key: cord-1014147-199hlhjo
authors: Berman, Noah Chase; Fang, Angela; Hoeppner, Susanne S.; Reese, Hannah; Siev, Jedidiah; Timpano, Kiara R.; Wheaton, Michael G.
title: COVID-19 and obsessive-compulsive symptoms in a large multi-site college sample
date: 2022-05-02
journal: J Obsessive Compuls Relat Disord
DOI: 10.1016/j.jocrd.2022.100727
sha: b0a1ee41aac84abda89bab9181dc1698b7c61e4a
doc_id: 1014147
cord_uid: 199hlhjo

The COVID-19 pandemic poses unique risks to college students’ mental health, and specifically to symptoms of obsessive-compulsive disorder (OCD). To better understand the relationship between COVID-19 impact and OC symptoms in this population, six colleges from across the US administered a battery of questionnaires and an emotion differentiation paradigm to eligible students (N = 841). We examined whether degree of pandemic-related disruption was associated with OC severity, and if so, whether this relationship was explained by trait (poor emotion regulation and differentiation) and state risk factors (poor sleep quality, less exercise frequency, less social support, thwarted sense of belongingness, and greater loneliness). Results indicated that the positive relationship between COVID-19 impact and OC severity was mediated by trait emotion-related processes (e.g., emotion regulation and differentiation), but no state risk factors emerged as significant mediators. Our findings contribute to the literature demonstrating a significant relationship between COVID-19 impact and OC severity, and highlight that emotion regulation difficulties may help explain this association. Our findings can inform evidence-based interventions on college campuses; however, the cross-sectional design precludes causal inferences. Future research should evaluate these relationships longitudinally and incorporate other psychosocial factors that may operate as mechanisms.

. Despite growing interest in the relationship between COVID-19 and OC symptoms, only a few studies have addressed this link among college students (e.g., Dixon et al., 2021; Ji et al., 2020; Jiang, 2020; Zhang et al., 2021) and no studies have examined this link across multiple universities in the U.S. Given the variability in COVID-19 responses from local and state governments, research must elucidate how COVID-19 relates to OC symptoms beyond a single university. In this vulnerable population, the relationship between COVID-19 impact and OC symptoms must be clarified, and the psychological risk factors associated with OC symptom severity should be identified. Fittingly, leaders in the field (Grant et al., 2022) have called for cross-sectional studies that use general population samples (where most do not have OCD, but some might) to provide insight into the dimensional mechanisms (e.g., emotion dysregulation) by which the pandemic may relate to OC symptoms. Such knowledge can serve as the foundation for future research on university-based prevention and intervention programs that work to minimize or prevent adverse psychological outcomes due to the ongoing pandemic.

Toward this end, our broad objective was to examine psychological trait and state risk factors associated with OC symptoms in college students during the COVID-19 pandemic. In terms of trait risk factors, we expected that emotional processing difficulties, specifically poor emotion regulation and emotion differentiation would be associated with more severe OC symptoms. Our hypothesis rests upon empirical evidence that difficulties in multiple aspects of 6 COVID-19 IN COLLEGE outcomes and internalizing psychopathology in adolescents and young adults (Erbas et al., 2014; Nook et al., 2018; Nook, 2021) . Emotion regulation and differentiation deficits may help explain why some college students struggle with OC symptoms during COVID-19. For instance, a student who is bombarded by messaging about the dangers of COVID-19 may be unable to differentiate the resultant emotional experience (e.g., cannot disentangle the sadness caused by her isolation, from the guilt of potentially infecting her loved ones, from the fear of contracting the illness herself, from the anger at her peers who are ignoring mandated safety measures). If she cannot disentangle her internal experience, then she would not be able to implement emotion-specific skills that effectively delay impulsive urges or facilitate acceptance of unwanted emotional experiences. Consequently, she may rely on ritualistic behaviors, like excessive handwashing, to regulate her emotional arousal.

In terms of state risk factors, we expected that five psychological constructs would be positively associated with OC symptom severity, based on the unique challenges that the pandemic imposed on college students' lives. Maladaptive health behaviors, such as poor sleep and lack of physical exercise, are considered general risk factors for mental health in college students (Ghrouz et al., 2019) . Findings suggest that these health behaviors are also relevant risk factors during the COVID-19 pandemic, as: (a) individuals who experienced worsening OC symptoms reported greater sleep disturbances, when compared to those who did not experience worsening (Benatti et al., 2020) , (b) pre-pandemic insomnia symptoms predicted OC symptoms 7 COVID-19 IN COLLEGE 2021; Zhang et al., 2021) . In addition to maladaptive health behaviors, social state-based risk factors, such as lack of social support, lack of belongingness, and loneliness, have also been associated with poor health outcomes (Leigh-Hunt et al., 2017; Palgi et al., 2020) and may be important risk factors for OC severity. Accordingly, a lack of perceived social support and being single has been linked to an increased risk of OCD during the pandemic (Alonso et al., 2021; Zheng et al., 2020) . Recent research also suggests that thwarted belongingness (e.g., fundamental need for connectedness is not being met; Cacioppo & Patrick, 2008) and loneliness may underlie suicide risk during COVID-19 (Gratz et al., 2020; Raj et al., 2021) .

Based upon the extant literature, our team of investigators generated two hypotheses regarding the relationship between COVID-19 and OC symptoms. First, we hypothesized that greater impact from COVID-19 (operationalized as higher scores on our COVID-19 Impact Measure) would be positively associated with OC symptom severity in college students from across the country. Second, we hypothesized that the positive association between COVID-19 impact and OC symptom severity would be explained by trait (poor emotion regulation and differentiation) and state risk factors (poor sleep quality, less exercise frequency, less social support, thwarted sense of belongingness, and greater loneliness). We addressed past limitations by adopting a multi-modal approach (i.e., questionnaires and paradigms), and recruiting participants across six academic institutions in the United States (US) during the pandemic.

Data collection occurred at six higher education institutions (four liberal arts colleges and two universities) in the US. Participants were recruited from the Introduction to Psychology course at each institution and inclusion criteria were broad -at least 18 years of age, fluent in J o u r n a l P r e -p r o o f 8 COVID-19 IN COLLEGE English, and willing to provide informed consent. In total, 841 students participated in the current study. Larger samples were recruited from the two universities (n's = 147 and 235) than the four liberal arts colleges (range of n's = 73 to 176; site-specific demographic information can be found in Table 1 ). Students were allowed to enroll in the research study throughout the entire fall semester, with data collection occurring between October 2 to December 16, 2020.

To assess risk factors, as well as OC symptoms, we administered a battery of psychometrically valid and reliable self-report questionnaires and paradigms. Additionally, we designed a measure assessing the impact of COVID-19 on participants' psychological health.

To characterize the sample, we assessed the following characteristics: age, biological sex, gender identity, race, ethnicity, sexual orientation, and academic year (i.e., first-year, secondyear, etc.). We also assessed where students resided for Fall 2020, who they lived with, and their ideal academic arrangement for the term (e.g., complete virtual learning; a hybrid approach). We used the single-item MacArthur Scale of Subjective Social Status -Adult Version (Adler et al., 2000) to assess subjective social status, a validated alternative to more objective indices, such as education or income. Participants marked their perceived social status on a ladder with 10 rungs, 9 COVID-19 IN COLLEGE This 14-item self-report measure assesses the impact of the COVID-19 pandemic on psychological health. We modified a previous version of this questionnaire (Fang et al., 2021) to be relevant for college students. Respondents rated the impact of COVID-19 on their everyday lives (e.g., "I found it hard to concentrate on work and/or leisure activities because I was preoccupied with thoughts about COVID-19") on a Likert scale from 1 ("Strongly Disagree") to 5 ("Strongly Agree"), with higher scores reflecting greater impact. The factor structure of the COVID-19 Impact Measure, and additional psychometrics, are reported in the Data Preparation section. In addition to the 14 quantitative ratings, participants also reported their degree of personal exposure (e.g., tested positive; hospitalized) to COVID-19 since January 2020, as well as that of their roommates and close friends/family. (DOCS; Abramowitz et al., 2010) This 20-item self-report measure assesses four dimensions of OCD: contamination, responsibility for harm, unacceptable thoughts, and symmetry/"just right" phenomena. For each dimension, respondents rate: (a) time occupied by obsessions and rituals, (b) avoidance behavior, (c) distress, (d) functional interference, and (e) difficulty disregarding obsessions and resisting compulsions. In both clinical and college student samples, the DOCS has shown strong internal consistency, test-retest reliability, convergent validity, divergent validity, sensitivity to treatment, and diagnostic sensitivity (total score ≥ 18 differentiates individuals with OCD from students; 10 COVID-19 IN COLLEGE behavior and refrain from impulsive behavior when experiencing negative emotions, and access to emotion regulation strategies perceived as effective. Each item (e.g., "I'm confused about how I feel") is rated from 1 ("Almost Never) to 5 ("Almost Always"). For the analyses, we used DERS-SF total scores, where higher scores reflect greater emotion dysregulation. This measure has demonstrated good internal consistency, convergent, and divergent validity in both adolescent and college student samples (Kaufman et al., 2016) . Cronbach's α in the current sample was 0.89. Thirty-five participants (4.2%) had >20% missing data for this measure. (Marshall et al., 2005) To assess respondent's exercise behaviors, participants were asked: "[Since the beginning of the academic term], about how many times a week, have you done 30 minutes (or more) of moderate physical activity that makes you breathe harder than normal? (for example, jogging, heavy lifting, aerobics, or fast bicycling)." Participants then chose from the following: 0 ("None"), 1 ("1-2 times/week"), 2 ("3-4 times/week"), or 3 ("5 times/week or more"). This tool has good reliability and convergent validity with accelerometer measurements of physical activity (Marshall et al., 2005) . It has been widely used in clinical and college student samples (e.g., Zimmerman et al., 2021) . Thirty-eight participants (4.5%) had missing data for this item. (Van Orden et al., 2012) This 15-item questionnaire assesses thwarted belongingness and perceived burdensomeness; however, only the 9-item belongingness sub-scale was used in this study. Each sub-scale item (e.g., "These days, I rarely interact with people who care about me") is rated from 11 COVID-19 IN COLLEGE al., 2012) and Cronbach's α in the current sample was 0.90. Thirty-six participants (4.3%) had >20% missing data for this measure. (MSPSS; Zimet et al., 1988) The MSPSS is a 12-item self-report scale measuring support from family, friends, and significant others. Each item (e.g., "I get the emotional help and support I need from my family") is rated on a 7-item Likert scale from 1 ("Very Strongly Disagree") to 7 ("Very Strongly Agree"). The total score (sum of all items) was used in analyses. The MSPSS has sufficient construct validity, test-retest reliability, and internal consistency (Zimet et al., 1988) and is widely used in college samples (e.g., Lopez-Castro et al., 2021) . Cronbach's α in the current sample was 0.92. Thirty-seven participants (4.4%) had >20% missing data for this measure.

To assess participants' ability to differentiate their own negative emotion states, we used a paradigm that parallels Boden et al. et al., 1995) and before each image appeared, participants saw a fixation cross in the center of the screen for two seconds. Each of the 10 images then appeared on the screen for six seconds and participants were told to "look directly at the picture." The order of the images was randomized to minimize potential order effects. Following each image, participants rated the intensity of the following emotions: fear, worried, lonely, sad, guilty, ashamed, jealous, embarrassed, angry, disgusted on a scale from 1 ("Not at all felt") to 6 ("Extremely felt"). A negative emotion differentiation index was then calculated using the average intraclass correlation (ICC (3,k); 12 COVID-19 IN COLLEGE that ratings for different emotions across pictures were strongly correlated and without notable distinctions (i.e., low emotion differentiation). Thus, paralleling past research, we reverse-scored the ICC values (i.e., 1 -ICC), so that larger values correspond to greater emotion differentiation among negative emotions. For this measure, we excluded observations from participants with more than 20% missing picture ratings (n=41; 4.9%), participants with no variance between any emotion ratings (n=2; 0.2%), and participants who had negative ICC scores (n=187; 22.1%), because negative scores are beyond the theoretical lower limit for the ICC. (Snyder et al., 2018) Based upon the single-item sleep quality scale (Snyder et al., 2018) , participants were asked to "rate the quality of [your] sleep" on a scale of: 0 ("Very Poor"), 1 ("Poor"), 2 ("Fair"), 3 ("Good"), and 4 ("Very Good"). The SQS has demonstrated adequate test-retest reliability, convergent and divergent validity, and responsivity to treatment (Snyder et al., 2018) . Thirtyeight participants (4.5%) had missing data for this item. (Hughes et al., 2004) This 3-item self-report scale assesses respondents' degree of loneliness. Each item (e.g., "How often do you feel isolated from others?") is rated from 1 ("Hardly Ever") to 3 ("Often"), with higher scores reflecting greater loneliness. This shortened version is unidimensional, possesses adequate internal consistency, convergent, and divergent validity, was developed specifically for large scale surveys (Hughes et al., 2004) , and has been widely used in clinical, community, and college student samples (e.g., Vaterlaus, 2022) . Cronbach's α in the current sample was 0.82. Thirty-six participants (4.3%) had missing data for this measure.

Individuals at all six sites were enrolled in Introduction to Psychology and were given the opportunity to participate in our study examining the relationship between COVID-19 and psychological difficulties. Interested participants were sent a unique link to a Qualtrics survey.

The same online questionnaire battery was administered to participants at all sites and was approximately 45-minutes in duration. At study completion, participants were given a debriefing form and compensated for their time. Each institution and/or instructor of Introduction to Psychology used a different approach for research compensation (e.g., lowest quiz dropped, extra credit), so compensation differed site to site. All measures and procedures were reviewed and approved by each site's local institutional review board (IRB).

To explore the factorial structure of our COVID-19 Impact Measure, we subjected all 14 items of the instrument to an exploratory factor analysis with oblique rotation (promax). Eight hundred and sixteen participants had complete data for all scale items and were used in this analysis. The Kaiser-Meyer-Olkin measure verified the sampling adequacy for the analysis, KMO = .84. Bartlett's test of sphericity Χ 2 (df = 91) = 4801.29, p < .0001, indicated that the correlation structure was adequate for factor analyses. Velicer's MAP analysis (O'Connor, 2000; Velicer et al., 2000) indicated that a two-factor solution was the best fit for the data. Initial eigenvalues indicated that the first two factors explained 77.84% and 17.37% of the common variance, respectively. The two factors were: (a) COVID-19 apprehension (fear, anxiety, and uncertainty) with 7 items; and (b) pandemic social distress (loneliness, stress, mental health) with 5 items. Two items (8 and 9) had no clear factor loadings on either factor (Supplementary Table   S1 ) and were excluded from the final factor scores. The inter-factor correlation was moderate Supplementary Table S1 .

Both factors 1 (α =0.80) and 2 (α =0.72) possessed acceptable internal consistency.

For all questionnaires, we calculated scale scores using within-person mean replacement (Downey & King, 1998; Shrive et al., 2006) if no more than 20% of the scale items were missing; for scales with existing sub-scales, we imputed scale scores at the sub-scale level before totaling them into total scores. In most cases of missing data in questionnaires, scale items were completely missing and less than 1% of the scores used mean-item replacement.

To test whether COVID-19 impact scores (COVID-19 apprehension sub-scale and social distress sub-scale) were associated with DOCS scores, we used generalized linear mixed models with COVID-19 impact scores as the independent variables (fixed effect) in separate models.

Both models included site as a random effect and DOCS total scores as the outcome. We used a z-score transformation of the COVID-19 impact scores in both models, so that the model coefficients could be interpreted as the expected change in raw DOCS scores with an increase of one standard deviation in the COVID-19 impact sub-scale scores.

To investigate which risk factors mediated the association between either of the two COVID-19 impact sub-scales and DOCS scores, we used a multiple mediation model using ordinary least squares path analysis with either of the COVID-19 impact sub-scale scores as the independent variable (in separate models), our hypothesized risk factors (i.e., thwarted belongingness, loneliness, perceived social support, difficulties with emotion regulation, negative emotion differentiation, sleep quality, and exercise frequency) as the mediators, and DOCS total score as the outcome. Preliminary review revealed, however, that our measure of J o u r n a l P r e -p r o o f 15 COVID-19 IN COLLEGE negative emotion differentiation had substantially less usable data than our other hypothesized risk factors, due to the number of negative ICC values that needed to be excluded (i.e., only n=610 possessed complete data). Because the mediational model only uses scales with no missing data, including this variable in the full model would have substantially reduced the available data for every other risk factor, thereby limiting our power. Therefore, we conducted two parallel mediational models. The first model (n = 800) included six of our seven hypothesized risk factors as mediators (i.e., thwarted belongingness, loneliness, perceived social support, difficulties with emotion regulation, sleep quality, and exercise frequency; Figure 1 ) and the second model (n = 610) included only negative emotion differentiation as a mediator ( Figure   2 ). We then repeated the mediation models with each of the DOCS sub-scales. We intended these additional mediation models to be hypothesis-generating explorations of the different OCD symptom dimensions and did not adjust them for family-wise error. In all mediation models, site was a covariate, and the significance of indirect effects was evaluated through percentile confidence intervals based on 10,000 bootstrap samples. All variables in the mediation models were also standardized using a z-score transformation to aid in the interpretation of path coefficients. The mediation analysis was performed using the PROCESS macro (Hayes, 2013) in SAS 9.4 for Windows. Results were interpreted as statistically significant when p < 0.05 or bootstrap confidence intervals of effect estimates excluded zero. to a hybrid learning model (n=399, 49.0%), while less hoped for a complete transition to virtual learning (n=212, 26.0%), and a few wanted their institution to "do something else" (n=3, 0.4%).

Roughly half of participants lived away from campus (n=432, 53.1%), though this percentage varied widely between schools ( Table 1 ). The majority of those who lived away from campus reported living with their parents or primary caregivers (n=402, 93.3%). The majority of students perceived themselves to be of medium (53.7%) or high (34.8%) social standing, with only 11.6%

reporting comparatively low social standing. Given the geographic variability, composition of each school, and type of higher education (university vs. liberal arts), we expectedly observed demographic differences between sites, as well as site differences in mean DOCS scores, COVID-19 impact, and various risk and resilience factors reported by students (Table 1 ).

The majority of students did not report being personally infected with the COVID-19 virus; only 56 participants (6.7%) reported testing positive for COVID-19 since January 2020 and only 2 (0.2%) had been hospitalized. However, 413 participants (49.1%) knew of at least one person they lived with, or a close friend or family member they did not live with, who tested positive, needed hospitalization, or died of COVID-19 (Table 1) . Of note, the average DOCS total score at all six sites exceeded 18, which is the cut-off score for differentiating OCD from student samples (Abramowitz et al., 2010) .

There were significant associations between the two COVID-19 impact sub-scales and DOCS total scores. The association was stronger for the COVID-19 apprehension sub-scale than for the social distress sub-scale. A one standard deviation increase in COVID-19 apprehension 

As shown in Table 2 , higher COVID-19 apprehension sub-scale scores and social distress sub-scale scores were significantly associated (p ≤ .004) with worse outcomes for all but one of the hypothesized mediators (perceived social support), but only one mediator had a significant association with DOCS total scores. Specifically, only the DERS-SF had a significant association with DOCS total scores, such that a one standard deviation increase in DERS-SF total scores was associated with one third of a standard deviation higher DOCS total score in either COVID-19

impact sub-scale model ( Table 2) . The multiple mediation model with COVID-19 apprehension sub-scale scores as the predictor accounted for about 35% of the variation in DOCS total scores, and the direct association between the COVID-19 apprehension sub-scale on DOCS total scores was greater than that of the only significant mediation pathway (82.9% vs. 14.5%). The multiple mediation model with the social distress sub-scale scores as the predictor accounted for about 27% of the variation in DOCS total scores, and the direct association between the social distress sub-scale on DOCS total scores was only slightly greater than that of the only significant mediation pathway (48.3% vs. 39.3%). In the simple mediation model, negative emotion differentiation emerged as a significant mediator only between the COVID-19 apprehension subscale and DOCS total scores, with the whole model explaining about 17% of the variation in DOCS total scores. In this model, higher COVID-19 apprehension was associated with decreased negative emotion differentiation (a-path; p = .0029), while higher (i.e., more successful) negative emotion differentiation was associated with a lower severity of DOCS total scores (b-path; p = J o u r n a l P r e -p r o o f 18 COVID-19 IN COLLEGE .0158). The net (indirect) effect of this mediation pathway was that the association of COVID-19 apprehension with negative emotion differentiation accounted for approximately 3.1% of the total association between COVID-19 apprehension and DOCS total scores (Table 2) . We were unable to detect an association between the COVID-19 social distress sub-scale and negative emotion differentiation (p = .1169).

These mediation models were repeated with each DOCS sub-scale as an individual outcome, and the results were very similar to the DOCS total score model (Supplementary   Tables S2-S5 ). This is not surprising since the correlations of the individual DOCS sub-scales with the total score were expectedly high (ranging from 0.73 [contamination sub-scale] to 0.85 [responsibility for harm sub-scale]). In these exploratory models, DERS-SF was the only consistent mediator of both COVID-19 impact sub-scales, though negative emotion differentiation, exercise frequency and loneliness sporadically emerged as weaker secondary mediators in models. Overall, the models with the COVID-19 apprehension sub-scale as the predictor explained greater variation in DOCS scales than models using the COVID-19 social distress sub-scale as the predictor. We were unable to detect any direct effects of thwarted belongingness, perceived social support, or sleep quality on any DOCS outcomes in any model.

In this study, researchers at six higher education institutions (two large public universities and four liberal arts colleges) investigated the association between COVID-19 impact and OC symptom severity. As expected, college students who reported greater COVID-19-related apprehension and social distress (i.e., greater pandemic-related disruption) also reported more OC symptoms. This finding was relatively large in magnitude and remained significant after accounting for site effects. Our pattern of results aligns with research demonstrating a positive J o u r n a l P r e -p r o o f 19 COVID-19 IN COLLEGE relationship between pandemic fears and OC symptoms (e.g., Wheaton et al., 2012) , as well as with research on the relationship between COVID-19 and OC symptomology in some clinical samples (e.g., Benatti et al., 2020; Wheaton et al., 2021) and college students at specific universities (Dixon et al., 2021; Ji et al., 2020; Jiang 2020; Zhang et al., 2021) . Thus, using a heterogenous sample of American college students from across the country, the present results add to our field's understanding of the relationship between the COVID-19 pandemic and OC symptoms in a uniquely vulnerable population (Arnett et al., 2014) .

Supporting our first hypothesis, positive associations emerged between OC symptom severity and both the apprehension and social distress sub-scales of the COVID-19 Impact Measure. However, the relationship between the COVID-19 apprehension sub-scale and OC outcomes was stronger than the observed relationships between the social distress sub-scale and OC symptom severity. This pattern aligns with recent research demonstrating that pandemicrelated stress and fear (i.e., the foundation of our apprehension sub-scale) is associated with elevated OC symptoms in the general population (Albertella et al., 2021; Fontenelle et al., 2021; Seçer and Ulaş, 2020) . Moreover, past research has consistently linked elevated stress levels to heightened OC symptoms (e.g., Coles et al., 2008; Vidal-Ribas et al., 2015) . The resultant stress due to COVID-19 may help explain why both sub-scales (apprehension and social distress) were positively associated with overall OC symptom severity, as well as with most OC dimensions (contamination, responsibility for harm, unacceptable thoughts, and symmetry/"just right" concerns). Although the COVID-19 social distress sub-scale possessed significant associations with OC outcomes, we anticipate that this sub-scale would be more strongly associated with depressive symptoms. Aligning with the loss of reinforcement theory (Lewinsohn, 1974) , the items assessing pandemic-related social distress (e.g., isolation; missed important social J o u r n a l P r e -p r o o f 20 COVID-19 IN COLLEGE opportunities, like a wedding) reflect a reduction in response-contingent positive reinforcement, which could ultimately increase dysphoria and depressive symptomology.

Aligning with predictions, there were significant associations between COVID-19 impact and nearly all of the individual vulnerability factors. Indeed, individuals who reported greater COVID-19 apprehension and social distress also reported worse sleep quality, getting less exercise, more loneliness and thwarted belongingness, and worse emotion regulation (e.g., Alonso et al., 2021; Ghrouz et al., 2019; Leigh-Hunt et al., 2017; Palgi et al., 2020; Zheng et al., 2020) . Somewhat surprisingly, neither COVID-19 apprehension nor social distress was significantly associated with perceived social support. One interpretation for this non-significant finding is that our measure of this construct (MSPSS; Zimet et al., 1988) did not assess social support via social media or networking sites, which was a predominant source of communication among college students during the pandemic (Zhen et al., 2021) . As a result, we may not have tapped into the true relationship between COVID-19 impact and social support.

The present study also highlighted meaningful relationships between our examined vulnerability factors and OC symptom severity. Of the indices investigated, impaired emotion regulation was the only factor that consistently related to overall OC symptom severity (and all four OCD dimensions in our exploratory analyses). These findings parallel past research demonstrating that individuals with deficits in emotional awareness and affect regulation also report greater OC symptom severity (Berman et al., 2018; Khosravani et al., 2017; Roh et al., 2007; Yap et al., 2018; Yazici & Yazici, 2019) . Partially supporting our second hypothesis, the mediational models indicated that both of our emotion-related vulnerability factors (emotion dysregulation and negative emotion differentiation) accounted for a significant portion of the direct effect between COVID-19 apprehension and overall OC symptom severity. It may be that J o u r n a l P r e -p r o o f 21 COVID-19 IN COLLEGE students who had difficulty differentiating the negative emotions provoked by pandemic-related apprehension (e.g., feeling frightened of becoming sick) may have relied upon maladaptive emotion regulation strategies (e.g., ritualistic behaviors) to regulate their undifferentiated emotional distress. Interpreted in the other direction, individuals who successfully identified their emotion states (e.g., guilt), and could then implement emotion-specific regulatory strategies (e.g., affect labeling; opposite action; problem-solving; Linehan, 2014), may have been less likely to turn toward maladaptive short-term regulatory behaviors that were ritualistic in nature (e.g., compulsive washing routines). Further mechanistic research is needed to test out these effects; however, the current findings do align with evidence that emotional intelligence mediated the relationship between pandemic exposure and the severity of psychological disorders in college students (Li, Li, & Fan, 2021) . In further support of this effect, Hong and colleagues (2022) reported that a poor understanding of one's affective experience (i.e., limited emotion differentiation abilities), non-acceptance of one's emotions, and trying to control impulses in the face of high affect (i.e., emotion dysregulation) helped explain the relationship between healthrelated concerns during the pandemic and the severity of OC symptoms.

The present results may have important clinical implications. Emerging work clearly suggests that the pandemic is having deleterious mental health effects on college students, including on their OC symptom severity. Fortunately, effective treatments exist for OCD, including both serotonin reuptake inhibitor (SRI) medications and cognitive-behavioral therapy (Koran et al., 2007) . Therefore, it is important that these treatments are available to students experiencing clinically significant OC symptoms as a result of the pandemic. Our findings also suggest that interventions aimed at improving emotion regulation (e.g., Affect Regulation (2021) recently demonstrated that using adaptive coping strategies during the pandemic, specifically future-oriented planning (e.g., identify the necessary next steps to manage difficult events; Carver, 1989) may interrupt the emotion emergence process, and decrease both emotion regulation difficulties and internalizing symptoms.

The present findings should be interpreted in light of several important study limitations.

First, although the current study included a geographically diverse set of schools from across the country, the present sample may not generalize to all college students. The participating schools were not randomly selected, and all students were selected from psychology classes at academically rigorous four-year institutions. Further study is needed to generalize to other college settings (e.g., community colleges). Moreover, given the disproportionate mental health impact of COVID-19 on communities of color, future research should examine whether the reported associations between COVID-19 impact and OC symptom severity persist in a more representative, diverse sample of U.S. college students.

Second, data were collected at a single time point in a cross-sectional design. A longitudinal framework would be needed to establish that participants' OC symptoms increased following the emergence of COVID-19 (as some participants' OC symptoms may have preceded the pandemic). Moreover, a longitudinal design would help elucidate the progression and course of college students' OC symptoms over time. As previously described, participants' self-reported OC symptom severity exceeded the threshold that distinguishes past student samples from those with clinically significant OCD. Notably, this pattern aligns with recent research demonstrating that a year into the pandemic, individuals considered to be "healthy controls" reported more severe internalizing symptoms compared to their pre-pandemic levels (Kok et al., 2022) . A J o u r n a l P r e -p r o o f 23 COVID-19 IN COLLEGE longitudinal design would identify whether this elevation reflects (a) a temporary (and potentially adaptive) increase in obsessive pre-occupation and ritualizing that dissipates as restrictions and pandemic-fears subside (Grant et al., 2022) or (b) a more enduring pattern of elevated OC symptoms in college students that may not naturalistically remit as the pandemic recedes (bringing implications for large-scale interventions at the community-level).

A third limitation relates to the emotion differentiation paradigm. Given the number of negative ICC scores on this paradigm, including this variable in the multiple mediation model would have substantially reduced the available data. Therefore, we ran a separate simple mediation model to test whether emotion differentiation explained the relationship between COVID-19 impact and OC symptom severity. Despite the fact that emotion differentiation was not in competition with other risk factors, this model does highlight that this trait risk factor significantly explains the relationship between some pandemic-related disruption and OC symptom severity. That being said, we encourage researchers to further investigate this paradigm to determine "best practices" for managing negative ICC scores, and whether online administrations (i.e., without supervision) can yield valid and reliable outcomes. A fourth limitation is that our mediation models only explained part of the variance between COVID-19 impact and OC symptom severity, highlighting the need to identify other explanatory variables (e.g., cultural orientation; identity factors) in future research. COVID-19 IN COLLEGE Note. To calculate the proportion of respondents with a specific demographic factor (e.g., % female), the available data from each site is used as the denominator, and this may differ from the total sample listed. Table 2 . Path and effect estimates for the multiple (n=800) and simple mediation (n=610) models examining the direct and indirect associations between COVID-19 Impact and OCD symptom severity.

Impact and mediator (a paths)

Relationship between mediator and DOCS total symptom severity (b paths)

Paths ( n/a n/a n/a n/a n/a n/a n/a n/a c 0. n/a n/a n/a n/a n/a n/a n/a n/a c 0. n/a n/a n/a n/a n/a n/a n/a n/a c 0. 3.1 COVID-19 Impact n/a n/a n/a n/a n/a n/a n/a n/a c 0.240 [0.163, 0.317] 96.9

Note: a Confidence intervals based on 10,000 percentile bootstrap samples were used for indirect effects (ab paths); Bolded areas reflect statistically significant effects. Cross-sectional multiple mediation model examining the association between the COVID-19 Impact apprehension and social distress sub-scales and overall OC symptom severity (DOCS total scores), mediated by six potential mechanisms: thwarted belongingness, loneliness, perceived social support, difficulties in emotion regulation, sleep quality, and exercise frequency.

Notes: Solid lines represent significant associations, dotted lines represent non-significant associations in the two multiple mediation models that use different COVID-19 Impact sub-scale scores as predictors. For the site effect, solid lines indicate that at least one site has a significantly different effect (p < .05) compared to the largest site (#4). Path estimates for specific models are presented in Table  2 . Cross-sectional mediation model examining the association between the COVID-19 Impact apprehension or social distress sub-scales and overall OC symptom severity (DOCS total scores), mediated by negative emotion differentiation in the sub-sample with valid negative emotion differentiation scores (n=610).

Notes: Solid lines represent significant associations in at least one mediation model, dotted lines represent non-significant associations in both mediation models. For the site effect, solid lines indicate that at least one site has a significantly different effect (p<.05) compared to site 4 in at least one mediation model. Specific path estimates for each model are presented in Table 2 .

Physical activity and sleep quality in relation to mental health among college students

Obsessive-compulsive symptoms and the COVID-19 pandemic: A rapid scoping review

Thwarted belongingness and perceived burdensomeness explain the associations of COVID-19 social and economic consequences to suicide risk

When social isolation is nothing new: A longitudinal study on psychological distress during COVID-19 among university students with and without preexisting mental health concerns

Introduction to mediation, moderation, and conditional process analysis: A regression-based approach

How health anxiety affected obsessive-compulsive symptoms during the COVID-19 pandemic in China: The mediation of difficulties in emotion regulation and the moderation of pathological personality traits. Personality and 30 COVID-19 IN COLLEGE

Reliability and validity of a brief physical activity assessment for use by family doctors

Acute impact of COVID-19 pandemic on phenomenological features in fully or partially remitted patients with obsessivecompulsive disorder

Emotion differentiation and youth mental health: Current understanding and open questions

The nonlinear development of emotion differentiation: Granular emotional experience is low in adolescence

SPSS and SAS programs for determining the number of components using parallel analysis and Velicer's MAP test

The loneliness pandemic: Loneliness and other concomitants of depression, anxiety and their comorbidity during the COVID-19 outbreak

Secondary psychometric examination of the dimensional obsessive-compulsive scale: Classical testing, item response theory, and differential item functioning

Thwarted belongingness and perceived burdensomeness: Construct validity and psychometric properties of the Interpersonal Needs Questionnaire

College student loneliness and the reopening of campuses during the covid-19 pandemic. Family and Consumer

Construct explication through factor or component analysis: A review and evaluation of alternative procedures for determining the number of factors or components

Are stressful life events causally related to the severity of obsessive-compulsive symptoms? A monozygotic twin difference study

How is the COVID-19 pandemic affecting individuals with obsessive-compulsive disorder (OCD) symptoms?

Emotion regulation difficulties in obsessive-compulsive disorder

Decreased theory of mind skills, increased emotion dysregulation and insight levels in adolescents diagnosed with obsessive compulsive disorder

Risk factors of psychological responses of Chinese university students during the COVID-19 outbreak: Cross-sectional web-based survey data

College students coping with COVID-19: Stress-buffering effects of self-disclosure on social media and parental support

The Multidimensional Scale of Perceived Social Support

Initial impact of the COVID-19 pandemic on college student mental health: A longitudinal examination of risk and protective factors

Risk Factors Thwarted belongingness, M(SD) 30.1 (10.8)

UCLA-LS-3 total, M(SD)

MSPSS total, M(SD)

DERS-SF total, M(SD)

M(SD)

Sleep Quality, M(SD)

Exercise Frequency, M(SD)