Participants and data collection
In this cross-sectional study, an online survey was conducted between October and November 2020. The target population included registered users of a social survey company (approximately 1.07 million users as of October 2020) between the ages of 20 and 69 years. The users registered with this company are an open-access panel recruited in various ways that does not rely on a specific platform. Among these users, 7,000 responses were requested to be collected. The sample size was determined to be as large as possible based on available funding resources. The registered participants, stratified by age group (e.g., 20–29, 30–39 years), sex (men/women), and education (junior high school/ high school/ 2-year college or vocational school/ 4-year college or more), were randomly selected so that the study population would match the population distribution in Japan. The survey company sent an e-mail to 23,188 registered users with the details of the study and a link to the response web page. The survey was closed when the number of respondents reached the target number in each stratum and when the total number of respondents reached 7,000.
This study was conducted after obtaining approval from the ethical review committees of Arakawa campus, Tokyo Metropolitan University (No. 20039), and Graduate School of Human Development and Environment, Kobe University (No. 455). All methods were carried out in accordance with the Japanese ethical guidelines for medical and biological research involving human subjects and the ethical standards of the Declaration of Helsinki. Consent for this study was obtained by presenting an explanatory statement at the beginning and requesting that only those who fully understood and consented to the statement would access the survey site. This survey has multiple research objectives and two papers have already been published using this survey [19, 23]. The first paper examined how each construct of the PA guidelines (awareness, knowledge, beliefs, and behavioral intentions), which consists of a logic model, was related to PA and sedentary behavior [23]. The second paper examined the relationship between health literacy, awareness of the PA guidelines, and physical activity levels [19]. Based on these results, the novelty of this present study is that it examined the psychological pathways from awareness of PA guidelines to PA behaviors, considering health literacy levels.
Measures
Participants’ characteristics
Based on previous studies, we chose participant characteristics associated with guideline awareness and knowledge, PA behavior, and health literacy [16,17,18,19,20,21,22,23,24]. Classification of educational background and household income was based on the information in other studies and surveys in Japan [19, 23, 28]. The characteristics of the participants were as follows: age, sex, current marital status, current working status, educational background (junior high school /high school /2-year college or vocational school /4-year college or more), and household income level (< \ 2 million / < \ 4 million / < \ 6 million / > \ 6 million).
Awareness of PA guidelines
Awareness of the Japanese PA guidelines was surveyed using the unprompted and prompted recalls. The unprompted recall method does not present clues, such as options or pictures, and the participant is asked to respond freely based on their memory [29]. For this recall, the participants were asked whether they had heard of the Japanese PA guidelines. Those who answered “yes” to this question were asked to freely provide the name of the guidelines. The first and second authors judged these free responses as correct, classified as “aware,” or incorrect. Prompted recall is a method to evoke the participants’ memory by presenting hints, such as options or pictures, and asking them to respond [18, 22, 30]. For this recall, two types of recall were used: written and illustrated recalls. In the written recall, we investigated the awareness of the Active Guide and “Plus-ten.” Participants were asked whether they had ever heard of these two terms. Among four options, including “I know the contents,” “I have heard of it but do not know the contents,” “I have never heard of it,” and “I learned it for the first time in this survey,” those who answered, “I know the contents” or “I have heard of it but do not know the contents” were classified as “aware” [18, 30]. In the illustrated recall, we presented an illustration of the Active Guide and asked the participants to respond either “yes” or “no” if they had ever seen or heard of a campaign that encouraged them to exercise or move their bodies. Those who answered “yes” were classified as “aware” [22]. Finally, we defined those who recognized any of the recall as “aware” of the PA guidelines.
Knowledge of PA guidelines
Next, we assessed knowledge of the PA guidelines. While awareness of the guidelines assessed whether participants had seen or recognized the guidelines, knowledge investigated whether they understood the recommended quantities, which are an essential component of the guidelines. Knowledge of the PA guidelines was based on numerical responses with open-text fields to the following three items: (1) recommended PA amount for individuals aged 18–64 years, (2) recommended amount of physical activity for individuals aged 65 years or older, and (3) PA amount that should currently be increased. For each item, 60 min/day, 40 min/day, and 10 min/day were considered correct answers. Among respondents aged 18–64 years, those who correctly answered both items (1) and (3) were classified as “adequate,” and those who incorrectly answered item (1) and/or (3) were classified as “inadequate.” Similarly, among respondents aged 65 years and older, those who correctly answered both items (2) and (3) were classified as “adequate”, and those who incorrectly answered item (2) and/or (3) were classified as “inadequate.”
Beliefs about PA guidelines
Eight items, which were developed based on the information in the Active Guide, were used to assess beliefs about PA guidelines (e.g., I think increasing the amount of time spent on physical activity, even if only a little, helps improve my health). The beliefs included items related to PA and health promotion and the development of PA habits (Additional file 1). The fitness of the model was confirmed by confirmatory factor analysis [23]. The goodness of fit of the model was root mean square error of approximation (RMSEA) = 0.070, comparative fit of index (CFI) = 0.985, and Tucker-Levis index (TLI) = 0.979, all of which were generally acceptable fit indices. The test–retest reliability conducted 2 weeks later was r = 0.45. Each item was surveyed using a 5-point scale. The calculated total score after each item ranged from 1 (completely disagree) to 5 (very strongly agree).
Behavioral intentions regarding PA participation according to recommended guidelines
For behavioral intentions, two items were developed: (1) “Do you intend to move your body for 60 min (18–64 years)/ 40 min (65 years and older) per day?” and (2) “Do you intend to move your body 10 min more than you do currently?” The options for both items were on a five-point scale from 1 (not at all) to 5 (very strongly). The average of the scores of the two items was calculated and used to score behavioral intention.
PA
PA was assessed using two questionnaires used in the Japan Public Health Center-based prospective Study (JPHC Study) [31] and in specific medical checkups and health guidance [32]. As the Active Guide prioritizes the message “Plus-ten” rather than the recommended levels of PA, the present study did not dichotomize the PA behaviors by meeting the recommended level. From the questionnaire of the JPHC study, moderate-to-vigorous-intensity PA (MVPA) was calculated using the method reported by Kikuchi et al. [33]. This questionnaire calculates MVPA using two items from occupation/household activities (walking/strenuous work) and three items from leisure time activities (walking fast/ light to moderate exercise such as golf, gardening/ vigorous exercise such as tennis, jogging, aerobics, and swimming) [31, 33]. Compared to 24-h activity records, the questionnaire showed moderate correlation (rho = 0.672) with MVPA, and moderate reliability (rho = 0.645) with retests conducted 3–6 months after the survey [33]. From the questionnaire of the specific medical checkups and health guidance, activity level was calculated using three items related to physical activity (exercise, daily physical activity, and walking speed). The participants were asked about their exercise, PA, and walking speed as follows: (1) “Do you engage in light sweaty exercise for at least 30 min for at least 2 days a week for at least 1 year?”, (2) “Do you engage in walking or similar PA for at least 1 h/day in your daily life”, and (3) “Do you walk faster than your peers of about the same age?”. The participants answered each question with a “yes” or “no,” and the activity level was classified into two groups: low (the number of “yes” responses was 0 or 1) and high (the number of “yes” responses was 2 or 3). It is because Kawakami et al. showed that this stratification approach yielded the highest discriminant validity (73% sensitivity and 68% specificity) to detect a recommended level of accelerometer-measured PA [32].
Health literacy
For health literacy, we used the Japanese version of the Communicative and Critical Health Literacy (CCHL) [34]. The CCHL consists of three assessment items of interactive health literacy (e.g. seeking information from various sources) and two assessment items of critical health literacy (e.g. considering the credibility of the information). Each item was answered using a five-point scale ranging from “Not at all” (1 point) to “Strongly agree” (5 points). The average score of the five items was used as the scale score (1–5 points). The internal consistency (Cronbach’s alpha) was 0.86 [34]. Construct validity of this scale was confirmed by examining its associations with health behaviors, coping styles, and somatic symptoms [34].
Order of measures
To minimize the influence of each survey item on the others, the following order was used for the survey: awareness of the Active Guide (unprompted recall), awareness of the Active Guide (written prompted recall), knowledge of the Active Guide, awareness of “Plus-ten” (written prompted recall), awareness of the Active Guide (illustrated prompted recall), behavioral intentions, and beliefs. Each question was displayed on a separate web screen. Additionally, it was not possible to return to the previous question.
Statistical analysis
Pathways from awareness of the guidelines to physical activity behavior among all respondents
The present study conducted the structural equation modeling to examine the PA pathways from the guideline awareness to behavior. The dependent variable was PA behavior. PA behavior was treated as a latent variable, defined by two observed variables: volume of MVPA (METs-hours per day) measured by the JPHC Study’s questionnaire [31, 33] and sufficient PA level (insufficient = 0, sufficient = 1) measured by Kawakami et al. [32]. PA beliefs and behavioral intention were also treated as latent variables from corresponding items. Awareness (not aware = 0, aware = 1) and knowledge (inadequate = 0, adequate = 1) of PA guidelines were treated as observed variables.
In the initial model, the present study examined the statistical significances of the following 10 standardized path coefficients based on the logic model: path from awareness to knowledge, belief, behavioral intention, and behavior; path from knowledge to belief, behavioral intention, and behavior; path from belief to behavioral intention and behavior, and path from behavioral intention to behavior (Fig. 1). Subsequently, we revised the model by removing insignificant paths. Chi-squared, CFI, TLI, RMSEA, and AIC values were evaluated as the model fit indices. The cutoff points for CFI/TLI and RMSEA were set as 0.95 and 0.06, respectively [35]. Using the bias-corrected bootstrap method (5,000 bootstrap samples), the present study estimated standardized direct, indirect, and total effects and 95% confidence intervals of awareness and knowledge of, beliefs about, and behavioral intention regarding PA behavior in the revised model. If the total and indirect effects of awareness on PA behavior were positive and statistically significant, the present study confirmed that there were mediating effects of this psychological pathway on PA behavior, which corresponded to the first research question.
Hypothesized models for pathways from PA guideline awareness to behavior. Awareness (not aware = 0, aware = 1) and knowledge (inadequate = 0, adequate = 1) of PA guidelines and PA amount (insufficient = 0, sufficient = 1) were dummy variables.
Moderating effects of health literacy on PA pathways from guideline awareness to behavior
Multi-group structural equation modeling was performed to examine the moderating effects of health literacy on the PA pathways from guideline awareness to behavior. The health literacy scores were dichotomized into low or high groups by the median split with the score of 3.40. The models were compared between low- and high-health literacy groups.
The present study implemented the multi-group structural equation modeling in three steps. In the first step, the unconstrained model was developed. The unconstrained model did not have any equality constraints for all parameters between the low- and high-health literacy groups. In the second step, we examined the model fit indices of the models constraining the following parameters between the low and high health literacy groups as equal: the path coefficients and variance within each latent variable (beliefs, behavioral intention, and PA behavior), variances of five main variables (observed variable of awareness, observed variable of knowledge, latent variable of beliefs, latent variable of behavioral intention, and latent variable of PA behavior), and path coefficients among main variables. For each parameter, a significant change in Chi-squared in the constrained model indicated that the corresponding parameter was different between the low- and high-health literacy groups. An insignificant change in Chi-squared in the constrained model indicated that equality constraint of the corresponding parameter between the two groups was reasonable. Thus, if the changes in Chi-squared were significant for the coefficients of the paths from awareness to knowledge, beliefs, intention, or PA behavior, the present study confirmed the moderating effects of health literacy on these psychological pathways, which addresses the second research goal. In the third step, the final model was constructed. The present study added the equality constraints identified in the second step on the unconstrained model. Using the bias-corrected bootstrap method (5,000 bootstrap samples), standardized direct, indirect, and total effects and 95% confidence intervals of awareness and knowledge of, beliefs about, and behavioral intention regarding PA behavior were estimated for the low- and high-health literacy groups.
Statistical significance was set at p < 0.05. Due to the nature of the web-based questionnaire survey, the database did not have missing values. The present study used AMOS v.25.0 (IBM Japan, Ltd., Tokyo, Japan) to perform the structural equation modeling.