Using the Risk Perception Attitude Framework to understand Nasopharyngeal Cancer risk reducing behaviors

Background: To reduce deaths due to nasopharyngeal cancer (NPC), it is important to understand factors that predict NPC risk reducing behaviors. This study employed the Risk Perception Attitude (RPA) framework to predict NPC risk reducing behaviors among a group of Malaysians. Methods: A sample of Malaysians (n=215) completed a questionnaire about perceived susceptibility, perceived severity, self-efficacy, response efficacy and intention to enact self-protective actions to reduce NPC risk. Results: A majority of the participants had responsive (high risk, high efficacy) and proactive attitudes (low risk, high efficacy). Hierarchical regression of mediation effect under structural equation model (SEM) approach was used to test the theory. Response efficacy and self-efficacy were negatively associated with perceived risk (p<0.01). Intention was negatively associated with perceived risk and positively associated with response efficacy and self-efficacy (p<0.01). The testing of the indirect effect from perceived risk effect yielded different conclusions for each efficacy measure. For the relationship between perceived risk and intention, there was no mediation effect of response efficacy but the mediation effect of self-efficacy was significant (p<0.1). Heightened perceived risk weakens efficacy beliefs and intention to enact self-protective behavior, suggesting that low risk messages may work better to avert fatalistic thinking for this group. Perceived risk and response efficacy explained 26.5% of the variance in self-efficacy, suggesting the importance of framing NPC risk messages to heighten the audience’s confidence to enact self-protective health behaviors.

Intention was negatively associated with perceived risk and positively associated with response efficacy and self-efficacy (p<0.01). The testing of the indirect effect from perceived risk effect yielded different conclusions for each efficacy measure. For the relationship between perceived risk and intention, there was no mediation effect of response efficacy but the mediation effect of self-efficacy was significant (p<0.1).
Heightened perceived risk weakens efficacy beliefs and intention to enact self-protective behavior, suggesting that low risk messages may work better to avert fatalistic thinking for this group. Perceived risk and response efficacy explained 26.5% of the variance in self-efficacy, suggesting the importance of framing NPC risk messages to heighten the audience's confidence to enact self-protective health behaviors.
Background Nasopharyngeal cancer (NPC) when diagnosed before Stage 3 has good prognostic outcomes. Considering that NPC is sensitive to chemo-radiotherapy and results in a twoand three-year survival rate of 84% and 78% respectively in cases of early detection [1], regular screening is important to reduce mortality. Worldwide, deaths due to NPC number 50,000 out of 86,000 cases [2] and 71% of new NPC cases are from East and Southeast Asia [3]. In Malaysia NPC is the fifth most common cancer, and the lifetime risk for males was 1 in 143 and for females it was 1 in 417 in the 2007-2011 period [4]. Most of the NPC cases were detected at Stages 3 and 4 (63% for males, 60% for females) [5], suggesting lack of awareness towards NPC symptoms.
Minimizing risk factors can reduce NPC deaths. It is widely reported among NPC researchers that Epstein-Barr virus infection, smoking, and frequent consumption of preserved food and salted fish are associated with a high incidence of NPC [6]. Nonenvironmental risk factors of NPC include gender, ethnicity, and family history [7]. Certain ethnic groups such as the Bidayuh [8], Cantonese [9] and Malaysian residents in Sarawak, Penang and Labuan have higher NPC incidences [5]. Exposure to formaldehyde, wood dust, smoke and chemicals may also be involved in the pathogenesis of NPC [10]. Awareness of NPC risk factors can lead to adoption of risk reducing behaviors such as screening and reduced intake of NPC-causing foods. However, taboos surrounding cancer discourage screening for early cancer detection [11,12,13]. However, little is known about NPC risk reducing behaviors and factors that predict these behaviors among Malaysians. Thus far, research on NPC in Malaysia are mostly clinical studies on the epidemiology of NPC [14,15,16,17,18,19,20,21]. Some studies examined the knowledge of the primary care doctors on NPC [22] and social impact of NPC [23]. The present study is the first to investigate individuals' reasons for enacting preventive or self-protective measures to reduce risk in a setting where incidence of NPC deaths is high: NPC ranks number five in cancer incidences in Malaysia period [4]. This study aimed to employ the Risk Perception Attitude (RPA) framework to predict NPC risk reducing behaviors among a group of Malaysians.

Risk Perception Attitude (RPA) Framework
The RPA framework was developed by Rimal and Real [24] based on Witte's Extended Parallel Process Model [25] to understand the relationship between health risk perceptions and health behaviors, moderated by their efficacy beliefs. Risk perception is related to beliefs about perceived severity and perceived susceptibility, while risk prevention behaviour is related to efficacy beliefs, operationalised as the product of self-efficacy and response efficacy. Health behaviors refer to self-protective measures that individuals can take or intend to take to avoid certain diseases, including information-seeking behaviours.
Based on their risk perceptions and efficacy beliefs, individuals are categorized into one of the four attitudinal groups: responsive (high risk, high efficacy), avoidance (high risk, low efficacy), proactive (low risk, high efficacy) and indifference (low risk, low efficacy). Their non-intervention study on skin cancer-related behaviours showed that the responsive group reported healthier outcomes than the avoidance group, and the proactive group reported healthier outcomes than the indifferent group [24].
The use of RPA has identified diverse risk factors influencing health behavioral intentions.
For example, people who reported high levels of perceived cancer risk and strong selfefficacy were more motivated and able to engage in various health actions such as cancer-prevention diets [26], information seeking on cancer, and cancer screening [27,28,29]. Besides efficacy beliefs, other studies found that non-environmental risk factors moderated breast cancer risk perceptions and intention to undergo mammography screening such as rural-urban locality, religion, social-cultural beliefs [30], family history [31], age [32], and health literacy [33]. Spiritual health locus of control, for example, influenced American women's perceived risk and efficacy, which resulted in several health outcomes: message acceptance, talking about breast cancer, information seeking and behavioural intentions [34]. In addition, cancer fatalism, mistrust of health care providers, and previous negative experiences with the medical system have also been found to be associated with people's invariable attitudes towards cancer precautionary measures such as cancer screening and adherence to health care recommendations [35]. In their review of the field concluded that the causal relationship between people's health risk perceptions and their health behaviours is "more tenuous" than expected, given the "discrepanc(ies) in (research) findings" while some studies found a positive connection between people's perceptions of health risks and their health-related precautionary behavior, others discovered otherwise [27]. They further argued that the causes for these discrepancies could be attributed to researchers' own failure to take into account the various moderators that may influence the causal link between risk perception and health behavior. Consistent with the RPA and the findings of studies that have employed this framework, we hypothesise that -efficacy beliefs, in addition to various modifying factors, will predict NPC risk reducing behaviors.

Participants and Procedure
Eligible participants were Malaysian residents ages 15 or older from various ethnic and socio-economic status groups. Given that the focus of the larger study was to investigate NPC risk reducing behaviors and factors that predict these behaviors, Malaysians with and without experience of NPC were eligible to be enrolled into the study. Based on the general rule, the minimum number of respondents or sample size is five-to-one ratio of the number of independent variables to be tested. However, Hair et al. [36] proposed that the acceptable ratio is ten-to-one. In the end, our sample size was 215 respondents. Since we could not get a list of all the elements of the population, we used a non-probability sampling of purposive sampling. The characteristics of 215 participants are reported in Table 1.

Measures
The survey comprised a questionnaire eliciting self-reports of independent measures: perceived risk (three items, [37]); response efficacy (four items, [38]); and self-efficacy (six items, [39]). The dependent measures consisted of items on intended behaviour to reduce NPC risk (six items, [39,40]. The questionnaire also contained demographic and health status questions.
All the items used 7-point Likert-type scales with a neutral option. The questionnaire items are shown in Table 2. The instrument was developed by first conducting a pilot study in which the researchers surveyed their contacts for appropriateness and clarity of the items, and then refining the items based on responses received.  Table 3 shows that the squared correlation for each construct is less than the AVE values implying adequate discriminant validity.

Results
Based on Rimal and Real's RPA framework, the participants were categorized into one of the four attitudinal groups using their means scores in perceived risk, response efficacy and self-efficacy beliefs [24]. Over half of the participants had a proactive attitude (118 low risk, high response efficacy; 114 low risk, high self-efficacy); "they are not motivated by their perceived risk status, but rather by their desire to remain disease free" [24]. Table 4 shows that about 44% of the participants had a responsive attitude (96 high risk and high response efficacy; 92 high risk and high self-efficacy), characterised as "being aware of their risk status and believing they have the requisite skills to avert the threat of the disease" and "motivated in enacting self-protective behaviour" [24]. Very few participants had avoidant (high risk, low self-efficacy), and indifferent (low risk, low selfefficacy) attitudes towards averting the NPC threat. Path analysis was employed to analyze the seven hypotheses. Figure 1 presents the results. In term of goodness of fit, our R2 value is 0.265 suggesting that 26.5% of the variance in the extent of self-efficacy can be explained by the hierarchical mediation of perceived risk and response-efficacy. Hence, we can conclude that our model has good explanation power.
We divided the results into three stages, namely, 1) direct effect between independent variables and mediator including the dependent variable; 2) direct effect between mediator and dependent variable; and 3) indirect effect between independent variable and dependent variable.
Hypothesis 1 stated that perceived risk has an effect on response efficacy. Our results show the negative effect of perceived risk on response efficacy with the coefficient value of -0.1960. The effect is significant at 1% level (p < 0.01) supporting the H1. Hence, higher perceived risk leads to lower response efficacy. In RPA framework, this behavior is categorized as avoidance. The same conclusion is found for the relationship between perceived risk and self-efficacy.
Hypothesis 2 stated that perceived risk has an effect on self-efficacy. We found that perceived risk has a negative effect on self-efficacy at the significance level of 1% with the coefficient value of -0.1968 supporting the H2. This implies that higher perceived risk leads to lower value of self-efficacy which is categorized as avoidance attitudes in the RPA framework. Thus, heightened risk perceptions may weaken efficacy beliefs. Perceived vulnerability to NPC seems to weaken confidence in taking precautionary measures and confidence in the measures.  (Rimal, 2000).

Hypothesis 4 stated that response efficacy has an effect on intention and Hypothesis 5
stated that self-efficacy has an effect on intention. In terms of the effect of efficacy effects on intention, Fig. 1 shows that both efficacy measures have significant effects on intention at 1% significance level. First, response efficacy has a significant effect on intention with a coefficient value of 0.214 implying higher response efficacy strengthens the intention. Self-efficacy is also found to have a significant effect on intention with the coefficient value of 0.3667 implying that a high level of self-efficacy strengthens intention to enacting NPC risk reducing behaviors. These findings provide support for H4 and H5.
Participants who believe they can take effective measures to avert the threat of NPC may be motivated to engage in more extensive precautionary behavior.
We ran the indirect effect for testing the path analysis of response efficacy. The findings show that there is a significant effect of response efficacy as the mediator in linking perceived risk and self-efficacy. Therefore, confidence in the effectiveness of precautionary behavior induces self-confidence to adopt NPC risk reducing behaviors.
We also tested the indirect effect from perceived risk effect. Hypothesis 6 stated that response efficacy mediates the relationship between perceived risk and intention whereas Hypothesis 7 stated that self-efficacy mediates the relationship between perceived risk and intention. Our estimations have different conclusions for each efficacy measure. When we take response efficacy as the mediator, the result depicts that there is no mediation effect of response efficacy for the relationship between perceived risk and intention.
Meanwhile, when self-efficacy is used as the mediator, there is a significant mediation effect to the extent that perceived risk influences intention. Yet, the relationship is arguably significant due to its 10% significance level. The mediation effect is negative implying that self-efficacy significantly carries the influence of perceived risk to reduce the intention of participants to enact NPC risk reducing behaviors. It seems that while efficacy behaviors play a limited role in mediating the effect of perceived risk upon intention, self-efficacy or confidence in adopting precautionary measures have some mediation effect.

Discussion
The current study examined the viability of RPA in predicting intention to enact NPC risk reducing behaviors in a sample of Malaysians. To our knowledge, this is the first study to use the RPA framework to understand factors influencing individuals' motivation to enact preventive or self-protective measures to reduce NPC risk in a setting of high NPC incidence. There were no similar surveys in Malaysia or neighboring Asian countries with high incidence of NPC such as Indonesia and Hong Kong as the studies are also on epidemiology of NPC [45,46,47,48,49]. Hence, one strength of our study is that it has obtained information that is hitherto not available on perceptions of the severity of NPC, susceptibility to NPC and efficacy beliefs pertaining to reducing risk of NPC. Despite National Cancer Registry records showing high incidence of NPC among Malaysians (NPC ranking number five in cancer incidences [4]), the participants in our study perceived themselves to be at low risk to NPC and rated the severity of NPC as marginally low.
However, their efficacy beliefs and intentions to enact self-protective behaviors were moderately high. Their perceptions of low threat of NPC could be due to lack of awareness of NPC incidences. The latest National Cancer Registry showed that age-standardised rates for nose and throat cancer incidences among men and women in Malaysia are 6.4 and 2.2 respectively in the 2007-2011 period [4]. The results on the lack of awareness of NPC are consistent with findings on the insufficient knowledge on breast cancer [50,51], colorectal cancer [52], oral cancer [53], and cervical cancer [54,55] in Malaysia. In fact, Samat et al.'s [55] survey showed that only 68% of the participants have some knowledge of cancer risk factors, and the study involved urban and educated Malaysians. In our study, 57% of the participants reported having some knowledge of NPC (Table 1). It is expected that the level of awareness on NPC would be even lower among the rural population, indicating the need for education.
Our study yielded somewhat unexpected results on the negative relationship between perceived risk with intention, response efficacy, and self-efficacy. In the RPA framework, Rimal and Real [24] predicted that high risk would increase motivation to enact selfprotective health behaviours but in our study, it had the opposite effect. Two possible reasons are discussed here. First, there could be a strong sense of fatalism because the participants believed that there was not much they could do to avert NPC since cancer is unlike infectious diseases whereby people can take measures to minimize their risk. Ting et al. [13] found taboos surrounding cancer. The participants avoided the topic of NPC and were apprehensive when talking about it, for fear of inviting cancer into their lives [12].
They were mostly resigned to the fact that chance determined whether they got cancer.
Elsewhere, taboos have been found to hinder women from taking up breast cancer screening [11]. Second, the participants could be unaware of their risk. As a group, the participants had a low level of perceived risk. Few had a family history of cancer (only 14 out of 215, Table 1), and they believed they led a generally healthy lifestyle, and were not exposed to potential risk factors such as preserved food and environmental pollutants.
Furthermore, these risk factors of NPC are not established because of mixed results in past studies. Our study indicated that a high level of perceived risk weakens efficacy beliefs and intention to enact self-protective measures. However, in information campaigns, it is still necessary to impart information on the risk so that the public is informed of the correct facts on NPC risk and severity but the framing should take account their psychological profile, determined using the RPA framework. In addition to health risk, the framing of the NPC risk messages can take into account financial risk because cancer treatment and management is costly [57], and the awareness may motivate the public to consider screening and lifestyle changes (e.g., preserved food, exposure to environmental pollutants) although the latter may not be confirmed risk factors of NPC.

Conclusions
The study provides important information on how risk and efficacy perceptions of a sample of the Malaysian population with respect to intention to enact self-protective behaviors.
The most striking finding from our study is the stronger mediating influence of selfefficacy on perceived risk and intention, compared to response efficacy. In fact, response efficacy does not mediate between perceived risk and intention. Previous studies examined the relationship between perceived cancer risk and self-efficacy for intention to engage in cancer prevention diets [26], information seeking and cancer screening [27,28,29]. However, these studies treated efficacy beliefs as one construct. By treating response efficacy and self-efficacy as distinct constructs, our study has indicated that it is selfconfidence to enact self-protective health behaviors which motivated participants to act on perceived risk of NPC. Surprisingly, response efficacy did not influence participants to alleviate their risk of NPC. In our study, the self-protective measures examined were screening, reducing preserved food (e.g., preserved vegetables, salted eggs), reducing exposure to environmental pollutants, and adopting a healthy lifestyle (e.g., exercise).
The selection of these measures was based on research which has shown the association of NPC with preserved food [15,58] and environmental pollutants [23,59]. In our study, the participants had confidence that these self-protective measures were somewhat effective but this was not a factor in heightening their intention to adopt these measures.
Instead their intention to enact these measures was influenced by their self-efficacy or their confidence to exert personal control over their health behaviors. The group of participants had high response and self-efficacy, with almost equal numbers having proactive and responsive attitudes. They had different motivations to enact self-protective behavior; the proactive group seeks to remain disease free whereas the response group seeks to avert the disease threat [24]. Hence, taking into view the mediating effect of self-efficacy between perceived risk and intention, education to increase awareness of NPC needs to focus on elevating the confidence of the target audience to take selfprotective measures, and perhaps convincing them of the ease of performing those measures.
However, since it is the first of such studies to examine factors influencing individuals' motivation to enact self-protective measures to reduce NPC risk using the RPA, further studies should be conducted to test the model with other samples within the Malaysian population to find out whether it can explain intention to enact NPC risk reducing behaviors. While the current model has identified self-efficacy as an important construct to target in NPC risk message design, one limitation of the study is that the findings were based on self-reports of perceived risk of NPC which is variable and subject to an individual's prior beliefs. This was also clouded by the participants' taboos in assessing their own risk for fear that it would invite cancer incidences. In view of this, an important angle of investigation to undertake in future is message effects on perceptions of risk and efficacy to obtain evidence-based information on individual's conceptualization of perceived threat of NPC as a property of the message. The findings will inform the design of group-specific risk messages, taking into consideration the audience segmentation afforded by the survey findings on the influence of risk and efficacy beliefs on motivation to enact self-protective health behaviors.
Abbreviations NPC Nasopharngeal cancer RPA Risk Perception Attitude

Ethics approval and consent to participate
The enumerators explained the study to participants and asked them to sign a consent form indicating their willingness to participate in the study. They were told that no personal information would be revealed in any reports on the study. Written consent was obtained from participants who agreed to participate in the study. Because this was a lowrisk survey study, ethical approval was not needed at the university. However, to ensure the participants' safety, enumerators were trained to answer questions about NPC, and contact details of the researchers were provided to participants. There were no participants under 16 years old in the study.

Consent for publication
Not applicable

Availability of data and materials
The data used to support the findings of this study are available from the corresponding author upon request.

Competing interests
The authors declare that they have no competing interests  Structured model of the RPA framework predicting intention to adopt NPC risk reducing behaviors among Malaysians.