The COVID-19 pandemic necessitated various measures, including pharmaceutical interventions and changes in nonpharmaceutical preventive behaviors. Media sources play a crucial role in guiding individual decision-making during uncertain times like pandemics. Previous research indicated that social networking service (SNS) use promotes preventive behavior, but these studies were limited to single time points early in the pandemic. This study aimed to investigate the changing effects of media use on preventive behaviors over time, specifically focusing on the shift in individuals' attitudes towards going out and self-restraint during the pandemic in Japan. The prolonged nature of the pandemic and the evolving situation necessitate a longitudinal analysis to understand the sustained impact of media on voluntary preventive behaviors.

Existing research highlights the importance of media in influencing behavior during crises. Studies have shown that social media, in particular, can promote preventive behavior by providing information on the severity of infectious diseases and inducing fear, leading to reduced outings. However, social media is also characterized by selective exposure and echo chambers, potentially influencing the impact of information. Traditional media, on the other hand, is characterized by higher reliability and uniform information broadcasting. Previous studies primarily focused on single time points, lacking a longitudinal perspective on the evolution of media's impact on pandemic-related behaviors. This study addressed this gap by examining media's effects over time.

A two-wave panel survey was conducted in 2020 and 2021 in Japan to explore the temporal changes in the effects of media on individuals' preventive behaviors. The survey included questions on media usage (mass media and social media), media literacy, media suspicion, political interest, selective exposure, and Twitter use (browsing and posting). Participants were clustered into three groups based on their self-restraint behaviors (self-restraint, moderate, and going out) in each year. Stepwise logistic regression models were used to predict cluster affiliation in 2020 and 2021, and also to predict shifts in cluster affiliation between the two years. The independent variables were the media related variables for both 2020 and 2021 to model the 2021 cluster affiliation to capture potential changes in the relationship. Independent variables from both 2020 and 2021 were used for modeling 2021 cluster affiliations as lifestyle changes and media usage may have differed across individuals. The stepwise method ensured selection of variables with predictive power among numerous candidates.

Cluster analysis revealed three consistent groups across both years: a self-restraint group, a moderate group, and a going out group. The proportion of individuals in the self-restraint group decreased from 2020 to 2021. Stepwise regression analyses showed that in both 2020 and 2021, Twitter browsing positively predicted self-restraint, while Twitter posting negatively predicted it. However, the effect of SNS browsing on preventing going out was only observed in 2020. In 2021, higher media suspicion and selective exposure were associated with a greater likelihood of belonging to the going out group. These results highlight that while SNS browsing consistently influenced self-restraint, its effect on going out behaviors diminished over time. Age and gender consistently predicted cluster affiliation in both years. Political interest was a significant predictor in 2020 but not in 2021, potentially reflecting shifts in news coverage from hard news to soft news. In predicting cluster shifts between 2020 and 2021, the analyses revealed the factors influencing the transition from self-restraint to going out behaviors. The data is available through OSF at https://doi.org/10.17605/OSF.IO/YRBCG

This study's findings challenge previous research by demonstrating the temporal dynamics of media effects on COVID-19 preventive behaviors. The diminishing effect of SNS on going-out behaviors suggests that the initial impact of information dissemination waned as the pandemic progressed. The influence of cognitive factors like media suspicion and selective exposure in the later stage highlights the complexity of media effects and the role of individual biases. The longitudinal design of this study offers a more nuanced understanding compared to single-time-point analyses. The results have implications for public health communication strategies. Effective messaging should consider the evolving information landscape and adapt strategies to maintain compliance with preventive behaviors over time.

This study provides valuable insights into the dynamic interplay between media use, cognitive factors, and preventive behaviors during the COVID-19 pandemic. The temporal changes in media effects highlight the need for continuous monitoring and adjustments in public health communication strategies. Future research could explore the specific types of information on social media that influence behavioral change and the role of social influence in shaping preventive actions during protracted public health crises. Further investigation could also examine the interplay of different media types and their varying impact across diverse populations.

The study is limited to a Japanese sample, and the generalizability to other cultures might be limited. The reliance on self-reported data may introduce biases. The cross-sectional nature of the data limits the ability to establish causal relationships between media use and behavioral change definitively. Future research could address these limitations by including a broader sample and employing experimental designs.