Psychological impacts from COVID-19 among university students: Risk factors across seven states in the United States methods
Aim. Evidence-backed execution summary for Psychological impacts from COVID-19 among university students: Risk factors across seven states in the United States methods from Psychological impacts from COVID-19 among university students: Risk factors across seven states in the United States.
Show snapshot details
On this page
This experiment, in seven questions
Jump straight to the part of the recipe you need. Data and provenance labels stay close to the action they support.
Shopping and prep list
What do I need before I start?
human
Subject model for the experiment.
- Use
- confirm full cohort details in the source paper
2.3 Analyses
reagent used in the protocol.
- Use
- Last, using the composite scores from the EFA, we used the identified latent constructs from the psychological impact survey items as input variables in a latent profile analysis (LPA) [ ]. Criteria for determining the number of profiles in the LPA included statistical adequacy of the solution and interpretability o...
3.2 Psychological impact profiles
reagent used in the protocol.
- Use
- Mean values of the psychological impact survey items are shown in. Eight of these were included in the EFA. (Feeling guilty demonstrated low communality ( h 2 =.21) and was removed from further analyses.) All eight items displayed relatively normal distributions ( ). Criteria of the resulting model were acceptable...
3.2 Psychological impact profiles
reagent used in the protocol.
- Use
- A three-profile solution fit the data best for the LPA. Information criteria decreased with additional profiles up to a five-profile solution, indicating a better model fit (, ). The elbow plot suggested minor improvements in model fit after a three-profile solution. Adding a fourth or fifth profile provided less i...
2.3 Analyses
We imputed missing values by bag imputation, which fits a machine learning regression tree model for each predictor as a function of all others [ ]. In our dataset, 5.2% of the quantitative data were missing and imputed.
- Use
- We imputed missing values by bag imputation, which fits a machine learning regression tree model for each predictor as a function of all others [ ]. In our dataset, 5.2% of the quantitative data were missing and imputed.
4.2 Recommendations for universities
Students in this study also expressed stress and anxiety associated with changes in education mode during the pandemics. As previous research has found, academic success may be supported with virtual town halls, regular email check-ins, virtual office hours, and peer mentoring [ ]. Globally, students' satisfaction w...
- Use
- Students in this study also expressed stress and anxiety associated with changes in education mode during the pandemics. As previous research has found, academic success may be supported with virtual town halls, regular email check-ins, virtual office hours, and peer mentoring [ ]. Globally, students' satisfaction w...
Before you run
What should be confirmed before execution?
First confirmation
Equipment is listed but no product mappings are linked.
Confirm before execution
This page is backed by a publishable Replication Data Ledger package with zero critical source-verification issues.
Confirm before execution
Open the source paper before finalizing run-specific details.
Procurement checkpoint
Use source-stated vendors where present. Treat mapped products as sourcing options unless the page marks an exact source match.
Open quote workflowStep-by-step procedure
What do I do, in order?
2.2 Measures
2. 2. 1. 2 Quantitative assessment. Regarding our selection of quantitative impacts to measure, we chose nine survey items based on information gathered from a review of previous research and new interview data. These nine survey items measured the following concepts: negative emotion states, preoccupation with COVID-19, feeling stressed, worry, and time demands.
2.2 Measures
Regarding the review of previous research, we examined studies of other large-scale disasters (i.e., the World Trade Center terrorist attacks on September 11, 2001; previous epidemics requiring quarantine), which are almost always associated with psychological impacts on the general population [ ]. These studies provided some guidance on what impacts to measure for the impacts of COVID-19 on college students.
2.2.2 Risk factors
Sociodemographic factors were self-reported and allowed identification of potential differences in impact levels by gender, age, race/ethnicity, socioeconomic status (SES), and academic status (undergraduate vs. graduate-seeking). SES was measured with perceived social class, which has been shown to accurately represent SES in student populations, using a battery of seven questions on class, parental education, and relative family income [, ]. To measure academic status, we asked respondents whether they were in pursuit of an undergraduate or graduate degree.
2.2.2 Risk factors
Another set of plausible lifestyle-related risk factors was time use. We utilized a recent recall question structure from the American Time Use Survey that strongly predicts objective time use and activity measures [ ]. Three items were used to ask respondents to indicate how many hours they spent outdoors (at a park, on a greenway/trail, in a neighborhood/yard, etc.), in front of a screen (on a smartphone/computer, watching television, online gaming, etc.), and engaged in moderate or vigorous physical activity that caused an increase in breathing or heart rate (fast walking, running, etc.) in the past 24 hours [, ].
2.3 Analyses
Last, using the composite scores from the EFA, we used the identified latent constructs from the psychological impact survey items as input variables in a latent profile analysis (LPA) [ ]. Criteria for determining the number of profiles in the LPA included statistical adequacy of the solution and interpretability of each profile [ ]. Indices used to determine statistical adequacy included the Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC) and sample-size adjusted Bayesian Information Criterion [ ]. For each of these indices, lower values represented better model fit. Also, the entropy criterion was calculated as a measure of classification precision [ ]. We favored a parsimonious solution with fewer profiles over a more complex solution if this improved the interpretability of the LPA [ ]. Z-scores of the input variables were used to interpret the profiles....
2.3 Analyses
For the unadjusted results, risk factors were evaluated with chi-squared contingency tables. Residuals from observed versus expected count comparisons determined the direction of the effect of the risk factors (i.e., more or less likely that a group was classified to a higher impact profile than another profile). Statistical significance of risk factors was calculated with Bonferroni adjustments to reduce Type I Error [ ]. Continuous measures were reduced to dichotomous or categorical factors based on clinically meaningful levels, past research, and data distributions. BMI was classified into four categories (less than 18 = underweight; 18 to 24.99 = normal; 25 to 29.99 = overweight; 30.0 and over = obese) [ ]. General health was separated into two groups: poor/fair health and good/very good/excellent health [ ]. Screen time was separated into less than eight hours on a device and eig...
3.2 Psychological impact profiles
Mean values of the psychological impact survey items are shown in. Eight of these were included in the EFA. (Feeling guilty demonstrated low communality ( h 2 =.21) and was removed from further analyses.) All eight items displayed relatively normal distributions ( ). Criteria of the resulting model were acceptable: Tucker Lewis Index = 0.95; Kaiser-Meyer-Olkin (KMO) factor adequacy measure of sampling adequacy (MSA) =.89 [ ]; significant Bartlett's test of sphericity, χ 2 (28) = 10503, p <.001. The VSS Criterion [ ] achieved a maximum of.93 with a two-factor solution, compared to.89 for a one-factor solution or.94 for a three-factor solution (, and Figs). We labelled the first factor as "Emotional Distress" since it was composed largely of negative affect items (afraid, irritable, sad, preoccupied and stressed). The second factor was composed of three items dealing with h...
3.3 Risk factors
Sociodemographic (a), lifestyle (b), and COVID-19 victim awareness (c) risk factors associated with high, moderate, and low psychological impact profiles for students across the United States. Residuals from Pearson's chi-squared tests depict likelihood of profile membership based on risk factor. Only significant factors ( p <.05) are reported. Reference groups include men; over 32 age; other race/ethnicity; average/above average SES (social class and relative family income); good/very good/excellent general health; less than 2 hours of time outdoors; less than 8 hours of screen time; and not knowing someone infected (COVID-19).
Measurement outputs
What raw and processed outputs should exist?
As a sensitivity analysis, we ran a logistic regression model with a subsample of respondents from the university that obtained a representative sample (North Carolina State Uni...
- Raw artifact
- Per-sample or per-animal endpoint measurements collected during the experiment
- Processed artifact
- Structured table with cleaned measurements ready for comparison
- Reported as
- Summary statistics and between-group or across-timepoint comparisons
To accomplish Objective 1, qualitative data from the open-ended responses were analyzed using content analysis with an inductive approach [, ]. Two independent researchers exam...
- Raw artifact
- Per-sample or per-animal endpoint measurements collected during the experiment
- Processed artifact
- Structured table with cleaned measurements ready for comparison
- Reported as
- Summary statistics and between-group or across-timepoint comparisons
Next, we reduced the survey items related to levels of psychological impact into latent constructs using exploratory factor analysis (EFA) with oblimin rotation [ ]. Scree plots...
- Raw artifact
- Per-sample or per-animal endpoint measurements collected during the experiment
- Processed artifact
- Structured table with cleaned measurements ready for comparison
- Reported as
- Summary statistics and between-group or across-timepoint comparisons
Last, using the composite scores from the EFA, we used the identified latent constructs from the psychological impact survey items as input variables in a latent profile analysi...
- Raw artifact
- Per-sample or per-animal endpoint measurements collected during the experiment
- Processed artifact
- Structured table with cleaned measurements ready for comparison
- Reported as
- Summary statistics and between-group or across-timepoint comparisons
Analysis plan
How should the outputs become interpretable results?
Acquisition
Collect raw experimental outputs with enough metadata to preserve sample identity, condition, and timing.
inferred from protocolPreprocessing / cleaning
Of the 14,174 students invited to participate in the survey, we received 2,534 responses with data on most of the relevant variables; thus, this sample size was available for most of the descriptive statistics and bivariate associations.
from paperScoring or quantification
Quantify the primary readouts for this experiment: As a sensitivity analysis, we ran a logistic regression model with a subsample of respondents from the university that obtained a representative sample (North Carolina State Uni...; To accomplish Objective 1, qualitative data from the open-ended responses were analyzed using content analysis with an inductive approach [, ]. Two independent researchers exam...; Next, we reduced the survey items related to levels of psychological impact into latent constructs using exploratory factor analysis (EFA) with oblimin rotation [ ]. Scree plots...; Last, using the composite scores from the EFA, we used the identified latent constructs from the psychological impact survey items as input variables in a latent profile analysi....
from paperStatistical comparison
Of the 14,174 students invited to participate in the survey, we received 2,534 responses with data on most of the relevant variables; thus, this sample size was available for mo...; Last, using the composite scores from the EFA, we used the identified latent constructs from the psychological impact survey items as input variables in a latent profile analysi...; For the unadjusted results, risk factors were evaluated with chi-squared contingency tables. Residuals from observed versus expected count comparisons determined the direction o...; Mean values of the psychological impact survey items are shown in. Eight of these were included in the EFA. (Feeling guilty demonstrated low communality ( h 2 =.21) and was re...
from paperReporting output
Report representative outputs alongside summary comparisons for As a sensitivity analysis, we ran a logistic regression model with a subsample of respondents from the university that obtained a representative sample (North Carolina State Uni..., To accomplish Objective 1, qualitative data from the open-ended responses were analyzed using content analysis with an inductive approach [, ]. Two independent researchers exam..., Next, we reduced the survey items related to levels of psychological impact into latent constructs using exploratory factor analysis (EFA) with oblimin rotation [ ]. Scree plots..., Last, using the composite scores from the EFA, we used the identified latent constructs from the psychological impact survey items as input variables in a latent profile analysi....
inferred from protocolStructured statistical methods
Of the 14,174 students invited to participate in the survey, we received 2,534 responses with data on most of the relevant variables; thus, this sample size was available for mo...; Last, using the composite scores from the EFA, we used the identified latent constructs from the psychological impact survey items as input variables in a latent profile analysi...; For the unadjusted results, risk factors were evaluated with chi-squared contingency tables. Residuals from observed versus expected count comparisons determined the direction o...; Mean values of the psychological impact survey items are shown in. Eight of these were included in the EFA. (Feeling guilty demonstrated low communality ( h 2 =.21) and was re...
source structuredSource and audit
What supports the facts on this page?
Evidence quotes (8)
2. 2. 1. 2 Quantitative assessment. Regarding our selection of quantitative impacts to measure, we chose nine survey items based on information gathered from a review of previous research and new interview data. These nine survey items measured the following concepts: negative emotion states, preoccupation with COVID-19, feeling stressed, worry, and time demands.
Regarding the review of previous research, we examined studies of other large-scale disasters (i.e., the World Trade Center terrorist attacks on September 11, 2001; previous epidemics requiring quarantine), which are almost always associated with psychological impacts on the general population [ ]. These studies provided some guidance on what impacts to measure for the impacts of COVID-19 on college students.
Sociodemographic factors were self-reported and allowed identification of potential differences in impact levels by gender, age, race/ethnicity, socioeconomic status (SES), and academic status (undergraduate vs. graduate-seeking). SES was measured with perceived social class, which has been shown to accurately represent SES in student populations, using a battery of seven questions on class, parental education, and relative family income [, ]. To measure academic status, we asked respondents whether they were in pursuit of an undergraduate or graduate degree.
Another set of plausible lifestyle-related risk factors was time use. We utilized a recent recall question structure from the American Time Use Survey that strongly predicts objective time use and activity measures [ ]. Three items were used to ask respondents to indicate how many hours they spent outdoors (at a park, on a greenway/trail, in a neighborhood/yard, etc.), in front of a screen (on a smartphone/computer, watching television, online gaming, etc.), and engaged in moderate or vigorous physical activity that caused an increase in breathing or heart rate (fast walking, running, etc.) in the past 24 hours [, ].
Last, using the composite scores from the EFA, we used the identified latent constructs from the psychological impact survey items as input variables in a latent profile analysis (LPA) [ ]. Criteria for determining the number of profiles in the LPA included statistical adequacy of the solution and interpretability of each profile [ ]. Indices used to determine statistical adequacy included the Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC) and sample-size adjusted Bayesian Information Criterion [ ]. For each of these indices, lower values represented better model fit. Also, the entropy criterion was calculated as a measure of classification precision [ ]. We favored a parsimonious solution with fewer profiles over a more complex solution if this improved the interpretability of the LPA [ ]. Z-scores of the input variables were used to interpret the profiles. The criteria to assign low and high values is not established and so we adopted previous studies' thresholds [ ]. These included standardized scores between +0.5 and -0.5 being labelled as moderate, scores above 0.5 being labelled as high, and scores below -0.5 being labelled as low levels of psycho...
For the unadjusted results, risk factors were evaluated with chi-squared contingency tables. Residuals from observed versus expected count comparisons determined the direction of the effect of the risk factors (i.e., more or less likely that a group was classified to a higher impact profile than another profile). Statistical significance of risk factors was calculated with Bonferroni adjustments to reduce Type I Error [ ]. Continuous measures were reduced to dichotomous or categorical factors based on clinically meaningful levels, past research, and data distributions. BMI was classified into four categories (less than 18 = underweight; 18 to 24.99 = normal; 25 to 29.99 = overweight; 30.0 and over = obese) [ ]. General health was separated into two groups: poor/fair health and good/very good/excellent health [ ]. Screen time was separated into less than eight hours on a device and eight or more hours on a device [ ]. Time outdoors was split into three groups: Less than 1.00 hour, 1.00 to 1.99 hours, and 2.00 hours or more [, ]. Time spent exercising was also split into three groups: 30.00 minutes or less, 30.01 to 59.99 minutes, and 1.00 hour or more [ ]. In addition, social cl...
Mean values of the psychological impact survey items are shown in. Eight of these were included in the EFA. (Feeling guilty demonstrated low communality ( h 2 =.21) and was removed from further analyses.) All eight items displayed relatively normal distributions ( ). Criteria of the resulting model were acceptable: Tucker Lewis Index = 0.95; Kaiser-Meyer-Olkin (KMO) factor adequacy measure of sampling adequacy (MSA) =.89 [ ]; significant Bartlett's test of sphericity, χ 2 (28) = 10503, p <.001. The VSS Criterion [ ] achieved a maximum of.93 with a two-factor solution, compared to.89 for a one-factor solution or.94 for a three-factor solution (, and Figs). We labelled the first factor as "Emotional Distress" since it was composed largely of negative affect items (afraid, irritable, sad, preoccupied and stressed). The second factor was composed of three items dealing with how time was spent presumably in worry during the pandemic (worry, too much time and a lot of time), and so we labelled it "Worry Time." This is a term from clinical psychology that describes time spent reflecting on all the possible impacts of a health concern, including those worries that an indivi...
Sociodemographic (a), lifestyle (b), and COVID-19 victim awareness (c) risk factors associated with high, moderate, and low psychological impact profiles for students across the United States. Residuals from Pearson's chi-squared tests depict likelihood of profile membership based on risk factor. Only significant factors ( p <.05) are reported. Reference groups include men; over 32 age; other race/ethnicity; average/above average SES (social class and relative family income); good/very good/excellent general health; less than 2 hours of time outdoors; less than 8 hours of screen time; and not knowing someone infected (COVID-19).
Machine-readable layer
[
{
"@context": "https://schema.org",
"@type": "HowTo",
"name": "Psychological impacts from COVID-19 among university students: Risk factors across seven states in the United States methods",
"description": "Evidence-backed execution summary for Psychological impacts from COVID-19 among university students: Risk factors across seven states in the United States methods from Psychological impacts from COVID-19 among university students: Risk factors across seven states in the United States.",
"totalTime": "PT1560M",
"step": [
{
"@type": "HowToStep",
"position": 1,
"name": "2.2 Measures",
"text": "2. 2. 1. 2 Quantitative assessment. Regarding our selection of quantitative impacts to measure, we chose nine survey items based on information gathered from a review of previous research and new interview data. These nine survey items measured the following concepts: negative emotion states, preoccupation with COVID-19, feeling stressed, worry, and time demands."
},
{
"@type": "HowToStep",
"position": 2,
"name": "2.2 Measures",
"text": "Regarding the review of previous research, we examined studies of other large-scale disasters (i.e., the World Trade Center terrorist attacks on September 11, 2001; previous epidemics requiring quarantine), which are almost always associated with psychological impacts on the general population [ ]. These studies provided some guidance on what impacts to measure for the impacts of COVID-19 on college students."
},
{
"@type": "HowToStep",
"position": 3,
"name": "2.2.2 Risk factors",
"text": "Sociodemographic factors were self-reported and allowed identification of potential differences in impact levels by gender, age, race/ethnicity, socioeconomic status (SES), and academic status (undergraduate vs. graduate-seeking). SES was measured with perceived social class, which has been shown to accurately represent SES in student populations, using a battery of seven questions on class, parental education, and relative family income [, ]. To measure academic status, we asked respondents whether they were in pursuit of an undergraduate or graduate degree."
},
{
"@type": "HowToStep",
"position": 4,
"name": "2.2.2 Risk factors",
"text": "Another set of plausible lifestyle-related risk factors was time use. We utilized a recent recall question structure from the American Time Use Survey that strongly predicts objective time use and activity measures [ ]. Three items were used to ask respondents to indicate how many hours they spent outdoors (at a park, on a greenway/trail, in a neighborhood/yard, etc.), in front of a screen (on a smartphone/computer, watching television, online gaming, etc.), and engaged in moderate or vigorous physical activity that caused an increase in breathing or heart rate (fast walking, running, etc.) in the past 24 hours [, ]."
},
{
"@type": "HowToStep",
"position": 5,
"name": "2.3 Analyses",
"text": "Last, using the composite scores from the EFA, we used the identified latent constructs from the psychological impact survey items as input variables in a latent profile analysis (LPA) [ ]. Criteria for determining the number of profiles in the LPA included statistical adequacy of the solution and interpretability of each profile [ ]. Indices used to determine statistical adequacy included the Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC) and sample-size adjusted Bayesian Information Criterion [ ]. For each of these indices, lower values represented better model fit. Also, the entropy criterion was calculated as a measure of classification precision [ ]. We favored a parsimonious solution with fewer profiles over a more complex solution if this improved the interpretability of the LPA [ ]. Z-scores of the input variables were used to interpret the profiles...."
},
{
"@type": "HowToStep",
"position": 6,
"name": "2.3 Analyses",
"text": "For the unadjusted results, risk factors were evaluated with chi-squared contingency tables. Residuals from observed versus expected count comparisons determined the direction of the effect of the risk factors (i.e., more or less likely that a group was classified to a higher impact profile than another profile). Statistical significance of risk factors was calculated with Bonferroni adjustments to reduce Type I Error [ ]. Continuous measures were reduced to dichotomous or categorical factors based on clinically meaningful levels, past research, and data distributions. BMI was classified into four categories (less than 18 = underweight; 18 to 24.99 = normal; 25 to 29.99 = overweight; 30.0 and over = obese) [ ]. General health was separated into two groups: poor/fair health and good/very good/excellent health [ ]. Screen time was separated into less than eight hours on a device and eig..."
},
{
"@type": "HowToStep",
"position": 7,
"name": "3.2 Psychological impact profiles",
"text": "Mean values of the psychological impact survey items are shown in. Eight of these were included in the EFA. (Feeling guilty demonstrated low communality ( h 2 =.21) and was removed from further analyses.) All eight items displayed relatively normal distributions ( ). Criteria of the resulting model were acceptable: Tucker Lewis Index = 0.95; Kaiser-Meyer-Olkin (KMO) factor adequacy measure of sampling adequacy (MSA) =.89 [ ]; significant Bartlett's test of sphericity, χ 2 (28) = 10503, p <.001. The VSS Criterion [ ] achieved a maximum of.93 with a two-factor solution, compared to.89 for a one-factor solution or.94 for a three-factor solution (, and Figs). We labelled the first factor as \"Emotional Distress\" since it was composed largely of negative affect items (afraid, irritable, sad, preoccupied and stressed). The second factor was composed of three items dealing with h..."
},
{
"@type": "HowToStep",
"position": 8,
"name": "3.3 Risk factors",
"text": "Sociodemographic (a), lifestyle (b), and COVID-19 victim awareness (c) risk factors associated with high, moderate, and low psychological impact profiles for students across the United States. Residuals from Pearson's chi-squared tests depict likelihood of profile membership based on risk factor. Only significant factors ( p <.05) are reported. Reference groups include men; over 32 age; other race/ethnicity; average/above average SES (social class and relative family income); good/very good/excellent general health; less than 2 hours of time outdoors; less than 8 hours of screen time; and not knowing someone infected (COVID-19)."
}
],
"tool": [
{
"@type": "HowToTool",
"name": "2.3 Analyses"
},
{
"@type": "HowToTool",
"name": "4.2 Recommendations for universities"
}
],
"supply": [
{
"@type": "HowToSupply",
"name": "2.3 Analyses"
},
{
"@type": "HowToSupply",
"name": "3.2 Psychological impact profiles"
},
{
"@type": "HowToSupply",
"name": "3.2 Psychological impact profiles"
}
],
"isBasedOn": {
"@type": "ScholarlyArticle",
"headline": "Psychological impacts from COVID-19 among university students: Risk factors across seven states in the United States",
"datePublished": "2021",
"author": [
{
"@type": "Person",
"name": "Matthew H. E. M. Browning"
},
{
"@type": "Person",
"name": "Lincoln R. Larson"
},
{
"@type": "Person",
"name": "Iryna Sharaievska"
},
{
"@type": "Person",
"name": "Alessandro Rigolon"
},
{
"@type": "Person",
"name": "Olivia McAnirlin"
},
{
"@type": "Person",
"name": "Lauren Mullenbach"
},
{
"@type": "Person",
"name": "Scott Cloutier"
},
{
"@type": "Person",
"name": "Tue M. Vu"
},
{
"@type": "Person",
"name": "Jennifer Thomsen"
},
{
"@type": "Person",
"name": "Nathan Reigner"
},
{
"@type": "Person",
"name": "Elizabeth Covelli Metcalf"
},
{
"@type": "Person",
"name": "Ashley D'Antonio"
},
{
"@type": "Person",
"name": "Marco Helbich"
},
{
"@type": "Person",
"name": "Gregory N. Bratman"
},
{
"@type": "Person",
"name": "Hector Olvera Alvarez"
},
{
"@type": "Person",
"name": "Christine M Wickens"
},
{
"@type": "Person",
"name": "Chung-Ying Lin"
},
{
"@type": "Person",
"name": "Chung-Ying Lin"
}
],
"identifier": "10.1371/journal.pone.0245327"
}
},
{
"@context": "https://schema.org",
"@type": "BreadcrumbList",
"itemListElement": [
{
"@type": "ListItem",
"position": 1,
"name": "Experiments",
"item": "https://replicatescience.com/experiments"
},
{
"@type": "ListItem",
"position": 2,
"name": "Psychological impacts from COVID-19 among university students: Risk factors across seven states in the United States methods",
"item": "https://replicatescience.com/experiments/psychological-impacts-from-covid-19-among-university-students-risk-factors-across-seven-states-in-the-united-states-methods-matthew-h-e-m-browning-pmc7790395/psychological-impacts-from-covid-19-among-university-students-risk-factors-across-seven-states-in-th-mlph6kj7"
}
]
}
]