The Impact of Information Sources on Mental Health During the COVID-19 Pandemic

Hypotheses and research questions:

Hypothesis 1: Media consumption across all information sources will significantly be
associated with depression and anxiety symptoms, with increased media consumption in general
associated with higher levels of health anxiety, depression, and general anxiety.

Hypothesis 2: Using social media and online interactive platforms to obtain news about the
pandemic in comparison to using traditional media (e.g., TV, radio, and newspapers) will be
associated with higher levels of health anxiety, depression, and general anxiety. Actively
staying away from information will further significantly be associated with higher levels of
higher levels of health anxiety, depression, and general anxiety.

Research Question 1: Is there a differential effect among different information sources on
health anxiety, depression and general anxiety? To what extend and how are different
information sources related to symptoms of health anxiety, depression, and general anxiety.

Exploratory:

Additionally, the investigators will exploratory investigate to what extent the amount of use
of different information sources impact health anxiety, depression, and general anxiety. The
investigators will also examine effect sizes with part correlations, to investigate
information sources with the most and least detrimental impact on health anxiety, depression,
and general anxiety.

Furthermore, the investigators will report which information platforms participants reported
as most useful with regards to information concerning how to best deal with the pandemic.

Statistical analysis:

Three multiple regression analyses will be conducted. 1) with PHQ-9 as the dependent
variable; and 2) the second with GAD-7 at as the dependent variable; and 3) the third with
Health Anxiety as the dependent variable. All analyses will include the different sources of
information as predictor variables, while controlling for the effects of pre-existing mental
health condition, gender, age, and education. Part correlations will be reported for each
regression analysis, presenting the effect size of the hypothesized predictors associated
with health anxiety, depression and general anxiety. A part (semi-partial) correlation gives
the least biased and easiest interpretable estimate of the strength of a predictive
relationship (Dudgeon, 2016). It is the correlation between the outcome and the aspects of
the predictor unique from all the other predictors. As a type of correlation, its size can be
evaluated according to Cohen's (1988) criteria: small >=0.10, medium >=0.30, large >=0.50.

The sources of information include: 1) National television and radio channels; 2) national
and regional, and local newspapers; 3) Use of social media for information; 4) use of blogs,
podcasts, forums or other Internet sources; 5) information obtained from peers (i.e., friends
and family); 6) Other sources of information; 7) actively staying away from information.

Multicollinearity and other statistical assumptions will be checked using examined.
Multicollinearity will be assessed with common guidelines (VIF < 5 and Tolerance > 0.2;
Hocking, 2003; O'Brian, 2007).

All analyses and questions addressed in the forthcoming paper that are not pre-specified in
this pre-registered protocol will be defined as exploratory.

Sensitivity analyses and random subsample replications of the main findings will be conducted
following selection of a random sample of participants that ensure a proportionate ratio
between the collected sample and the adult population of Norway.

Possible transformations:

All variables will be assessed in their original and validated format as is recommended
practice, as long as this is possible with regards to statistical assumptions underlying the
pre-defined analyses (i.e., multiple regression). However, if this is not possible with
regards to the statistical assumptions behind the analyses, transformation (e.g., square root
or log-transformations) may be needed to apply interval-based methods. The investigators will
examine the degree of skewness and evaluate this against the assumptions and analyses before
choosing the appropriate analysis. The pre-registered and planned analyses include multiple
regression as long as assumptions are met. Alternatively a non-parametric test will be used.

Inference criteria Given the large sample size in this study, the investigators pre-define
their significance level: p < 0.01 to determine significance.

Sample size and power calculation:

The present study is part of a larger project with the first part aiming to investigate
information source predictors of mental health through regression analyses, and the second
part aiming to examine directional relations amongst specific sources of information and
their centrality through complex systems approaches (i.e., network analysis). Consequently,
power calculations are based on power required for network analyses. Following power analysis
guidelines by Fried & Cramer (2017), it is recommended that the number of participants are
three times larger than the number of estimated parameters. However, more conservative
recommendations by Roscoe (1975) for multivariate research, recommends sample size that is
ten times larger than the number of estimated parameters. Thus, following these two
approaches respectively, between 1053 to 3510 participants are required. Data will be
collected for three weeks, and participants are based on a representative and random sample
of Norwegian adults, randomly selected and provided equal opportunity to partake in the
study, providing digital consent.

Missing data:

The TSD system (Services for Sensitive Data), a platform used in Norway to store
person-sensitive data verifies participants officially through a kind of national ID number
to give them full right to withdraw their data at any time, following the European GDPR
(General Data Protection Regulation) laws. Accordingly, participants are allowed to withdraw
their own data at any time. The survey includes mandatory fields of response. Participation
is voluntarily, and withdrawal of provided data is possible at any moment. The investigators
do not expect participants to withdraw their data and thus expect no missing data. However,
if participants do withdraw their data, the investigators will conduct state-of-art missing
data analyses and investigate whether data is missing at random.