A five-wave Czech study offers a more complicated picture of digital media and depression than the familiar claim that more screen time automatically means worse mental health. Researchers examined how different ways of engaging with news and online content corresponded with self-reported depressive symptoms between 2020 and 2025. The associations changed over time, and some even reversed direction.
The research, published in Frontiers in Psychology on 20 October 2026, drew on survey waves containing 2,192 responses in 2020, 2,164 in 2021, 1,717 in 2022, 1,707 in 2023 and 2,072 in 2025. Its central finding is not that digital media causes depression. Rather, what people do with media, and the social circumstances in which they do it, may matter more than an undifferentiated count of minutes online.
An important complication appears within the paper itself: its abstract and detailed results do not consistently report the same direction for some 2025 correlations. This article prioritises the published numerical correlation table and flags that discrepancy rather than treating a contested summary as a settled conclusion.
Why a single screen-time number misses the point
Digital engagement includes activities that differ substantially. Reading economic reporting, browsing reader comments, discussing news with other users and feeling emotionally overwhelmed by coverage are not equivalent experiences. Yet public debates often put all these activities under the same heading of screen time.
The Czech research team separated media engagement into three components: serious news and economic reporting; reader comments and interactive content; and emotional impact and media stress. The goal was to examine whether these categories related to depressive symptoms in the same way across different years.
That distinction is especially relevant when news cycles are dominated by uncertainty. During the pandemic, information could be practically valuable and emotionally demanding at the same time. In later years, the same interactive channels might be used differently, including for discussion, support or ordinary information seeking. Those possibilities are plausible explanations, not effects that this observational study independently proved.
How the five-wave study was conducted
Areeba Syed, Marek Preiss, Dominika Grygarová, Zuzana Podaná and Jiří Horáček analysed data from the Czech National Panel. Respondents completed computer-assisted web interviews, with repeated collection in 2020, 2021, 2022, 2023 and 2025. The researchers describe the design as longitudinal, although the reported associations were calculated separately for each survey wave.
The study used a shortened five-item version of the Patient Health Questionnaire to measure recent depressive symptoms. Its items covered reduced pleasure, depressed mood, sleep problems, low energy and changes in appetite. These are useful symptom indicators, but a questionnaire score is not the same as a clinical diagnosis of major depressive disorder.
Media responses came from a questionnaire about media use and perceived impact. The researchers applied principal component analysis to identify related patterns of responses, then calculated Pearson correlations between the resulting media components and depressive symptom scores. Correlations indicate whether two measurements tend to move together. They cannot establish which one caused the other.
The wave sizes should not be added together and presented as a count of unique participants. Some people participated repeatedly. The detailed methods report 1,109 complete cases across all five waves, while the abstract gives 1,120, another small reporting inconsistency that readers should bear in mind.
Depressive symptom scores were higher in the final wave
The mean score on the study’s five-item depression measure was 7.413 in 2020 and 8.604 in 2025. Intermediate wave means were 7.593 in 2021, 7.641 in 2022 and 7.891 in 2023. These figures describe average symptom scores among respondents in each survey wave.
A higher average in 2025 is not proof that media exposure produced the increase. The social environment changed, respondents were not identical across every wave, and other influences on mental health were not experimentally controlled. The paper does not identify a single causal explanation for the observed pattern.
It is also important not to confuse changes in the survey mean with an estimate of how many people had clinically diagnosed depression. The authors measured symptoms through a brief questionnaire. That is valuable for detecting broad patterns, but the results do not directly establish a national clinical prevalence rate.
Reader comments showed the largest shifts
Reader comments and interactive content were positively correlated with depressive symptoms in the first four reported waves. The numerical table lists r = 0.233 in 2020, 0.275 in 2021, 0.256 in 2022 and 0.305 in 2023. These correlations are not enormous, but they were more pronounced than those reported for serious news.
In the published correlation table, the association changes sign in 2025, reaching r = -0.276. That means higher scores on this media component coincided with lower depressive symptom scores in that wave, rather than higher scores. However, the paper’s abstract describes a positive association in 2025, making the direction inconsistent across sections of the article.
The table-based result could reflect changes in how people engaged with online discussions, differences in respondents or the surrounding news environment. It does not demonstrate that reading comments protects mental health. It also does not show that hostile comment sections caused depression during the earlier years.
For digital publishers and platform designers, the practical lesson is to investigate the nature of interaction rather than assuming all comments are beneficial or harmful. Moderation quality, social support, harassment and topic sensitivity may be relevant, but the study did not separately test these mechanisms.
Serious news was not consistently linked to greater distress
Economic reporting and serious news showed very small correlations with depressive symptoms in the first four waves. The study reports r = 0.011 in 2020, 0.020 in 2021, 0.055 in 2022 and -0.021 in 2023. Only the 2022 result among these early waves was reported as statistically significant, and its magnitude was still small.
In 2025, the numerical results table reports a modest negative association of r = -0.104. This should not be read as proof that consuming serious news improves mood. It means that within that wave, respondents reporting more of that component tended to report slightly fewer symptoms.
This finding challenges the oversimplified assumption that keeping up with important news necessarily harms mental health. News consumption may support informed decisions, but people also differ in what they read, why they read it and how they respond. The study’s observational design cannot separate all these possibilities.
Research Today has previously covered a study linking reduced social media use with friendship outcomes. That is a different research question, and its findings should not be merged with the Czech correlations as if the studies tested the same intervention.
Media stress and emotional responses require careful reading
The component measuring emotional impact and perceived media stress showed small positive correlations in the early waves: r = 0.069 in 2020, 0.125 in 2021, 0.094 in 2022 and 0.110 in 2023. The detailed correlation table also lists a positive association of r = 0.150 in 2025.
The abstract, by contrast, reports a negative 2025 correlation for this component. Because the two sections disagree, a cautious account cannot claim with confidence that the emotional-stress relationship reversed in 2025. Independent clarification or correction from the authors would be needed.
There is another distinction between a media-related stress questionnaire and direct observation of a biological stress response. Participants reported how media affected them; researchers did not establish that reading a particular article produced a measurable physiological change or that any one media platform was responsible.
The authors discuss rumination, the repeated return to distressing thoughts, as a possible pathway connecting digital engagement and wellbeing. But the correlations reported here do not constitute a randomised test of that mechanism. A separate randomised study of career rumination demonstrates how different the evidence becomes when an intervention is experimentally assigned.
What the study can and cannot tell us
One strength of the research is its repeated measurement over several years, including the unusual social conditions of the pandemic and later periods. Rather than relying on a single cross-sectional snapshot, it offers evidence that media-depression associations are not fixed. The component-based approach is also more informative than a blanket measure of total screen time.
However, repeated observational surveys do not eliminate reverse causation. People experiencing depressive symptoms may seek out particular content, participate in discussions differently or feel more stressed by the same news. Media engagement could be a response to distress, a contributor to distress or both.
The survey skipped 2024, so changes between 2023 and 2025 cannot be located precisely. Participant attrition and renewed recruitment complicate direct comparisons. The authors also did not analyse the specific content of posts, articles or comments people encountered, which limits any conclusions about misinformation, hostility or supportive interactions.
The paper has internal numerical discrepancies, including conflicting descriptions of 2025 correlations and a difference in the number of complete five-wave cases. These do not erase the underlying data, but they weaken confidence in sweeping interpretations of the apparent reversals. Readers should place more weight on the clearly reported wave-specific numbers than on broad causal language.
What this means for readers and publishers
For individual readers, the findings support a more precise question than whether media is good or bad: what kind of engagement leaves a person informed, connected or distressed? An activity that helps someone understand events is different from repeatedly revisiting comments that provoke anxiety, even if both take the same amount of time.
For publishers, the study provides a reason to think about the experience surrounding a story. Clear context, responsible framing and well-managed discussions may matter. But the data do not establish that any specific platform design or moderation intervention would reduce depression, and such claims would require controlled testing.
For researchers, a valuable next step would be to test media experiences directly, distinguish active social support from conflict-driven interaction, and follow the same participants with consistent measures. Replication outside Czechia would help establish whether the observed shifts are specific to its recent history or appear in other populations.
The strongest conclusion is therefore narrower than a headline blaming social media for declining mental health. Media habits and depressive symptoms were associated in ways that varied by activity and year. Understanding those differences is a better starting point than treating every minute online as interchangeable.
Source Information
Study: Syed, A., Preiss, M., Grygarová, D., Podaná, Z., and Horáček, J. (2026). Digital media engagement and its relationship with depression severity.
Journal: Frontiers in Psychology, volume 17. Published 20 October 2026.
DOI: 10.3389/fpsyg.2026.1926446.
Method: Five-wave Czech online panel surveys (2020, 2021, 2022, 2023 and 2025); five-item depressive symptom questionnaire, media use components and wave-specific Pearson correlations. The article contains discrepancies between its abstract and numerical tables, noted above.
