

Additionally, young people like to socialize and indulge in parties and celebrations, which have been restricted in pandemic times, adding to their frustration levels [52,53,57]. Some students who receive counseling services have not been able to receive such support. Many students who work part-time jobs have lost their employment (voluntarily or employer initiated) during COVID-19, causing financial distress [24,52,53,57]. According to a study of 69,054 French students, nearly 42.8% of students reported having at least one negative mental health outcome; of those, only 12.4% sought assistance from healthcare professionals [24]. The stigma associated with seeking mental health support has been cited as a primary factor of underreported mental health issues among adolescents [62]. Among risk factors, the female gender is associated mainly with psychosocial health [24,53]. Females were twice as likely as males to experience mental health issues [24]. Our study found a significant gender gap in psychological morbidities. Females had significantly higher anxiety levels (34.6% vs. 22.9%) and depression (32.4% vs. 26.0%) than males. This finding was consistent with previous studies [24,63,64]. The gender differences may be attributed to a higher prevalence of pre-existing mental health conditions among females

than males, complicated by introversion, higher sensitivity to traumatic events, and other factors, including hormonal imbalances and genetic vulnerability, and a higher mental health stigma among men [64–66]. Additional evidence reported that it is likely that mental health issues among men are underreported because of their tendency not to seek help from others [67].

We found a wider variation while making country comparisons. Anxiety and depression reported out of Asian countries were lower compared to other countries. Traditional close-knit family systems in Asia can be a protecting factor overriding one significant risk factor of social isolation, which has shown to contribute to increased risk of mental health issues [66]. Additionally, Asian countries, especially China and India, have traditional medicine with products and services widely available that are acceptable, affordable, and culturally appropriate. Most importantly, these have been adopted by the various Asian countries' health care systems [68]. However, the efficacy of traditional medicine has not been fully proven in counteracting mental health problems.

### *4.1. Strengths and Limitations*

This meta-analysis is the first to assess the psychological impact of COVID-19 among students. It is urgent and essential to know the global scope of the issue. This population group is already facing a disproportionate burden of psychological morbidities even before the pandemic. This study also has some limitations. First, the self-reporting nature of the data collected by the studies in our meta-analysis might not be an accurate representation of the clinical diagnosis of the psychological illness. Second, sampling bias may exist because nearly 66.6% (18 out of 27) of the studies were conducted in Asia and predominantly China (51.8%; 14 out of 27). The larger pool of studies from China may presumably be due to the greater interest of the Chinese researchers in unfolding the epidemiology of COVID-19, as China was the first country to be affected by COVID-19. Other countries might have other research priorities prior to the pandemic inception, which occurred two months following the pandemic emergence in China. Third, all studies included in this metaanalysis were cross-sectional, which only account for prevailing circumstances, thereby lacking a longitudinal aspect to encounter temporality. Fourth, the studies included were only published in the English language, which might have introduced a language bias. Last, most of the studies included in this meta-analysis did not provide the year-wise, program (undergraduate/postgraduate), and type of course (e.g. STEM vs. non-STEM) stratifications of the students, which restricted our ability to determine differences in psychological morbidities among these groups. Future studies can be designed to account for differences in psychological outcomes across different groups of students to design a more targeted interventional approach.
