Data and alternative models describing the associations among non-infection pandemic stress, event-related rumination, depression, and anxiety

Here we present cross-sectional data collected from 1507 participants through the Qualtrics online survey platform. Participants were recruited from Reddit, Facebook, and the Queen's University undergraduate participant pool, and were instructed to complete a pandemic stress survey, the Beck Depression Inventory-II (BDI-II) [1], the Beck Anxiety Inventory (BAI) [2], a modified version of Event-Related Rumination Inventory (ERRI) [3], and a demographics questionnaire. For the 1069 participants who were not exposed to COVID-19 infection, we calculated the sum of each scale/subscale and performed a multiple mediation analysis using MPlus. The results indicated that three models (one primary model and two alternative models) had comparable statistical power to explain the variance as we tested different configurations of predictor, mediator, and outcome variables. Given the cross-sectional nature of the present study, we could not conclude which model was most valid. Therefore, we share our original data and tested models here for others to use. They are useful for researchers who wish to replicate our results, conduct new analyses with these data, or design future studies.


a b s t r a c t
Here we present cross-sectional data collected from 1507 participants through the Qualtrics online survey platform. Participants were recruited from Reddit, Facebook, and the Queen's University undergraduate participant pool, and were instructed to complete a pandemic stress survey, the Beck Depression Inventory-II (BDI-II) [1] , the Beck Anxiety Inventory (BAI) [2] , a modified version of Event-Related Rumination Inventory (ERRI) [3] , and a demographics questionnaire. For the 1069 participants who were not exposed to COVID-19 infection, we calculated the sum of each scale/subscale and performed a multiple mediation analysis using MPlus. The results indicated that three models (one primary model and two alternative models) had comparable statistical power to explain the variance as we tested different configurations of predictor, mediator, and outcome variables. Given the crosssectional nature of the present study, we could not conclude which model was most valid. Therefore, we share our original data and tested models here for others to use. They are useful for researchers who wish to replicate our results, conduct new analyses with these data, or design future studies. © 2022 Published by Elsevier Inc. This is an open access article under the CC BY license ( http://creativecommons.org/licenses/by/4.0/ ) Table   Subject Clinical Psychology Specific subject area Pandemic stress, internalizing psychopathology, and event-related rumination Type of data Table  Figure How the data were acquired Data were collected with an online survey comprised of the Beck Depression Inventory-II (BDI-II) [1] , the Beck Anxiety Inventory (BAI) [2] , a modified version of the Event-Related Rumination Inventory (ERRI) [3] , and a questionnaire of pandemic stress. We used the Qualtrics online survey platform to collect and filter these data.

Value of the Data
• Due to the cross-sectional nature of these data, the alternative models are useful for considering other, equally plausible explanations of our data besides the main model we explored in Squires et al. [5] . • Researchers working on the dissociation of intrusive versus deliberate rumination may benefit from these data since our models showed a consistent dissociation between intrusive and deliberate rumination no matter where they are located in the model. • Our data may also benefit scholars who are interested in the psychological impact of public health crises more broadly. These data may be compared to and integrated with other COVID-19, pandemic-related, or other public health datasets. • Researchers who would like to conduct longitudinal studies on the mediation effects of event-related rumination may be informed by these data. Data presented here encourage future studies to examine stress, event-related rumination, and internalizing psychopathology in a longitudinal design to directly test the direction of causality. • Finally, this dataset can inform clinical research that aims to treat depression through techniques that target rumination, such as by redirecting ones' intrusive, unconstructive ruminative thoughts into more deliberate, constructive forms [6] .

Data Description
The following files are available in the Mendeley Data online repository: dataset_filtered_n1507.sav: An SPSS file that contains dataset collected from 1507 participants who completed all items of the COVID-19 Impact Scale, ERRI, BDI-II, and BAI. Includes participants who experienced COVID-19 infection themselves or among family. Only fully complete data is presented here. Incomplete (not 100% complete) data has been filtered out. Table  1 summarizes the demographic characteristics of this COVID-infected sample. dataset_filtered_n1507_CSV.csv: The previously described filtered dataset, but in commaseparated format.
dataset_analyzed_n1069.sav: An SPSS file containing the dataset used in our multiple mediation analysis, collected from 1069 participants. Includes all participants in filtered dataset except those with experience of COVID-19 infection, whether themselves or among family. Table  2 summarizes the demographic characteristics of this COVID-free sample. ItemCodes.docx: The descriptions of each item and their associated ID codes for the dataset files.
Questionnaires.docx: The instructions and item lists for the COVID-19 Impact Scale, ERRI, BDI-II, and BAI. model1.txt: The MPlus input file that defines our main multiple mediation model, which was investigated in Squires et al. (2022) (Model 1). In this model, the predictor is COVID-19 Impact Scale total score, the outcomes are BDI-II and BAI total scores, and the mediators are Intrusive Rumination and Deliberate Rumination subscale scores; age and sex are included as covariates.
model2.txt: The MPlus input file that defines an alternative model to Model 1. In this model (Model 2), the predictor is COVID-19 Impact Scale total score, the outcomes are Intrusive and Deliberate Rumination subscale scores, and the mediators are BDI-II and BAI total scores; age and sex are included as covariates.
model3.txt: The MPlus input file that defines and tests another alternative model. In this model (Model 3), the predictors are BDI-II and BAI total scores, the outcome is COVID-19 Impact Scale total, and the mediators are Intrusive Rumination and Deliberate Rumination subscale scores; age and sex are included as covariates.
results_models2+3.docx: A summary of the multiple mediation results for Models 2 and 3. The results of Model 1 are presented in Squires et al. [5] .

Experimental Design, Materials and Methods
1507 participants were recruited from Reddit, Facebook, and the Queen's University undergraduate psychology participation pool as part of a larger study about rumination. Data were collected via the Qualtrics survey platform.
The questionnaire started with demographic questions asking participants about their age, gender, ethnicity, native language, marital status, occupational status, medication, etc. Then they were presented with a modified version of the Event-Related Rumination Inventory (ERRI) [3] . The ERRI includes 20 items, 10 for each of its two subscales: deliberate rumination (e.g., "I think about what the event might mean for my future.") and intrusive rumination (e.g., "Thoughts about the event come to mind and I cannot stop thinking about them."). Item scores range from 1 (Not at all) to 4 (Often). For the ERRI in this dataset, "…the event…" was replaced by "…the pandemic and its consequences…" to assess rumination about the COVID-19 pandemic specifically. The scores of each subscale were summed up to represent the level of engagement in the corresponding type of rumination.
ERRI was followed by a survey of pandemic-related stress, the COVID-19 Impact Scale, which has four domains -infection of oneself or family, social restrictions, impact on daily life (e.g., employment, education), and shortage of resources (e.g., food, medicine, etc.). Each item's score ranged from 1 (No Impact) to 5 (Severe Impact). Great internal consistency was shown in our sample (Cronbach's α = .82) for this scale. For the analyzed dataset, participants' data were excluded if the participant indicated that they or a family member had ever been infected by COVID-19. A total of 438 participants reported that they or a family member had a history of COVID-19 infection, and thus data from 1069 participants remained in the analyzed dataset. Two total scores were calculated for this scale by summing up all the relevant items (and are present in the filtered dataset): one that included the two items about personal and family COVID-19 infection, and another that excluded these items. The total with these items removed was what we used in our multiple mediation analyses.
Lastly, the Beck Depression Inventory-II (BDI-II) [1] and the Beck Anxiety Inventory (BAI) [2] were used to measure the severity of depression and anxiety respectively. The BDI-II contains 21 items, with the scores ranging from 0 (the least severe) to 3 (the most severe). It evaluates cognitive, affective, and behavioral changes in the last two weeks (e.g., worthlessness, sadness, changes in appetite). Similarly, BAI also has 21 items ranging from 0 (Not at all) to 3 (Severely). It concerns psychological and physical fluctuations such as numbness, dizziness, heart pounding, sweating, etc. The sum of the scores on each scale represented the severity of depression/anxiety.
None of the items required recoding. Altogether, we calculated five total scores to use as variables in our multiple mediation models: non-infection pandemic stress (PS), deliberate rumination (DR), intrusive rumination (IR), depression severity, and anxiety severity. Age and gender were included as covariates. Our multiple mediation models were tested using MPlus statistical modeling software.

Ethics Statements
This study was approved by the General Research Ethics Board at Queen's University (GPSYC-765-16). We confirm that informed consent was obtained from participants and that participant data has been fully anonymized. Participants were informed with a letter of information presented immediately after clicking the study link. Participants indicated their consent to participate by advancing to the next page of the online survey (and checking the Captcha box in the process).

Funding
This work was supported by a Natural Sciences and Engineering Research Council of Canada (NSERC) Postgraduate Scholarship (Doctoral), awarded to Scott Squires, M.Sc.; and an NSERC Discovery Grant ( 03637 ), awarded to Jordan Poppenk, Ph.D.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.