No significant association of repeated messages with changes in health compliance in the COVID-19 pandemic: a registered report on the extended parallel process model

Introduction

One year has passed since the COVID-19 outbreak, but the health threat has not been terminated yet. As of February 2, 2021, there are over 100 million confirmed cases globally; in Japan, 389,518 cases and 5,722 deaths (mortality rate: 1.47%) have been reported (World Health Organization, 2021). During this period, while a heavy loss of both life and economy is caused, some countries and regions have achieved staged success in the fight against this disease. There is no doubt that public compliance with effective health proposals plays a crucial role in achieving this success.

It is suggested that functional fear, which is explained as one of the negative emotions serving protective functions in certain contexts, has promoted public health compliance during the COVID-19 pandemic (Harper et al., 2020). Nevertheless, this needs to be explored further, especially considering previous studies on health communication and the features of information dissemination in real life. The main question being addressed in our study is to examine how people’s health compliance intention is influenced by various factors in the COVID-19 context in an exhaustive way.

It is known that fear has an impact on health compliance. In fact, fear-arousing communication is considered an effective way to promote health campaigns and has been widely investigated for promoting health awareness related to various topics, including smoking (Leventhal & Watts, 1966), alcohol use (Wolburg, 2001; Moscato et al., 2001), AIDS (Treise & Weigold, 2001) and so forth. After several initial studies, inconsistent results indicated that a simple monotonic function of fear might not be expected in persuasive health communication. Specifically, despite a large number of studies indicating the positive main effect of fear on persuasion, which means that fear improves persuasiveness (Leventhal & Watts, 1966; Dabbs & Leventhal, 1966; Leventhal, Singer & Jones, 1965), relatively few report the negative main effect of fear, showing as less attitude change with stronger threat (Janis & Feshbach, 1953; Janis & Terwilliger, 1962). Later research confirmed that the effects of fear-arousing communication interact with various source variables, message variables, and receiver variables and thus cannot be described easily (Miller & Hewgill, 1966).

One of the most recent and prevalent theories on fear-arousing communication is the extended parallel process model (EPPM: Witte, 1992, 1994), which is based on former frameworks, including the Parallel Response Model (Leventhal, 1970) and protection motivation theory (Rogers, 1975). In the EPPM, there are four main factors, which influence the prediction of certain communication outcomes: perceived susceptibility and severity composing perceived threat, self-efficacy and response efficacy composing perceived efficacy. Perceived susceptibility refers to a belief about the risk of experiencing a threat, whereas severity refers to a belief about the magnitude of the threat. On the other hand, self-efficacy is defined as a belief about the ability to perform a recommended proposal to avert the threat; response efficacy is a belief about the effectiveness of the recommended proposal in deterring the threat (Witte, 1996). Concretely, when both perceived threat and perceived efficacy are high, people are most likely to engage in a danger control process, which means conforming to the recommended health proposal. Nevertheless, when perceived threat and perceived efficacy are high and low, respectively, people turn to a fear control process leading to coping responses that reduce fear and danger control responses. Meta-analyses on the results of fear-arousing research confirm the validity of the EPPM (Floyd, Prentice-Dunn & Rogers, 2000; Peters, Ruiter & Kok, 2013).

Currently, we are exposed to the same health proposals concerning COVID-19, in all likelihood, at least once in our daily lives, which makes message repetition an essential factor to be taken into consideration. Although research concerning stimulus repetition first emerged in the 1960s (Zajonc, 1968), there was no study on message repetition in persuasive communication until 10 years later. It is suggested that under moderate repetition (less than three times), agreement toward persuasive messages rises (Cacioppo & Petty, 1979). The rationale is explained as follows: Scrutiny is reinforced through a moderate level of message repetition, which enhances the understanding of message content and the merits advocated by the message, consequently improving supportive attitudes toward the message. Later research revealed that the mechanism mentioned above is applicable only when arguments in the message are perceived as strong and when the issue is of high personal relevance (Cacioppo & Petty, 1989; Claypool et al., 2004).

In health communication on COVID-19, we assumed that the interpretation of the content of a certain health proposal message might be related to factors in the EPPM. Concretely, response efficacy and perceived susceptibility might be perceived when reading a health proposal message. If the content of a health proposal was supported, high response efficacy should be found, resulting in a decrease in perceived susceptibility. Given the connection between the EPPM and message repetition in persuasive health communication and the actual state we have been through when confronted with health proposals during the COVID-19 pandemic, we built an integrative model (Fig. 1) to investigate the factors and their associations concerning an individual’s behavior intention to conduct effective health proposals to prevent the infection.

The present study is of comprehensive significance in revealing the mechanism in the purview of public compliance with effective health proposals in the COVID-19 pandemic. It is unique and necessary in several aspects.

First, considering that COVID-19 is a real-life and ongoing public health emergency, some of its properties are fixed and thus cannot be manipulated. Take perceived threat, for example: In previous research, perceived threat was always considered together with fear because without extra fear-arousing materials (e.g., explanation of a certain disease in text or video), the perceived threat might not be notable enough to act as an independent variable. Unlike other unfamiliar diseases (e.g., melanoma), even though the degree to which one is influenced by the pandemic may vary, COVID-19 should be one of the immediate health threats in several countries and regions, including Japan, where the present study was conducted. Considering that fear-arousing information is no longer needed in the message, we are interested in exploring how perceived threat alone for COVID-19 affects behavior intention when there is no intentional manipulation of fear arousal (i.e., Hypothesis 8 below).

Second, there is still no research to test a model that combines the EPPM with message repetition. Although studies on similar topics have been conducted (Skilbeck, Tulips & Ley, 1977; Treise & Weigold, 2001; Shi & Smith, 2016), the results have not been analyzed in an integrated way or using structural equation modeling (SEM). To be specific, when supportive arguments toward the content of a recommended proposal are enhanced due to moderate repetition, it can also be interpreted as the change in response efficacy and perceived susceptibility in the EPPM. Nevertheless, this connection between the two research topics has rarely been made clear, causing difficulty in making an exact prediction of the results.

Third, it remains essential to confirm the validity of EPPM’s established construction, namely the sub-dimensions of perceived efficacy and perceived threat, to enhance compliance with public health guidelines in the COVID-19 pandemic. We assume better health compliance is due to higher perceived efficacy, but what does perceived efficacy mean? By learning more about its two indicators, self-efficacy and response efficacy, we can better understand why people choose to conform or not to certain health proposals. For instance, even though physical distancing is considered efficient in preventing infection (high response efficacy), the difficulty in keeping physical distance may vary between people based on their socializing needs (divergent self-efficacy), which results in different levels of health compliance. Similarly, if we are aware that perceived threat is indicated by perceived susceptibility and severity, we may know which properties to emphasize in education on COVID-19 to boost public health compliance.

Our hypotheses are as follows.

• 1) Hypothesis 1: Perceived efficacy has a positive effect on self-efficacy.

  H0: Perceived efficacy has no effect on self-efficacy.

• 2) Hypothesis 2: Perceived efficacy has a positive effect on response efficacy.

  H0: Perceived efficacy has no effect on response efficacy.

• 3) Hypothesis 3: Perceived threat has a positive effect on perceived susceptibility.

  H0: Perceived threat has no effect on perceived susceptibility.

• 4) Hypothesis 4: Perceived threat has a positive effect on severity.

  H0: Perceived threat has no effect on severity.

• 5) Hypothesis 5: Repetition has a positive effect on response efficacy.

  H0: Repetition has no effect on response efficacy.

• 6) Hypothesis 6a: Repetition has a positive effect on perceived susceptibility.

  H6b: Repetition has a negative effect on perceived susceptibility.

  H0: Repetition has no effect on perceived susceptibility.

• 7) Hypothesis 7: Perceived efficacy has a positive effect on behavior intention.

  H0: Perceived efficacy has no effect on behavior intention.

• 8) Hypothesis 8a: Perceived threat has a positive effect on behavior intention.

  H8b: Perceived threat has a negative effect on behavior intention.

  H0: Perceived threat has no effect on behavior intention.

Materials & Methods

The present study received approval from the psychological research ethics committee of the Faculty of Human-Environment Studies at Kyushu University, Fukuoka, Kyushu, Japan (approval number: 2019-034).

Results

Tables 2, 3 presents the correlations among variables with means and standard deviations. We chose a robust maximum likelihood estimator to perform SEM for the model in Fig. 1, which produced Satorra–Bentler rescaled χ² to correct non-normality-induced bias. The model yielded an acceptable fit (Kline, 2015; Bentler & Bonett, 1980): χ² (34, N = 326) = 196.367, p < 0.001; comparative fit index (CFI) = 0.949; Tucker–Lewis index (TLI) = 0.934, root mean square error of approximation (RMSEA) = 0.059 (90% CI [0.047–0.070], p > 0.05). The p values of the coefficients of concern were adjusted using the false discovery rate (Benjamini & Hochberg, 1995). The results of the theoretical model are presented in Fig. 2.

It was shown that perceived efficacy had positive effects on both self-efficacy (β = 0.883, p < 0.001) and response efficacy (β = 0.726, p < 0.001). Similarly, perceived threat was found to positively influence perceived susceptibility (β = 0.522, p < 0.001) and severity (β = 0.904, p < 0.001). However, we did not detect a significant influence of message repetition on either response efficacy (p = 0.110) or perceived susceptibility (p = 0.680). Furthermore, the results confirmed the positive influence of perceived efficacy on behavior intention (β = 0.766, p < 0.001), but no effect of perceived threat on behavior intention (p = 0.313).

Discussion

In the present study, we performed SEM to examine whether the EPPM could be generalized to a naturalistic context like the COVID-19 pandemic. Furthermore, we explored how a repeated health proposal is involved with the EPPM. According to the results of our analyses, the sub-dimensions of perceived efficacy toward the target message and perceived threat toward COVID-19 were confirmed, which supported Hypotheses 1–4. There was no evidence revealing that message repetition significantly influenced response efficacy (Hypothesis 5) or perceived susceptibility (Hypothesis 6). Furthermore, while we found a positive effect of perceived efficacy on behavior intention (Hypothesis 7), we did not detect a significant effect of perceived threat on behavior intention (Hypothesis 8).

Regarding the generalization of the EPPM in the COVID-19 context, most of our hypotheses were confirmed, except for the path from perceived threat to behavior intention. According to the EPPM, high threat and low efficacy would lead to a fear control process where people try to control their fear instead of taking efficient actions. In contrast, high threat and high efficacy would result in a danger control process where people take efficient actions to avert the threat. Moreover, low threat-high efficacy condition and low threat-low efficacy condition are not sufficient triggers for behavior change compared with the former two conditions. To the best of our knowledge, there is limited research using SEM to test the EPPM, except for a study conducted by Ort & Fahr’s (2018). Considering that the differences in analysis strategies may bring us new insights into the result of the non-significant path from perceived threat to behavior intention, we would take into account other analysis strategies in the previous research when discussing the results. Nevertheless, the following discussion was based on several post-hoc analyses, which were not registered in advance. We conducted the analyses simply for exploration, and thus any implication based on them should be treated with caution and confirmed through further investigation.

In several studies (Gore & Bracken, 2005; Witte, 1996), the average score of the six threat items was used as the score for perceived threat, and the average score of the six efficacy items was used as the score for perceived efficacy. Then the discriminating value was calculated by subtracting these two scores (Z_efficacy – Z_threat = discriminating value). When the discriminating value was negative (i.e., perceived threat outweighed perceived efficacy), the participant was considered to get involved in the fear control process. When the discriminating value was positive (i.e., perceived efficacy outweighed perceived threat), the participant was considered to take danger control actions. In other words, the formula emphasized the comparison between perceived threat and perceived efficacy rather than a single variable’s level (high/low) alone in determining health communication outcomes. Applying this strategy, we divided our participants in the no-repetition condition (which had the same design as the previous research) into two groups—danger control group and fear control group (N_danger = 70, N_fear = 82)—according to their discriminating values, and then conducted independent samples t-test on their behavior intention scores. The behavior intention of participants in the danger control group was significantly higher than that in the fear control group (t(150) = 3.65, p < 0.001), which was theoretically consistent with the basic frame of the EPPM. Given that these analyses were not registered beforehand and that the discriminating value approach has been criticized for its flawed conceptual logic (Ort & Fahr, 2018), we consider it not appropriate to make any conclusion about the EPPM’s validity in the present study. Instead, we suggest treating the result’s limited implication as a mere reference or a hint when discussing the relation between perceived threat and behavior intention.

There are also studies using the comparison between scores and the midpoint of the scale (four in the present research) to determine the level of perceived threat and perceived efficacy (Shi & Smith, 2016). When we applied this strategy, the results showed that only 16 out of 326 participants had low perceived efficacy toward the target health proposal message, and 3 out of 326 participants with low perceived threat toward COVID-19. The fact that few participants rated perceived threat low confirmed our expectation for the fixed properties of COVID-19—even with no extra fear-arousing material, its perceived threat stays at a high level. Moreover, in line with our pilot study’s results that people have favorable attitudes toward the target message, the message’s perceived efficacy was at a high level as well. In other words, we have mainly focused on the high efficacy-high threat condition in the present study.

The exploratory analyses mentioned above provide several possible explanations for the non-significant path between perceived threat and behavior intention in our theoretical model. As one possibility, a perceived threat may act as a judgment criterion, meaning that the danger control process will not start until the perceived threat reaches a certain level. Subsequently, the fluctuation in its value does not influence behavior intention anymore. If this is the case, the increase in perceived threat might not affect behavior intention, for the perceived threat was already at a high level in this study. Another possibility is that we have to take perceived threat’s interrelation with perceived efficacy into consideration when evaluating its effect on behavior intention. When perceived threat increases and perceived efficacy is controlled to stay constant, there are three kinds of possible changes in the process people get involved in when facing a health threat, namely switching from the danger control process to the fear control process; staying unchanged in the fear control process; or staying unchanged in the danger control process. While behavior intention should decrease in the switch from the danger control process to the fear control process, it is unclear how behavior intention would change when the process stays the same. Consequently, the relation between perceived threat and behavior intention would not be monotonic. In any case, we prefer to think that threat has its own function even if not reflected in number because studies (Witte, 1992; Gore & Bracken, 2005) have shown that if individuals receive a no-threat and high-efficacy message, the message did not influence them, which is not predicted by the EPPM.

In all cases, we should not expect to enhance people’s behavior intention through increasing COVID-19’s perceived threat. Instead, it is more reliable to count on the effect of perceived efficacy on behavior intention toward effective health proposals, which is consistent with the results from previous research (Earl & Albarracın, 2007; Mongeau, 1998; Peters, Ruiter & Kok, 2013; Ruiter et al., 2014; Tannenbaum et al., 2015; Ort & Fahr, 2018).

Regarding message repetition, neither response efficacy nor perceived susceptibility were found to be influenced by message repetition. Based on the differences in message repetition manipulation between the present study and previous ones, there may be several possibilities accounting for the results. First of all, considering that there are 24–72 h between two exposures to the message, the stimulus was too simple to be impressive, making the manipulation of message repetition invalid. For example, in Cacioppo & Petty (1979), a simple message was repeated several times in one experiment section with almost no interval. In the studies where the interval lasted for several days or weeks (Shi & Smith, 2016; Horowitz, 1969; Skilbeck, Tulips & Ley, 1977), the materials serving as stimuli were more complicated (e.g., 5-min video clip; eight-page pamphlet; 20-min lecture). Similarly, in a recent study investigating the relation between short health reminders and the change in behavior intention during the COVID-19 pandemic (Yonemitsu et al., 2020), no effect of the reminders was reported with a 1-week interval. Therefore, it is reasonable to conjecture that in the present study, the complexity of the health proposal message and the length of time interval did not make a good match to test message repetition’s effect. Furthermore, considering that COVID-19 has been a severe health threat in Japan—where the present study was conducted—the large amount of relevant information people received from sources other than the survey may have become an extraneous variable. On all accounts, it was not detected that mere repetition of a short piece of health proposal at 24–72h’s interval significantly influenced its response efficacy and COVID-19’s perceived susceptibility.

Conclusions

Supplemental Information