Self-perceived food literacy in relation to the quality of overall diet and main meals: A cross-sectional study in Japanese adults

This cross-sectional study aimed to assess the relationship between self-perceived food literacy (SPFL) and quality of overall diet and main meals in Japanese adults. In total, 5998 adults aged 20 – 79 years were included in this analysis. The SPFL was assessed using the Japanese version of the 29-item Dutch SPFL scale (score range 1 – 5). Using validated dietary information, the Healthy Eating Index-2015 (HEI-2015) was calculated (score range 0 – 100). The mean SPFL was 3.18; the internal consistency of the overall scale was considered good (Cronbach ’ s alpha: 0.80). The mean HEI-2015 for overall diet was 50.4. The SPFL was significantly and positively associated with the HEI-2015. Using multiple linear regression, one point increase of SPFL corresponded to an increase in HEI-2015 by a point of 4.8 for overall diet, 6.2 for breakfast, 4.6 for lunch, and 3.6 for dinner (all P < 0.0001). Six of the eight domains of SPFL (i.e., food preparation skills, resilience and resistance, healthy snack styles, examining food labels, healthy budgeting, and healthy food stockpiling) were significantly associated with the HEI-2015 for overall diet. When the HEI-2015 for each meal was examined, the domains showing significant associations with all three meals included food preparation skills, healthy snack styles, and healthy budgeting. The healthy food stockpiling was associated with the HEI-2015 for breakfast and lunch, but not dinner. The social and conscious eating and daily food planning were associated with the breakfast HEI-2015 only, with the resilience and resistance associated with the dinner HEI-2015 only. In conclusion, the SPFL was cross-sectionally associated with the quality of overall diet and main meals in Japanese adults. Further observation and intervention studies are needed to confirm the associations observed here.


Introduction
Based on estimates from the Global Burden of Disease Study, poor dietary habits contribute to 22% of total deaths and 15% of disabilityadjusted life years globally each year (GBD 2017Diet Collaborators, 2019), surpassing those associated with any other risk factors such as tobacco smoking (Willett et al., 2019).These figures are even higher in East Asian countries including Japan (30% and 21%, respectively) (GBD 2017Diet Collaborators, 2019).Given that diet is a key modifiable risk factor, it is not surprising that the focus on prevention has become a priority (U.S.Department of Health and Human Services and U.S. Department of Agriculture, 2015), mainly through reducing the intake of foods high in sodium and added sugars (e.g., ultra-processed foods) (Lane et al., 2021a) and increasing the intake of plant foods, including whole grains, nuts, and legumes as well as fruits and vegetables (Schwingshackl et al., 2018;Willett et al., 2019).Eating behavior is complex and influenced by a wide range of factors, including personal, sociocultural, and environmental determinants (Mozaffarian, 2016).In the current food system, where energy-dense, nutrient-poor, ultra-processed foods dominate due to their palatability, availability, affordability, and market strategies, people face difficulties in achieving dietary intake recommendations (Baker et al., 2020).The extent to which people can make healthy food choices in such an environment is covered by the concept of food literacy (Sponselee et al., 2021).While various definitions of food literacy exist (Amouzandeh et al., 2019;Azevedo Perry et al., 2017;Truman et al., 2017), the most referenced definition was developed by Vidgen and Gallegos (Vidgen & Gallegos, 2014).They define food literacy as 'a collection of inter-related knowledge, skills and behaviors required to plan, manage, select, prepare and eat food to meet needs and determine intake' (Vidgen & Gallegos, 2014).In addition to this individual-level perspective, there have been recent attempts to look at food literacy in a broader sense, comprehensively considering cultural, community, societal, political, and environmental values of food (Rosas et al., 2022;Vettori et al., 2019;Yoo et al., 2022).
To our knowledge, however, no previous research has examined different aspects of food literacy in relation to the nutritional quality of each meal type (such as breakfast, lunch, and dinner), in addition to the overall diet quality.This kind of research is of paramount importance in the Japanese context.First, despite the perception that the Japanese diet is healthier than other diets (Sasaki, 2020), an analysis based on national dietary survey data showed that its overall diet quality as assessed using the Healthy Eating Index-2015 (HEI-2015) (Krebs-Smith et al., 2018;Panizza et al., 2018;Reedy et al., 2018) is comparable to that of Americans (mainly due to the high intake of refined grains and sodium and the low intake of whole grains and fruits) (Murakami et al., 2020a).Second, our previous cluster analysis based on approximately 1500 days of dietary record data showed that breakfast, lunch, and dinner had distinctive dietary patterns (Shinozaki et al., 2020).More specifically, we found two types of breakfast (i.e., 'bread-based' and 'rice-based') and five types of lunch (i.e., 'bread', 'rice-based', 'ramen' (Chinese wheat noodle), 'udon/soba' (Japanese wheat noodle/buckwheat noodle), and 'sushi/rice bowl dishes') (Shinozaki et al., 2020).This indicates that Japanese breakfast and lunch are largely characterized by the choice of staple foods (rice, bread, and noodles).In contrast, no specific dietary pattern could be found for dinner, with the 'miscellaneous' pattern predominating (81%) (Shinozaki et al., 2020), suggesting a wide variety of food combinations in dinner.Interestingly, however, diet quality (as assessed using the HEI-2015) was highest for dinner (mean: 53.0), followed by lunch (mean: 48.9) and breakfast (mean 45.0) (Murakami et al., 2022a).Moreover, different meals are associated with different motivational factors (Chambers et al., 2016;Peters et al., 1995;Phan andChambers, 2016, 2018).For example, one study reported that health and convenience criteria were more important predictors of preferences for morning meals than for midday and evening meals, while general 'liking' was most heavily weighted for midday and evening meals (Peters et al., 1995).In another study, choosing foods for breakfast were driven more by motives of price, health, and convenience, while choosing foods for lunch were driven more by motives of price and habits (Phan & Chambers, 2018).In contrast, food choices for dinner were driven more by variety seeking, traditional eating, and sociability (Phan & Chambers, 2018).Although findings are not necessarily consistent across studies, there is a consensus that people choose foods for different meals with different motivations.To sum up, there are substantial differences in food choices and combinations (Guan et al., 2018;Murakami et al., 2022a;Myhre et al., 2015;Shams-White et al., 2021;Shinozaki et al., 2020) as well as motivational factors (Chambers et al., 2016;Peters et al., 1995;Phan andChambers, 2016, 2018) between meal types (breakfast, lunch, and dinner), suggesting that it is plausible that different aspects of food literacy may influence the nutritional quality of each meal.
Therefore, the objective of this cross-sectional study was to extensively investigate the relationship between food literacy and the overall diet quality, as well as the quality of specific meal types.We formulated the following research questions.(1) Is food literacy associated with the overall diet quality, independent of other factors, and if so, which elements of food literacy are most influential in determining the overall diet quality?(2) Is food literacy associated with the quality of specific meal types, and if so, is there the difference in the strength of association among meal types?(3) Which elements of food literacy are most influential in determining diet quality for different meal types?

Study procedure and participants
The present cross-sectional analysis was based on data from an online questionnaire survey.Fig. 1 shows a flow diagram of participant selection in the present analysis.The target sample consisted of 6600 Japanese adults aged 20-79 years, including not only the general public but also health professionals allied to nutrition (dietitians, registered dietitians, doctors, dentists, etc.), with the intention of ensuring that food literacy and diet quality were well distributed within the data set (Murakami, Shinozaki, Okuhara, McCaffrey, & Livingstone, 2024).Exclusion criteria were individuals outside the 20-79 year-age group and those working in health professions unrelated to nutrition (e.g., veterinarians, dental hygienists, assistant nurses, clinical psychologists, and nurse practitioners).The reason for excluding dental hygienists and including dentists is that in Japan, dental hygienists can be certified in fewer years (3 years) than doctors, and dentists can be certified in the same number of years (6 years) as doctors.The main purpose of this K. Murakami et al. survey was to examine the factors related to the way Japanese people in general and nutrition-and diet-related professionals such as dietitians are exposed to nutrition and diet-related information in mass and social media (Murakami, Shinozaki, Okuhara, McCaffrey, & Livingstone, 2024).In a representative sample of Japanese adults, the prevalence of nutrition and diet-related information seeking was 52.3% for television, 5.0% for radio, 18.1% for newspapers, 23.1% for books/magazines, 16.6% for websites, 7.7% for social networking sites, and 13.5% for medical institutions (Ministry of Health and Labour and Welfare, 2020).Calculating the sample size based on the mean prevalence (19%), 95% confidence interval (CI), and 1% precision of the nutrition-and diet-related information seeking yielded 5913 individuals (Naing et al., 2006); assuming a 10% dropout rate and rounding, the final sample size was determined to be 6600.
Data collection was conducted by an internet research agency, Rakuten Insight (https://insight.rakuten.com/).Participants were recruited from the registered panelists of the research agency.An email with a survey cooperation request and a webpage link to the survey was sent to a randomized list (n = 676,329) of registered panelists aged 20-79 years (n = 2,603,155).The study was explained at the beginning of the webpage, and only individuals who agreed to participate proceeded to the screening stage based on age, sex, and occupation (n = 76,845).Participants could proceed to the main survey only if there was a vacancy in the relevant sampling category as shown in Fig. 1.As a result, 7722 individuals proceeded to the main survey, of whom 1122 did not complete all the question items.The data collection started on 10 February 2023, and completed on 16 March 2023 when the target number of individuals who completed the main survey was met in all sampling categories (total n = 6600).For analysis, we excluded individuals whose answers were considered unreliable based on the answer to the question 'This question is for the purpose of investigating "unreliable" answers when responding to the survey.Please select neither agree or disagree from the following options'.The response options were strongly agree, agree, neither agree or disagree, disagree, or strongly disagree.This resulted in the exclusion of 314 individuals whose response was either strongly agree, agree, disagree, or strongly disagree.Further, we excluded individuals whose data were considered unreliable with regard to body height (≥200 cm; n = 2) or energy intake (<800 or >4200 kcal/day for males and <500 or >3500 kcal/day for females; n = 286 (Bertoia et al., 2016)).Consequently, the final analysis sample comprised 5998 individuals (2687 males and 3311 females) aged 20-79 years.All the questionnaire items were prepared by the first, second, and third authors.
The study was conducted according to the guidelines laid out in the Declaration of Helsinki, and all procedures involving human participants were approved by the Ethics Committee of the University of Tokyo Faculty of Medicine (protocol code: 2022288NI; date of approval: 13 January 2023).Informed consent was obtained online from all individuals involved in the study.

Assessment of food literacy
The self-perceived food literacy (SPFL) was assessed using the Japanese version of the SPFL scale.First, the original SPFL scale, which was developed in Dutch with the English translation available (Poelman et al., 2018), was translated into Japanese by the second author.The Japanese translation was checked, and when necessary modified, by the first author.Then, back translation (to English) was conducted using DeepL Translator, a neural machine translation service (https://www.deepl.com/translator).The first and second authors further checked and then approved the Japanese version and its back-translated version.Finally, the backward translation was reviewed by the original researcher (Poelman et al., 2018), based on which relevant modifications were made such that the translated version better reflected the original scale.
The SPFL scale is an expert-based and theory-driven tool for measuring food literacy with respect to healthy eating (Poelman et al., 2018).The validity of the original Dutch version has been described elsewhere (Poelman et al., 2018).The SPFL scale, consisting of 29 items, measures 8 domains of food literacy: food preparation skills (6 items), resilience and resistance (6 items), healthy snack styles (4 items), social and conscious eating (3 items), examining food labels (2 items), daily food planning (2 items), healthy budgeting (2 items), and healthy food stockpiling (4 items) (Poelman et al., 2018).The questionnaire (in English) has been described in full elsewhere (Poelman et al., 2018).Participants were asked to answer all the questions based on a 5-point Likert scale (1 = 'not at all/never' to 5 = 'yes/always').The total score was calculated as the average of all the items, with negative items reversed, indicating that the higher the score, the higher food literacy is (possible scores ranging from 1 to 5) (Poelman et al., 2018).The score for each domain was also calculated as the sum of the scores divided by the number of items.In the present study population, the Cronbach's alpha for the assessment of internal consistency was 0.80 for total score, 0.88 for food preparation skills, 0.65 for resilience and resistance, 0.69 for healthy snack styles, 0.47 for social and conscious eating, 0.90 for examining food labels, 0.76 for daily food planning, 0.83 for healthy budgeting, and 0.77 for healthy food stockpiling, which was considered generally good or adequate (except for social and conscious eating) but also comparable to observations in the Dutch adults (Poelman et al., 2018).

Assessment of diet quality
In the present study, we used the HEI-2015(Krebs-Smith et al., 2018;Panizza et al., 2018;Reedy et al., 2018) as a measure of diet quality.The HEI-2015 is a 100-point scale to assess compliance with the 2015-2020 Dietary Guidelines for Americans (U.S.Department of Health and Human Services and U.S. Department of Agriculture, 2015), with a higher score indicating a better quality of overall diet.The HEI-2015 consists of nine adequacy components (e.g., total fruits, total vegetables, and whole grains) and four moderation components (e.g., sodium).
Information on dietary habits during the preceding month was collected using a short version of the Meal-based Diet History Questionnaire (MDHQ) (Murakami et al., 2022b(Murakami et al., , 2022c(Murakami et al., , 2023)).Briefly, the MDHQ consists of three parts: (1) consumption frequency (during the preceding month) of major food groups for each of main meals (breakfast, lunch, and dinner) and snacks (morning snack, afternoon snack, and night snack) separately (113 questions); (2) relative consumption frequency of sub-food groups within the major food groups (72 questions) with questions on consumption frequency and portion size for alcoholic beverages (10 questions); (3) general eating behaviors (22 questions) (Murakami et al., 2022b(Murakami et al., , 2022c(Murakami et al., , 2023)).In contrast, the short MDHQ (hereafter referred to as sMDHQ) asks only about consumption frequency of major food groups (except for non-caloric beverages) for the main meals (which were derived from Part 1 of the MDHQ; 66 questions) and alcoholic beverages (which were derived from Part 2 of the MDHQ; 10 questions).
Prior to the present study, the validity of sMDHQ was examined using a simulated analysis using a different dataset.The HEI-2015 was calculated following simulation procedures for 111 women and 111 men who completed the MDHQ (Murakami et al., 2022b(Murakami et al., , 2022c(Murakami et al., , 2023)).In the simulation, deleted items from Part 1 were deemed as no-consumption for all individuals, while the intermediate response was used for all individuals for items from Part 2 and Part 3. We then examined the correlation between the HEI-2015 obtained by this simulation procedure and that derived from the 4-non-consecutive-day weighed dietary record collected in the same population (Murakami et al., 2022b(Murakami et al., , 2022c(Murakami et al., , 2023b)).For women, the Spearman correlation coefficients between the HEI-2015 derived from the sMDHQ and that from the dietary record were 0.47 for overall diet, 0.54 for breakfast, 0.41 for lunch, and 0.39 for dinner.For men, the corresponding values were 0.62 for overall diet, 0.65 for breakfast, 0.38 for lunch, and 0.32 for dinner.These results suggest that the sMDHQ is comparable with the MDHQ in terms of the ability for ranking individuals according to the quality of overall diet, breakfast, lunch, and dinner (Murakami et al., 2023a).
In the present study, the HEI-2015 was calculated based on estimates of dietary intake derived from the sMDHQ.Component scores needed for the calculation of HEI-2015 were calculated using the Japanese version (Murakami et al., 2019a) of the US Food Patterns Equivalents Database (Bowman et al., 2014), except for fatty acids and sodium, for which the 2015 version of the Standard Tables of Food Composition in Japan (Council for Science and Technology; Ministry of Education, Culture, Sports, Science and Technology, 2015) was used.As described in detail elsewhere (Murakami et al., 2019a), we calculated the HEI-2015 component scores based on energy-adjusted values of dietary intake, namely amount per 1000 kcal of energy or percentage of energy, except for fatty acids, and then we summed up these scores to obtain the HEI-2015 score.These calculations were done for each meal type, and the score for overall diet was calculated using the sum of the intake of each meal type.

Assessment of other variables
In this study, sex (assigned at birth) was self-selected as either male or female.Age (in years) was also self-reported and categorized into three groups (20-39, 40-59, and 60-79).Based on body mass index (BMI; in kg/m 2 ) calculated using self-reported weight and height, three categories of weight status were created: underweight (<18.5),normal weight (≥18.5 to <25), and overweight (≥25) (World Health Organization, 2000).Self-reported information on the following variables was also used in this study (categorization shown in parentheses): education (junior high-school or high school, junior college or technical school, university or higher, and other), household income (<4, 4-7, and >7 million yen), employment status (none, student, part-time job, and full-time job), marital status (unmarried, married, and do not want to answer), living alone (no and yes), presence of chronic disease (no and yes), and smoking status (never, past, and current).Further, participants were categorized according to nutrition-and health-related occupation: none (i.e., general public), non-governmental qualification related to food and nutrition, media, dietitians/registered dietitians, doctors/dentists, and other health professionals (i.e., nurses, midwives, public health nurses, and pharmacists).Healthy eating motivation was assessed using the Japanese version of the 7-item original scale developed in Ireland (Naughton et al., 2015).The development process of the Japanese scale was similar to that for the SPFL scale described above; thus, the final back-translated version was reviewed by the lead researcher involved in the development and validation of the original English version (Naughton et al., 2015).The items were measured on a 7-point Likert scale (1 = 'strongly disagree' to 7 = 'strongly agree').The healthy eating motivation score was calculated as the average of all the items, with negative items reversed, indicating that the higher the score, the higher healthy eating motivation is (possible scores ranging from 1 to 7).In the present study population, the Cronbach's alpha for the seven items was 0.59, which was considered satisfactory (Taber, 2018) but low compared with that observed in the previous study (0.81) (Naughton et al., 2015).For analysis, participants were grouped into approximate quartiles according to the healthy eating motivation score.

Statistical analysis
All statistical analyses were performed using SAS statistical software (version 9.4, SAS Institute Inc., Cary, NC, USA).We considered 2-tailed P values < 0.05 statistically significant.Descriptive data are presented as means and standard deviations (SDs).Basic characteristics were shown for the whole sample and for males and females separately and for the general public and health professionals (i.e., dietitians/registered dietitians, doctors/dentists, and other health professionals) separately; differences were examined using independent t-test.Then, the mean values of the SPFL total score and HEI-2015 for overall diet were calculated according to categories of selected characteristics (covariates); difference was examined using either independent t-test or analysis of variance followed by Bonferroni's post hoc test.As the main analysis, we examined the association between the SPFL total score and HEI-2015 using multiple linear regression.Separate analyses were conducted for the HEI-2015 for overall diet, breakfast, lunch, and dinner.Covariates were determined prior to analysis on the basis of previous studies (Bonaccio et al., 2013b;Lavelle et al., 2020;McGowan et al., 2016;Murakami et al., 2023;Park et al., 2020;Poelman et al., 2018;Sexton-Dhamu et al., 2021;Tani et al., 2020;Zoellner et al., 2011) and included age, sex, weight status, education, household income, employment status, marital status, living alone, presence of chronic disease, smoking status, nutrition-and health-related occupation, and the healthy eating motivation score.The variance inflation factor scores for any variable in any model (range: 1.04-2.90)were within acceptable limits (<10) (O'Brien, 2007), suggesting that multicollinearity was not an issue.Regression coefficients (β) with 95% CIs were calculated as the change of the HEI-2015 with 1-point increase of the SPFL total score (and for other variables as the difference between each category and the reference category).The same analyses were also conducted for males and females separately and for the general public and health professionals separately, as well as with the SPFL treated as a categorical variable (quartiles).Finally, we examined the associations of SPFL domain scores with the HEI-2015 for overall diet, breakfast, lunch, and dinner (based on the total sample).

Results
This analysis included 5998 individuals (2687 males and 3311 females) aged 20-79 years (Table 1).Mean age was 46.8 years (SD: 15.1).The mean SPFL total score was 3.18 (SD: 0.43).The mean HEI-2015 for overall diet was 50.4 (SD: 7.5).As shown in Supplemental Table 1, males had higher mean age and BMI than females.Conversely, females had higher mean SPFL total score and domain scores (except for no difference in healthy snack styles and a lower mean score of resilience and resistance) and higher mean HEI-2015 for overall diet, breakfast, lunch, and dinner than males.When compared with health professionals, the general public had higher mean age and BMI (Supplemental Table 1).Conversely, the general public had lower mean SPFL total score and SD, standard deviation.a Calculated using self-reported body weight and height.b Scores ranging from 1 to 5, with a higher score indicating a higher food literacy.The total score was calculated as the average score of all the 29 items.c Scores ranging from 0 to 100, with a higher score indicating a higher diet quality.
K. Murakami et al. domain scores (except for no differences in healthy snack styles and healthy food stockpiling and a higher mean score of resilience and resistance) and lower mean HEI-2015 for overall diet, breakfast, lunch, and dinner.
Table 2 shows the SPFL total score and HEI-2015 for the overall diet according to categories of selected characteristics (covariates).Higher mean values for SPFL total score and HEI-2015 for overall diet were associated with being older, female sex, underweight, intermediate education (i.e., junior college or technical school), higher household income, being unemployed, being married, living with someone (only for the HEI-2015), presence of chronic disease, never smoking, being dietitians/registered dietitians, and higher healthy eating motivation.
After adjustment for potential confounding factors, the SPFL total score was significantly and positively associated with the HEI-2015 for overall diet and for each meal (Table 3).A one-point increase of SPFL total score corresponded to an increase in HEI-2015 by a point of 4.84 (95% CI: 4.37, 5.31) for overall diet, 6.22 (95% CI: 5.11, 7.33) for breakfast, 4.57 (95% CI: 3.81, 5.33) for lunch, and 3.58 (95% CI: 2.96, 4.20) for dinner.As shown in Supplemental Table 2, other factors significantly associated with a higher HEI-2015 for overall diet included a Scores ranging from 1 to 5, with a higher score indicating a higher food literacy.b Scores ranging from 0 to 100, with a higher score indicating a higher diet quality.c On the basis of independent t-test for sex, living alone, and presence of chronic disease and one-way analysis of variance (ANOVA) for other variables.When the overall P from ANOVA was <0.05, a Bonferroni's post hoc test was performed; different letters (i.e., a, b, c, and d) in the column P values mean significant difference between categories (P < 0.05).d Underweight, normal weight, and overweight were defined as participants having a body mass index (in kg/m 2 ) of <18.5, ≥18.5 to <25, and ≥25, respectively.
K. Murakami et al. older age, female sex, underweight (compared with overweight), higher education, higher household income, unemployment status (compared with part-time job and full-time job), presence of chronic disease, never smoking (compared with current smoking), dietitians/registered dietitians, doctors/dentists, and non-governmental qualification related to food and nutrition (compared with the general public), and higher motivation scores for healthy eating.The association between SPFL total score and HEI-2015 for overall diet and for each meal remained statistically significant when the analysis was restricted to males, females, the general public, and health professionals, regardless of treatment for the total SPFL score, or a combination of both conditions (continuous SPFL scores for Supplemental Table 3 and quartile SPFL scores for Supplemental Table 4).However, the effect size was consistently large in females compared with males, while the difference of effect size was unclear or inconsistent between the general public and health professionals.
Table 3 also shows the associations of SPFL domain scores with HEI-2015 for overall diet, breakfast, lunch, and dinner (in the total sample).Six of the eight domains of SPFL (i.e., food preparation skills, resilience and resistance, healthy snack styles, examining food labels, healthy budgeting, and healthy food stockpiling) were significantly and positively associated with the HEI-2015 for overall diet.For breakfast, all domains except for resilience and resistance and examining food labels showed significant positive associations.The following domains were positively associated with the HEI-2015 for lunch: food preparation skills, healthy snack styles, healthy budgeting, and healthy food stockpiling.The domains positively associated with the HEI-2015 for dinner included food preparation skills, resilience and resistance, healthy snack styles, and healthy budgeting.

Discussion
To our knowledge, this is the first study to comprehensively explore the associations between various aspects of food literacy and the overall diet quality, as well as the quality of individual meal types.In this crosssectional study in Japanese adults, we found that the SPFL total score was positively associated with the HEI-2015 for overall diet, breakfast, lunch, and dinner, even after adjustment for a wide range of factors associated with diet quality.More specifically, six of the eight domains of SPFL (i.e., food preparation skills, resilience and resistance, healthy snack styles, examining food labels, healthy budgeting, and healthy food stockpiling, but not social and conscious eating and daily food planning) were significantly associated with the HEI-2015 for overall diet.When the HEI-2015 for each meal was examined, different domains of SPFL showed significant associations, with breakfast most notably (food preparation skills, healthy snack styles, social and conscious eating, daily food planning, healthy budgeting, and healthy food stockpiling) compared to lunch (food preparation skills, healthy snack styles, healthy budgeting, and healthy food stockpiling) and dinner (food preparation skills, resilience and resistance, healthy snack styles, and healthy budgeting).
Our finding that a higher SPFL total score was independently associated with a higher HEI-2015 for overall diet is consistent with the majority of previous studies in which a higher composite measure of food literacy was associated with higher intakes of healthy foods (vegetables (Kawasaki et al., 2022;Lee et al., 2023;Poelman et al., 2018;Yoo et al., 2023), fruits (Lee et al., 2023;Poelman et al., 2018;Yoo et al., 2023), whole grains (Lee et al., 2023;Yoo et al., 2023), fish/shellfish (Kawasaki et al., 2022;Poelman et al., 2018)), lower intakes of less healthy foods (such as biscuits, pizza, and sugar-sweetened beverages) (Poelman et al., 2018), and higher scores of diet quality (Blaschke et al., 2023;Boslooper-Meulenbelt et al., 2021;Park et al., 2020).Notably, we found that the strength of the association for the SPFL total score was larger than any other factors included in the model such as sex, age, education, household income, occupation, and motivation for heathy eating.Taken together, these results suggest that food literacy is a key  a Models with the SPFL total score or domain scores, age, sex, weight status, education, household income, employment status, marital status, living alone, presence of chronic disease, smoking status, nutrition-and health-related occupation, and healthy eating motivation score as the explanatory variables and the HEI-2015 for overall diet, breakfast, lunch, or dinner as the response variable.The HEI-2015 ranges from 0 to 100, with a higher score indicating a higher diet quality.The SPFL score ranges from 1 to 5, with a higher score indicating a higher food literacy.
b All of the SPFL domain scores were entered simultaneously into each model for overall diet, breakfast, lunch, and dinner.
K. Murakami et al. factor to consider for both those who are eager and those who need to improve diet quality.An interesting finding is that the association between the SPFL total score and the HEI-2015 for overall diet was stronger in females than males.Reasonable explanations for this finding may include the fact that females are more responsible for cooking and possibly grocery shopping than males in many Japanese households (Tani et al., 2020), that females tend to be exposed to stronger sociocultural norms regarding body shape and thus foods (Buote et al., 2011), and that females are more involved and preoccupied with food than males (Konttinen et al., 2021).The present finding is generally consistent with our previous analysis, in which nutrition knowledge and cooking skills (important facets of food literacy) were positively associated with diet quality in females but not males (Murakami et al., 2023).In fact, the present study showed positive associations of some of the food literacy domains relevant to nutrition knowledge (examining food labels) and cooking skills (food preparation skills) with diet quality.We do not know why the present study (but not our previous study) (Murakami et al., 2023) observed a significant association in males as well, but the reason may be due to methodological differences in terms of measurement of both food literacy and diet quality, as well as due to characteristics of the present male participants such as health-conscious nature and higher socioeconomic status.Meanwhile, we observed that the association between the SPFL total score and the HEI-2015 for overall diet was generally similar between the general public and health professionals.This may not be surprising, however, considering that although both the SPFL and the HEI-2015 were on average high in the latter group, these differences were not large.A notable exception was that the general public had a higher mean score of resilience and resistance compared to health professionals, as well as no difference in healthy snack styles and healthy food stockpiling.This might reflect the fact that health professionals (e.g., nurses) also face similar barriers to self-care experienced by the general population, particularly busy schedules, competing demands on their time, availability of resources, and work-related stress (Akter et al., 2019;Muhlare & Downing, 2023).
We further found the positive associations between the SPFL total score and the HEI-2015 for each individual meal (breakfast, lunch, and dinner).This observation is important given considerable variations of food choice and food combination between meal types (Guan et al., 2018;Murakami et al., 2018;Murakami et al., 2020b;Murakami et al., 2022a;Myhre et al., 2015;O'Hara & Gibney, 2021;Shams-White et al., 2021) and relatively low correlations between diet quality of each meal (Murakami et al., 2022a).The association with the SPFL total score was strongest for the diet quality for breakfast, followed by lunch and dinner.The reasons for this hierarchy remain unknown but may be because it is easier to be consistent with eating healthy breakfast as it is typically consumed at home and there are specific foods marketed for breakfast, but for lunch and dinner there is typically more choice/effort into preparing and planning.Furthermore, prior research has shown that breakfast food choices are more influenced by health and convenience motives, whereas different motives dominate for lunch (likes/dislikes, price, and habits) and dinner (likes/dislikes, variety seeking, traditional eating, and sociability) (Peters et al., 1995;Phan & Chambers, 2018).This may explain the stronger association between breakfast quality and food literacy than lunch or dinner.Another explanation may be several characteristics of breakfast (compared with lunch and dinner), including its smaller mean but larger SD for HEI-2015 (Murakami et al., 2022a), the most consistent food choice pattern within individuals (Schwedhelm et al., 2018), and relative similarity to snacks in terms of percentage of energy consumed (Murakami et al., 2019b(Murakami et al., , 2022a)), the time spent per eating occasion (Murakami et al., 2022d), and nutritional quality (Murakami et al., 2022a) at least in Japanese.These inferences are consistent with the SPFL domain scores that showed significant associations with the HEI-2015 for breakfast.Particularly, the domain healthy snack styles showed the strongest association, followed by healthy budgeting, which seems reasonable considering somewhat similar natures between breakfast and snacks as mentioned above.Additionally, the domain daily food planning was associated with the HEI-2015 for breakfast only.Given that time restraint is the largest obstacle for eating healthy breakfast at least for Japanese (Melby & Takeda, 2014), daily food planning may be much more important for breakfast, compared with lunch and dinner.Alternatively, previous studies have shown that compared to lunch and dinner, breakfast food choices are more influenced by health and convenience motives (Peters et al., 1995;Phan & Chambers, 2018), which may explain the positive association between daily food planning and the quality of breakfast.For lunch and dinner, the domains most strongly associated with diet quality were food preparation skills and healthy snack styles.This may suggest that someone who has higher food preparation skills are more likely to produce higher quality meals (McGowan et al., 2016;Murakami et al., 2023;Tani et al., 2020), but also be more likely to choose healthier snacks.Mainly because of low validity of the MDHQ for assessing the quality of snacks (Murakami et al., 2023a), the present study exclusively focused on three main meals.Nevertheless, snack intake is quite different from three main meals (Hess et al., 2016) in terms of, for example, timing (Murakami et al., 2022d), food choice and portion size (Murakami et al., 2019b(Murakami et al., , 2022a)), and motivational factors (Chambers et al., 2016;Peters et al., 1995;Phan & Chambers, 2016, 2018).Snacks consumed by Japanese adults contribute to, on average, 8% to 10% of total energy, characterized by a variety of foods including confectioneries, soft drinks, tea/coffee, dairy products, and fruits (Murakami et al., 2019b(Murakami et al., , 2022a)).Thus, further studies for examining the association between food literacy and snack intake and behavior would be of interest.
The major strength of our study was the use of a more recent instrument for measuring food literacy (Poelman et al., 2018) to examine the association with the quality of each meal as well as the overall diet quality (Murakami et al., 2023).This study is novel and highly relevant to the growing interest in meal patterning or chrono-nutrition in the discussion of nutrition, health, and chronic disease prevention (Almoosawi et al., 2016;Phoi et al., 2022;St-Onge et al., 2017;van der Merwe et al., 2022).Furthermore, the present analysis considered a variety of potential confounding factors in a large sample, which also contributed to ensure that stratification by sex and nutrition-and health-related occupation (the most important confounding factors in this study) did not substantially alter the main results.
However, this study has several limitations.First, our sample consisted of individuals registered with an internet research agency, and only a small percentage of these individuals expressed interest in this study.In addition, based on the main objective of the survey, it was determined that about half of the sample were nutrition-and healthrelated professionals.Therefore, the participants of this study are not a nationally representative sample of the general Japanese population.For example, the education level and household income of the present participants were higher than those of the national representative sample (education: 54.6% for junior high school or high school, 20.8% for junior college or technical school, and 24.6% for university or higher (Statistics Bureau, 2018); household income: 45.0% for <4 million yen, 26.9% for ≥4 to <7 million yen, and 28.0% for ≥7 million yen (Ministry of Health, Labour & Welfare, 2017)).Moreover, our participants appeared different from the national representative sample, when it comes to mean, BMI and HEI-2015 for overall diet (males: 23.9 kg/m 2 and 51.3, respectively; females: 22.5 kg/m 2 and 52.9, respectively) (Ministry of Health, Labour & Welfare, 2017; Murakami et al., 2020a).Further research in a more representative sample is thus warranted.
Second, all variables used in this study were based on self-report.In particular, while diet quality was assessed using energy-adjusted values derived from an established dietary assessment tool (Murakami et al., 2023a) to minimize measurement error (Murakami et al., 2008), the measurement of dietary intake cannot be done without error (Livingstone & Black, 2003;Subar et al., 2015).Also, the HEI-2015, originally developed and validated for the American population (Krebs-Smith et al., 2018;Panizza et al., 2018;Reedy et al., 2018), may not fully K.Murakami et al. reflect the dietary patterns of the Japanese population (Murakami et al., 2020a).Nevertheless, we note that the efficacy of the HEI-2015 in assessing the overall diet quality of Japanese has been supported by our previous analyses: a higher total score in the HEI-2015 was associated with favorable patterns of the overall diet, including higher intakes of dietary fiber and key vitamins and minerals and lower intakes of saturated fats, added sugars, and sodium (Murakami et al., 2020a(Murakami et al., , 2020c)).Furthermore, cultural differences between Japan and the Netherlands were not taken into account during the development process of the Japanese version of the SPFL scale.For example, the mean SPFL total score was lower in this population than in samples of Dutch (3.83) (Poelman et al., 2018) and Italian (3.50) (Luque et al., 2022) adults.The reason is unclear but may be the low score of healthy snack styles derived from a low intake of snacks in Japanese (Murakami et al., 2019b(Murakami et al., , 2022a) ) as mentioned above.Consequently, although the internal consistency of the SPFL total score was good and comparable to that observed in the original Dutch study (Poelman et al., 2018), this tool may not be optimal for use with the Japanese population.Nevertheless, our findings in support of the hypothesis that food literacy is positively associated with diet quality can be considered valuable evidence regarding the criterion validity of the SPFL scale.It should be noted, however, that both food literacy and dietary intake were assessed consecutively on the same online questionnaire.This may have caused bias because responses to the food intake questions were primed by the initial food literacy items in the questionnaire and thus more in line with the initial responses (Braun et al., 2012;Poelman et al., 2018).However, these are common limitations in the field of behavioral nutrition research using surveys that measure behavioral determinants and eating behaviors (Poelman et al., 2018).Similarly, this study was conducted under a cross-sectional design.Because the temporality of associations is unknown, it is not possible to address causality or its direction.Nevertheless, it is logical to assume that food literacy necessarily precedes diet quality, so it is unlikely that the association observed here is due to reverse causality.
Third, primarily because the SPFL scale used in this study is designed to measure food literacy regarding healthy eating (Poelman et al., 2018), this study focused only on diet quality as a first step.Because food literacy can be situated in a broader paradigm of food and health, future studies should encompass multiple indicators to measure against food literacy (Amouzandeh et al., 2019), including ultra-processed food intake (Lane et al., 2021b), food security (Godfray et al., 2010), sustainable eating (Macdiarmid et al., 2012), and food waste (Lebersorger & Schneider, 2011).Finally, although we made adjustment for a variety of variables, we cannot rule out the possibility of residual confounding.
In conclusion, after adjustment for a variety of factors associated with diet quality, higher SPFL was associated with higher quality of overall diet and main meals in Japanese adults.Further, for each meal, different domains of SPFL showed significant associations.The findings of this study are important for nutrition education and behavioral interventions to improve the quality of diet in general population.More observational and experimental studies are needed to draw firm conclusions about the impact of food literacy on dietary quality, upon which strategies and campaigns to improve food literacy at the population level may be effectively developed and established.

Fig. 1 .
Fig. 1.Flow diagram of participant selection in the present analysis.

Table 2
SPFL total score and HEI-2015 for overall diet according to categories of selected characteristics.

Table 3
Associations of SPFL total and domain scores with HEI-2015 (n =