The COVID-19 pandemic presented a significant challenge in addressing the inconsistent adherence to recommended safety measures, particularly within university settings where students aggregated from diverse geographical locations. University campuses in the US were identified as hotspots for COVID-19 transmission. This study aimed to understand the factors influencing the variation in safety behaviors among university students to inform strategies for managing future outbreaks. In 2020, universities implemented various reopening plans, including modifications to housing and dining, reduced in-person activities, and expanded classroom space. However, behavioral-based prevention strategies remained crucial. Prior research has established links between safety adherence and factors such as social responsibility, perceived social norms, and the transactional nature of compliance. While a sense of duty and moral obligation have been associated with following guidelines, their impact on university students remains unclear. The perception of social norms among students, often misaligned with reality, also plays a significant role in their behavior. Therefore, shifting perceived social norms and providing accurate normative information could be effective strategies for improving compliance with COVID-19 precautions on university campuses. Past research demonstrates the effectiveness of educational courses in improving public health-related behaviors, creating cascading effects by fostering a culture of safety. Washington University in St. Louis (WUSTL) launched "The Pandemic: Science and Society" course in Fall 2020, aiming to provide comprehensive information on SARS-CoV-2, safety measures, and societal impacts. This large-scale course offered a unique opportunity to investigate the impact of such an intervention on campus safety behaviors. Early reports of outbreaks on other campuses highlighted the urgency of understanding and influencing student behavior. The study sought to understand students' information sources regarding COVID-19, the relationship between information sources, knowledge, risk perception, and safety behaviors, and the factors influencing adherence to COVID-19 precautions. The hypothesis was that knowledge and accurate risk perception would independently and jointly correlate with varying behaviors, predicting that official information sources (health organizations, the Pandemic course) would be associated with higher knowledge and safer behaviors.

Existing literature highlights the challenges in achieving consistent adherence to public health guidelines during the COVID-19 pandemic, especially within university settings. Studies emphasize the importance of understanding the factors influencing this variation to develop effective strategies for future outbreaks. Previous research has linked safety guideline adherence to social responsibility, perceived social norms, and the perception of risk. The role of education and information dissemination in shaping protective behaviors has also been highlighted. Studies have shown the effectiveness of educational interventions in improving public health-related knowledge and practices. However, the specific impact of such interventions on university students, and the role of different information sources, remains less explored. The influence of social media and traditional news sources, along with family and peer influence, on health behaviors is an area of ongoing research. The study aims to fill this gap by investigating the relationship between information sources, knowledge, risk perception, and behaviors among university students during the pandemic.

The study utilized an online survey administered at the end of the Fall 2020 semester at WUSTL. The survey collected data on respondents' demographics, information sources, COVID-19 knowledge, risk perception, and personal and close associate safety behaviors. A personal safety score (pSafety) and a closest associate safety score (caSafety) were created by assigning values to responses regarding personal and associate behaviors respectively. A risk perception score was developed based on respondents' assessment of the risk level of various activities. The top three information sources were also recorded. The final sample consisted of 928 completed questionnaires after applying exclusion criteria to remove incomplete or conflicting responses. Data analysis involved chi-square tests to compare the distribution of information sources across different COVID-19 knowledge scores (CKnow). Ordinal logistic regression was used to assess the impact of each survey question on the CKnow variable. The relationships among CKnow, pSafety, caSafety, and risk perception were analyzed using the polr function in R, with Nagelkerke's R² used to assess the model fit. The study also investigated differences in percentages of each primary source of information across CKnow, pSafety, and caSafety scores using chi-square tests.

The majority of respondents (83%) were WUSTL students, with 46% living off-campus, 36% on-campus, and 14.5% at home. 46.7% of respondents took "The Pandemic: Science and Society" course. COVID-19 knowledge scores were significantly associated with variations in attendance at large gatherings and restaurant visits, both personally and among close associates. However, the Nagelkerke's R² values for models relating safety scores to knowledge were low. The most frequently cited information sources were family (52%), CDC/WHO websites (50%), television/radio/print (47.4%), friends (38.6%), and "The Pandemic" course (32.4%). Official health organizations and the Pandemic course were associated with higher COVID-19 knowledge compared to other sources; those using family and friends as sources showed lower knowledge. Respondents who used the Pandemic course as a source had higher personal and close associate safety scores. Social media users had lower knowledge and safety scores. Government press briefings were associated with the lowest safety scores. Chi-square tests revealed significant differences in the percentages of each primary information source across different knowledge, personal safety, and close associate safety scores. Accurate risk perception was common (74.86% with scores of 9-10), with higher risk perception scores associated with using official sources rather than less formal ones. Safety behaviors were influenced by COVID-19 knowledge and accurate risk perception (R² = 0.33). All variables and interactions significantly explained variations in personal safety scores (Table 1).

The study revealed a strong correlation between information sources, COVID-19 knowledge, risk perception, and safety behaviors. Official sources like health organization websites and the Pandemic course were associated with higher knowledge and safer behaviors, underscoring the importance of providing reliable information. Conversely, reliance on informal sources like family and friends was linked to lower knowledge and less protective behavior. These findings align with previous research highlighting the impact of information source credibility on health outcomes. The role of social media was less prominent, and its use was associated with lower knowledge and safety scores, potentially due to the spread of misinformation. The study’s findings underscore the critical role of accurate risk perception in influencing protective behaviors. The Pandemic course at WUSTL served as a successful intervention by providing reliable information, highlighting risks, promoting safety measures, and building community. This approach aligns with strategies aimed at enhancing compliance with safety behaviors by fostering a culture of safety and emphasizing the collective benefits of protective actions. The study's limitations include reliance on self-reported behaviors and the lack of longitudinal data. However, the findings still offer important insights into designing effective public health communication strategies for universities.

This study demonstrated a significant relationship between information sources, COVID-19 knowledge, risk perception, and adherence to safety guidelines among university students. The Pandemic course at WUSTL, by providing reliable information and fostering a sense of community, proved effective in promoting safer behaviors. Official sources of information are critical for disseminating accurate knowledge and shaping protective actions. Future research could explore the long-term impact of such interventions, the role of specific communication strategies, and the effectiveness of interventions tailored to diverse student populations.

The study relies on self-reported data, which is subject to recall bias and social desirability bias. The cross-sectional design limits the ability to establish causality. The study population was predominantly comprised of WUSTL students, limiting the generalizability of the findings to other university contexts. The lack of longitudinal data prevents the assessment of the long-term effects of the Pandemic course and potential message fatigue.