The dynamics of risk perceptions and precautionary behavior in response to 2009 (H1N1) pandemic influenza

The medical outcomes of an infectious disease outbreak are affected by the behavior of individuals. Individuals who vaccinate, take anti-viral medications or stay home from work reduce not only their own risk of infection, but also those of others in the population. The dynamic nature of infectious disease transmission means that behavior by a modest number of individuals can have a significant impact on the trajectory of an outbreak [1]. Understanding individuals' behavior and its relation to their perceived risk is therefore important in terms of effective control of an infectious disease outbreak [2].

As the classic Health Belief Model and other health psychology models indicate, risk perceptions have been viewed as one of key drivers of health behaviors [3–7]. As an example, a meta-analysis showed that perceived risk likelihood and severity judgements predict vaccination behavior and that the relationship was stronger with better measures of perceived risk [4]. Precautionary behavior may also be influenced by the perceived costs and benefits of such behavior [8, 9], and the perceived impact of an individual's behavior on other individual's health outcomes [10]. Consequently, when predictors of precautionary behaviors, such as perceived risks and benefits change over time, we might expect the precautionary behavior to follow suite. An unexpected outbreak of infectious disease provides an opportunity to examine how new information affects risk perceptions and, hence, how changes in risk perceptions influence behavior. When the pandemic outbreak of H1N1 influenza started in Mexico City in April 2009, the massive media coverage in the early days of the outbreak coupled with global disease transmission instigated widespread fear initially [11, 12]. However, media attention declined gradually in the U.S [13]. Understanding the factors that drive individual behaviors in dynamic social contexts has both theoretical and practical implications.

In addition to the temporal change, the incidence of H1N1influenza in the US shows substantial geographical variation [14]. Consequently, objective risk of infection varies with geography. The outbreak therefore afforded the opportunity to examine whether perceived risk tracks objective risk. We examined the geographical differences in risk perceptions and precautionary behaviors.

Recent studies examined risk perceptions, emotions, and precautionary behaviors during the early stage of the 2009 (H1N1) influenza pandemic in Britain, Hong Kong, Australia, Malaysia and Europe, and the USA [13, 15–21]. These studies found that precautionary behaviors were associated with anxiety about H1N1 influenza [13, 16, 17, 19–21], risk perceptions [13, 15, 16, 18–21], perceived efficacy of the precautionary behaviors [16, 20], and prior experience of influenza vaccination [15, 18, 21].

Our study focuses on the temporal and geographic dynamics of risk perceptions and precautionary behaviors. We examined how risk perception, willingness to take pharmaceutical interventions, and engagement in precautionary activities changed over time and differed by geographical risk status in each state. In addition, we compared survey responses to the number of H1N1-related articles published in newspapers to provide a preliminary look at the relationship between precautionary behaviors and media attention. Our measures of risk perception included perceived likelihood of infection and perceived severity [4].

Our survey conducted at the initial stage of outbreak indicated that perceptions about the risks associated with 2009 (H1N1) pandemic influenza, as well as interest in pharmaceutical interventions and precautionary activities, showed changes over time and variations over geography and demography. Although the perceived likelihood of H1N1 infection increased over time, interest in preventive pharmaceutical interventions and engagement in information seeking activities declined. These declines were correlated with the decrease in media attention to H1N1 throughout May 2009. We did not observe the decline in engagement in quarantine measures partly because of the small number of respondents who reported the activities.

Perceived likelihood of infection also varied geographically. Respondents who lived in states with a greater number of H1N1 cases did indeed perceive a higher likelihood of infection, suggesting that respondents were aware of the number of cases in their geographical area. This result was robust when H1N1 per population was used as a measure for geographical risk. Engagement in precautionary activities and interest in pharmaceutical intervention, however, were not found to track this geography-driven difference in perceived risk likelihood.

We also observed a number of demographic differences. Women showed a higher general concern about H1N1 - they perceived higher risk likelihood, were willing to pay more to receive pharmaceutical interventions and more likely to engage in information seeking activities. This gender difference in risk perception is consistent with studies on risk perceptions on health [25, 26]. Respondents from larger households undertook more precautionary activities and were more interested in pharmaceutical interventions. Although our demographic data do not allow us to identify the structure of each household, a reasonable guess would be that larger households tend to include a child or children in the household. Influenza transmission from children to adults in a household is often emphasized [27–30], and the positive association between the degree of precautionary behavior and household size would be normatively predicted. In addition, in H1N1 influenza, studies reported that hospitalization rate and mortality caused by infection among children were higher than for seasonal influenza [31, 32], which may have further contributed the greater degree of interests and engagement in precautionary behaviors by respondents from larger household. There were few age differences, although older respondents perceived a higher death toll and were willing to pay less to receive pharmaceutical intervention.

We found that perceived likelihood of H1N1 influenza infection tracked objective risk both dynamically and geographically. The temporal dynamic change in risk perception on an infectious disease in response to the objective level of problem was previously found but in a longer time frame [33]. In an emergency situation such as a disease outbreak, however, individuals' risk perceptions could be adjusted in the time frame of days or weeks. In contrast to risk likelihood perceptions that increased over time, respondent's degree of precautionary behaviors declined in willingness to accept preventive intervention and in engagement in information seeking activities, following a pattern similar to the level of media attention. As discussed previously, this decline in precautionary behavior mirrors the decline in media attention. In addition, it also mirrors a sharp decline in Web searched about influenza. Google Insight ^® (Google Inc., Mountain View, CA) indicates that search volume for the search term "flu" showed a tremendous spike right after the outset of the pandemic in April 2009, but quickly returned close to before-pandemic levels in the following two weeks [34]. Data on risk perceptions and behavior change during the initial phase of a disease outbreak are rarely available [13], and our analysis provides useful information on individuals' response associated with the dynamic nature of an infectious disease outbreak.

Our survey results have several implications for successful response to a novel influenza outbreak. Effective vaccination strategies against influenza have been receiving considerable attention [30, 35, 36]. Successful implementation of an optimal vaccination strategy depends critically on individuals' willingness to accept pharmaceutical and non-pharmaceutical recommendations. As in previous questionnaire research on vaccination behavior [3–5, 15, 18], willingness to accept a pharmaceutical intervention was associated both with perceived risk and with an individual's adopting precautionary activities. Thus, acceptance of an H1N1 vaccine is likely to be highest among individuals who perceive high risk and who have already engaged in precautionary activities.

Our study also showed that changes over time in willingness to take action tracked temporal changes in media attention. Furthermore, the vast majority of the survey participants reported that they first knew about the H1N1 influenza outbreak through one form of media, suggesting the importance of mass media as an information source. Thus, campaigns to change public behavior may be most successful at the height of media attention, which may occur during the early stages of an outbreak. A previous study showed that national and international public health authorities were the most important source of information on H1N1 influenza in media reports [37]. Therefore a high emphasis should be given to the role of public health authorities in encouraging the public to take the preventive measures.

Finally, individuals in certain demographic categories may be most receptive to pharmaceutical interventions. Young women from large households expressed the highest level of interests in pharmaceutical interventions, and thus may be a potentially successful target of pharmaceutical intervention campaigns. Further study is needed to examine how perceptions and behavior change in response to intervention campaigns.

There are several limitations to our study. First, our sample size was limited and response rate is low. The lows response is partly because our survey was optional and respondents were free to decline any invitation they received. The survey company issued invitations in such a way that the age and gender distribution of participants would approximate that of the US adult population on each day. The survey procedure did not allow us to identify the exact proportion of those who did not participate and of those who were turned away after the daily quota was reached. Second, an optional survey is subject to self-selection bias. In particular, respondents may have been more interested in and concerned about H1N1 influenza than non-respondents, which may result in reporting higher degree of risk perception and/or interest in precautionary behavior than non-respondents. Also our survey was not exempt from limitations that web-based surveys commonly have. For example, gender, age, education background, social status may be related to access to computers or attitude towards web-based surveys [38]. However, most of our results were based on a comparative analysis, which should not be affected by baseline levels of risk perception. Third, we did not collect responses repeatedly from a single cohort but our respondents consist of a different cohort of individuals every day, as in a previous study [13]. Although this collection scheme provided greater sample size to analyze geographical and demographical variations, it resulted in a slightly different distribution on each day of the respondents in terms of sex and age given the limited sample size per day. We therefore controlled for demographic variables in our time-series and geographic analysis. Fourth, our analysis included sex, age, and household size as respondents' characteristics because the primary interest was the dynamics in risk perceptions and precautionary behaviours. Associations with additional respondents' socioeconomic characteristics could be potentially addressed in future research.

This study tracked individual's risk perceptions and precautionary behavior in an initial stage of a disease outbreak, and provides important insight into temporal and geographical dynamics of risk perception and precautionary behavior. Our result suggested that perceived risk of infection and precautionary behavior were dynamic in time and differed by demographic characteristics and geographical location. These patterns will likely influence the effectiveness of disease control measures.