Central indicators of psychological functioning such as difficulties in emotion regulation and habitual problems in one's relating to others are likely to have been substantially impacted by the COVID-19 amelioration measures of societal lock-down and physical (ne social) distancing. In turn, as these amelioration measures have been relaxed, that impact will presumably be reduced, gradually returning these factors to pre-crisis levels. Also, these factors are likely to predict mental health outcomes such as symptoms of depression and anxiety throughout the pandemic and beyond, so that levels of emotion regulation difficulties and interpersonal problems early on will predict later symptom status. Similarly reductions in such difficulties during the various phases of the outbreak will be associated with a concurrent reduction in psychological symptoms and reduced symptom levels at later stages.
Hypotheses/Research questions:
H1: There will be a significant decrease in emotion-regulation difficulties (DERS) and
interpersonal problems from T1 to T2.
Exploratory: Examine the difference in interpersonal problems and DERS among different
sub-groups before and after the COVID-19 pandemic
H2: Higher level at T1 and less reduction from T1 to T2 in emotion-regulation difficulties
and interpersonal problems will be associated with less reduction in anxiety and depression
from T1 to T2, above and beyond, demographic variables (age, gender, education).
Exploratory: Explore the different octants of IIP as predictors of changes for anxiety and
depression.
This study is part of 'The Norwegian COVID-19, Mental Health and Adherence Project',
involving multiple studies.
Statistical models:
Repeated surveys like the present one typically have high drop-out and substantial missing
data. Therefore, we will use mixed models instead of paired t-tests, repeated measures
ANOVAs, and ordinary linear regression to analyze the data. Mixed models use maximum
likelihood estimation, which is the state of the art approach to handle missing data (Schafer
& Graham, 2002). Especially if data are missing at random, which is likely in our survey,
mixed models give more unbiased results than the other analytic methods (O'Connel et al.,
2017).
In preliminary analyses, and for each of the dependent variables (GAD-7 and PHQ-9, DERS ,
IIP), the combination of random effects and covariance structure of residuals that gives the
best fit for the "empty" model (the model without fixed predictors except the intercept) will
be chosen. Akaike's Information Criterion (AIC) will used to compare the fit of different
models. Models that give a reduction in AIC greater than 2 will be considered better (Burnham
& Anderson, 2004). The program SPSS 25.0 will be used (IBM Corp, 2018).
First, H1 about decrease in DERS and IIP, GAD-7 and PHQ will be tested by using anxiety or
depression as dependent variable in a model using time (T1 period = 0, T2 period = 1) as a
predictor. Second, demographic group variables will be added as predictors. Third, the
initial (T1) levels of DERS and IIP as constant covariates will be added, together with the
interactions of these constant covariates with time. These interactions represent tests of H2
about the covariates predicting change in anxiety and depression. Finally, the T2 levels of
DERS and IIP as constant covariates will be added, together with the interactions of these
constant covariates with time. These interactions represent tests of H2 about the change in
the covariates from T1 to T2 predicting change in anxiety and depression from T1 to T2s.
Transformations Depending on degree of skewness compared to theoretical possibilities and
interpretations, variables will be assessed in their original and validated format as is
recommended practice, as long as this is possible. As this study examines psychopathology
levels amongst a general population (and not a clinical population), we do expect a skewed
data. We will attempt to assess these variables in their original and validated format as is
recommended practice, as long as this is possible. However, if this is not possible to the
statistical assumptions behind the analyses, transformation may be needed to apply
interval-based methods. Alternatively a non-parametric test will be used.
Inference criteria
Given the large sample size in this study, we pre-define our significance level:
p < 0.01 to determine significance
Missing data Maximum likelihood
Sensitivity analyses Sensitivity analyses will be conducted after selecting a random sample
of participants to reflect the proportion of subgroups in the Norwegian adult population.
Exploratory:
Questions addressed in the future paper which is not pre-specified will be defined as
exploratory.
Inclusion Criteria:
- Eligible participants are all adults including those of 18 years and above,
- Who are currently living in Norway and thus experiencing identical NPIs, and
- Who provide digital consent to partake in the study.
Exclusion Criteria:
- Children and adolescents (individuals below 18)
- Adults not residing in Norway during the measurement period
Sverre Urnes Johnson, PhD
41633313
s.u.johnson@psykologi.uio.no
Omid Ebrahimi, Mr
omideb@uio.no