Integrating Risk Perception Attitude Framework and Subjective Norms for Predicting Smokers’ Health Information Seeking

Protection motivation theory (Rogers, 1975, 1983) and the health belief model (Janz & Becker, 1984; Rosenstock, 1974; Sheeran & Abraham, 1996) are representative health communication theories, according to which perceived risk is a main factor for predicting health outcomes. In contrast, the risk perception attitude (RPA) framework argues that predictions of health outcomes must take into account efficacy as well as risk perception (Rimal, 2001).

The RPA framework is based on the extended parallel process model (EPPM) (Rimal, 2001; Rimal & Real, 2003), which combines the parallel process model (Leventhal, 1970, 1971) with protection motivation theory (Rogers, 1975, 1983) to compensate for the theoretical limitations of risk perception and its inconsistent relationship with health outcomes.

In the EPPM model, people with a high threat perception are divided into two different behavior pathways (the danger control process and the fear control process) and exhibit response behaviors to diseases in order to control the threat. Efficacy beliefs are thought to shape different behavioral pathways, which correspondingly lead to different health behavior outcomes for diseases. The danger control process leads to positive health behavior, while the fear control process leads to behavior that avoids or rejects health behavior. Perceived threats motivate a behavioral response, but at the same time, they can trigger a feeling of fear. If one’s efficacy is low, fear is amplified, prompting a defensive motivation for controlling fear; thus, the fear control process is undertaken to avoid or reject the threatening situation. On the contrary, high levels of efficacy weaken one’s sense of fear and prompt the danger control process, which enables one to overcome the perceived threat, leading to health behaviors that can prevent disease.

The RPA framework follows the EPPM model, which sees efficacy beliefs as a moderating variable between risk perception and health outcomes. Witte (1992, 1994) emphasized that preventive behaviors to protect an individual’s health from a disease are not stimulated by perceived risk and that efficacy beliefs are essential for preventing disease. In other words, while perceived risk motivates preventive behaviors to escape the threat of disease, the actual preventive measures to respond to the threat of disease are more directly controlled by an individual’s efficacy beliefs (Witte, 1992, 1994).

As such, the RPA framework based on the EPPM model conceptualizes risk perception as a motivator to avoid diseases and efficacy beliefs as facilitators of behavioral change. There are four possible behavior classifications depending on risk perception and efficacy beliefs: responsive (high risk perception, high efficacy), avoidant (high risk perception, low efficacy), proactive (low risk perception, high efficacy), and indifferent (low risk perception, low efficacy).

Responsive behavior is shown by people with a high level of risk awareness about smoking and a high level of belief in their knowledge and ability to prevent smoking-related diseases. Such individuals are likely to respond immediately to new threats and take precautions to protect themselves from smoking-related illnesses and minimize damage.

Avoidant behavior occurs in people with a high level of awareness of the risks of smoking, but a low level of belief in their knowledge and ability to prevent smoking-related diseases. Such individuals take the risk of smoking-related illness into consideration, but fail to take corresponding measures to reduce their risk, or intend to avoid the risky situation altogether.

Proactive behavior is shown by people who are not well motivated, due to their low level of risk perception, but are likely to take measures before developing smoking-related diseases due to their high level of belief in their knowledge and ability to prevent smoking-related diseases.

Finally, indifferent behavior occurs in people with low levels of both risk perceptions and efficacy beliefs. In other words, they have no interest in the damage caused by smoking, little knowledge of smoking-related illnesses, and a low ability to avoid dangerous behaviors.

Corcoran (2013) and numerous other researchers (Cummings, Jette, Brock, & Haefner, 1979) have pointed out that theories that do not account for social norms and social and psychological factors are limited in their ability to predict health outcomes. Furthermore, according to Bettenhausen and Murnighan (1985), norms are some of the least visible aspects of human behavior, but function as a powerful form of social control. For this reason, the influence of norms when predicting an individual’s health outcomes should not be overlooked.

According to the theory of planned behavior, health behaviors are influenced not only by cost-benefit analyses and rational decision-making, but also by perceived behavioral norms within a reference group. Therefore, norms provide a social context and exert a strong influence on individuals’ decision-making process concerning their health behavior (Sorensen, Emmons, Stoddard, Linnan, & Avrunin, 2002).

Individuals who feel that they have normative support for a particular issue or behavior from a reference group are more likely to act in a way that is consistent with their attitudes than those who are not (Terry & Hogg, 1996; Terry, Hogg, & White, 1999). Numerous studies on the drinking behavior of adolescents show that the influence of a reference group acts as an important variable in individuals’ drinking behaviors (Kuther, 2002). The drinking behavior of adolescents is guided by the perceived norms of others around them, rather than by a cost-benefit analysis based on rational decision-making.

The theory of planned behavior (Ajzen, 1991) views subjective norms as a predictor of health-related behavioral intention, along with attitudes toward behavioral control and health-related behavior itself. However, several studies dealing with the relationship between norms and health outcomes have argued that norms not only act as predictors of health-related behavioral intention, but also as mediators or moderators. The main impact of norms is to directly affect health-related behavioral intentions, but norms also have the effect of mediating or moderating the relationship between perceived risk, efficacy, and health-related behavioral intention (Cummings et al., 1979; Park & Smith, 2007).

With the increasing tendency to treat patients as active consumers of healthcare (Frosch & Kaplan, 1999), information seeking has become a key concept in health communication. Health-related information seeking can lead to a deeper understanding of disease symptoms, prevention tactics, and effective cures by reducing uncertainty (Turner, Rimal, Morrison, & Kim, 2006). Seeking health information has been proven to be related to various health outcomes such as changes in cognition, behavior, and affect (Lambert & Loiselle, 2007). More specifically, depending on a person’s judgments, beliefs, and attitudes, seeking health information can have an impact on decisions to undertake health-promoting behaviors (Lambert & Loiselle, 2007; Upadhyay, Lord, & Gakh, 2019).

In particular, seeking health information can provide motivation for more informed decisions and actions to maximize the positive effect of health behaviors (Anker, Reinhart, & Feeley, 2011; Ramirez et al., 2013). Health information seeking can include internet searches, discussions with healthcare providers, and hotline calls regarding health-related topics (Anker et al., 2011). Low interest in health information seeking has been associated with negative health outcomes, including delayed diagnosis or treatment of diseases and persistent unhealthy behaviors (Johnson, 2014).

Moreover, according to large health campaigns like the Stanford Five-City Project (Winkleby, Flora & Kraemer, 1994) and the Minnesota Heart Health Program (Viswanath & Finnegan, 1996), information seeking and retention of knowledge can lead to positive health outcomes. It also has been proven that the positive effects of these large health campaigns continue to be effective and stimulate information-seeking behavior even after the campaign ends (Rimal, Flora, & Schooler, 1999). Therefore, information seeking can be considered an important factor for self-protective behavior (Rimal & Real, 2003).

The research model presented in Figure 1 empirically examined the effect of risk perception, efficacy beliefs, and subjective norms on information seeking and the moderating effect of subjective norms on information seeking within the RPA framework.

Figure 1. Research model.

Download Original Figure

The RPA framework conceptualizes efficacy beliefs as a moderating variable to explain the lack of consistent support for the causal relationship between risk perception and self-protective motivations and behaviors in health communication research (Rimal & Real, 2003). However, tests of the RPA framework have produced inconsistent findings; specifically, survey-based research (Rimal, 2001; Rimal & Real, 2003; Rimal, Böse, Brown, Mkandawire, & Folda, 2009) has reported more theory-consistent results than experiment-based research (Rimal & Real, 2003; Turner et al., 2006; Zhao & Cai, 2009). From these findings, it is reasonable to infer that risk perception or efficacy beliefs can play a major role in the RPA framework. In addition, it has been widely reported that subjective norms, which account for individuals’ awareness about the beliefs and actions of others (Fishbein, 1980), have been considered as a main factor for improving predictions about the results of health-related behavioral intentions (Corcoran, 2013; Cummings et al., 1979). Thus, self-protective motivations and behavioral intentions can be affected by subjective norms as well as by rational decision-making stemming from individual risk perception and efficacy beliefs. Thus, the following hypotheses are posited:

In some studies, subjective norms showed direct and moderating effects between perceived risk, efficacy beliefs, and health-related behavioral intentions (Cummings et al., 1979; Park & Smith, 2007; Park, Klein, Smith, & Martell, 2009). Thus, the following hypotheses are posited:

In the RPA framework, risk perception consists of severity and susceptibility. Severity refers to the degree to which one perceives a disease as serious, and susceptibility indicates the degree to which one perceives himself or herself to be susceptible to a disease. Efficacy beliefs consist of response efficacy, which is one’s belief in the effectiveness of health behaviors for preventing diseases which may pose a threat, and self-efficacy, which considers one’s capability to perform preventive health behaviors.

Variables used in the study model were developed to be measured using a 5-point Likert scale. To ensure the face validity of each variable’s measurement items, the measurement items from existing theories and prior studies were derived, with appropriate modifications for application to smokers. Operational definitions and related studies of the variables are shown in Table 1.

In order to verify the impact of risk perception, efficacy beliefs, and subjective norms on health-related information seeking, data were collected through a survey of undergraduate and MBA students majoring in business administration at a large university in Seoul, Korea. To fulfill the purpose of the study, respondents were limited to current smokers who had smoked at least 100 cigarettes in his or her lifetime, as defined by the United States Centers for Disease Control and Prevention. Of the 350 survey responses collected, 311 responses were used for analysis after excluding 39 unreliable responses. The demographic data of the respondents are shown in Table 2.

Before verifying the structural model, the reliability and validity of the measurement model were verified by conducting CFA using SmartPLS 2.0 to confirm whether the questionnaire items were configured and measured according to the research purpose. Since the model contains second-order variables, the factor scores of the first-order constructs were used to create second-order constructs (Chin, Marcolin, & Newsted, 2003). Following the suggestion of Thornton, Henneberg, and Naudé (2014), the validity of second-order variables was examined at the level of the first-order constructs.

With regard to reliability, the composite reliability (CR) value and the Cronbach’s alpha value were measured, both of which are considered to indicate high reliability if the value is greater than 0.7 (Nunnally, 1975) and satisfactory reliability if the value is greater than 0.6 (Hair, Sarstedt, Pieper, & Ringle, 2012). As shown in Table 3, the Cronbach’s alpha and CR values were 0.661 and 0.803, respectively, confirming reliability.

Next, in the analysis of average variance extracted (AVE), all variables exceeded the standard value of 0.5 (Fornell & Larcker, 1981) and all factor loadings were found to be significant (with values greater than the standard value of 0.5), indicating convergent validity (Bagozzi, Yi, & Phillips, 1991), with the exception of SELF1 and SELF2, which were eliminated since they had values below 0.5.

Additionally, the discriminant validity was verified by determining whether the square root of the AVE of each construct was higher than the correlation between that construct and the other constructs. Table 4 demonstrates that the square roots of the AVE for constructs shown in bold italics were higher than the correlation between those constructs and the others, thus confirming the discriminant validity of the measurement model (Fornell & Larcker, 1981).

In this study, the partial least squares structural equation modelling (PLS-SEM) method was used to verify the hypotheses, and SmartPLS 2.0 was used as an analysis tool. PLS-SEM is recommended for when the number of samples is small, when the data do not follow a normal distribution, when the rationale is not clear, or when second-order variables are used. In this study, since risk perception and efficacy were measured as second-order variables, it was judged appropriate to use the PLS structural equation model rather than a covariance-based structural equation model employed in AMOS and LISREL.

The structural model was evaluated using the variance of the described endogenous variables (R²) and the values of the path coefficients (β) that passed the significance level test. As second-order variables, risk perception and efficacy beliefs are formed using factor scores of the first-order constructs of severity, susceptibility, self-efficacy, and response efficacy (Bock, Zmud, Kim, & Lee, 2005). A bootstrapping resampling method (311 samples, 5,000 repetitions) was performed, and three hypotheses (hypotheses 1, 2, and 3) were supported as shown in Figure 2, showing the paths’ coefficients and t-statistic values for the structural models. Both risk perception and efficacy beliefs, which comprise the RPA framework, were positively associated with information seeking (β = 0.114, p < .05 and β = 0.305, p < .001, respectively), thus supporting hypotheses 1 and 2. Subjective norms are also significantly associated with information seeking (β = 0.213, p < .001), supporting hypothesis 3. Risk perception, efficacy beliefs, and subjective norms explained approximately 25% of variance in information seeking (R² = 0.247).

Figure 2. Results of the partial least squares method. * p < .05, *** p < .0001 (two-tailed).

Download Original Figure

A further analysis was conducted to explore whether subjective norms moderated the effect of risk perception and efficacy beliefs on information seeking. No moderating effect was found, and hypotheses 4a and 4b were therefore rejected. The results of the moderating effect analysis are shown in Table 5.