The participants in this study were adolescents aged 12–18 yr living in the Seoul and Gyeonggi metropolitan areas of the Republic of Korea. The survey was conducted using an online self-administered questionnaire (accessed via Google QR code), which was available for 3 mon, from October 14, 2022, to December 22, 2022. The survey was distributed to middle and high schools recruited through convenience sampling. Participants voluntarily responded to the questionnaire. Of the total 396 questionnaire responses, 4 were excluded because they were missing answers. A total of 392 responses (99% response rate) were used for analysis. The purpose, significance, and intended use of the research were explained to both the participants and their legal guardians prior to the survey, and they provided consent to participate in the study through online signatures. This study was conducted after obtaining approval from the Sangmyung University Institutional Review Board (IRB No. IRB-SMU-C-2023-1-0015).

The demographic characteristics of interest included date of birth, sex, height, weight, grade level, type of residence, and household composition. Body mass index (BMI; kg/m²) was calculated using height and weight measurements [21].

The food-related media use measurement tool was constructed based on previous studies [2223]. The survey included questions about participants' experiences watching food-related content (such as Mukbang or Cookbang), the primary social media platforms they used for food-related content, the frequency with which they watched such content, their preferred types of content, their perception of the influence of food content on their eating habits, their reasons for watching, and their personal experiences with uploading food-related content.

Questions on food consumption behavior investigated the participants’ experiences of late-night eating, food and snack acquisition (specifically, the locations where participants typically purchased their meals and snacks), and level of awareness of nutrition information presented on food products.

The items regarding food-related lifestyle were constructed based on previous research [24], comprising a total of 25 questions categorized into 5 items each related to health, safety, taste, trend, and economic aspects. Participants were asked to rate their agreement with statements on a 5-point Likert scale ranging from “Strongly Disagree” (1 point) to “Strongly Agree” (5 points).

For food literacy, the study used the Food Information Comprehension Scale, a tool for measuring food literacy at each stage of life [142526]. Specifically, it used the scale for children and adolescents, which consists of 19 items categorized into production (3 items), distribution (3 items), selection (7 items), cooking (3 items), and intake (3 items). Participants responded to each item using a 5-point Likert scale, ranging from “Strongly Disagree” (1 point) to “Strongly Agree” (5 points). The weights for each domain and item were calculated. The domain weights were as follows: production (0.175), distribution (0.25), selection (0.20), cooking (0.175), and intake (0.20). The item weights were as follows: production item 1 (9.09), production item 2 (6.82), production item 3 (9.09), distribution item 4 (7.99), distribution item 5 (8.68), distribution item 6 (8.33), selection item 7 (3.66), selection item 8 (3.72), selection item 9 (3.34), selection item 10 (3.61), selection item 11 (3.66), selection item 12 (3.56), selection item 13 (3.45), cooking item 14 (7.80), cooking item 15 (8.78), cooking item 16 (8.41), intake item 17 (7.91), intake item 18 (7.29), intake item 19 (9.80). Sums of weights of 62 or higher were evaluated as “High,” those between 51 and 61 were evaluated as “Medium-High,” those between 41 and 50 were evaluated as “Medium-Low,” and those below 41 were evaluated as “Low” [142526].

For NQ, the study used the NQ-A 2021 scale, a nutritional assessment tool that comprehensively evaluates the quality of meals and eating behaviors in adolescents [18]. It consists of 20 items categorized into balance (8 items), moderation (9 items), and practice (3 items). Participants responded to each item using a 5-point Likert scale, ranging from “Almost never eat” (1 point) to “More than once a day” (5 points). The weights for each domain and item were calculated. The domain weights were as follows: balance (0.15), moderation (0.30), and practice (0.55). The item weights were as follows: balance item 1 (0.017), balance item 2 (0.027), balance item 3 (0.014), balance item 4 (0.025), balance item 5 (0.013), balance item 6 (0.025), balance item 7 (0.014), balance item 8 (0.015), moderation item 1 (0.038), moderation item 2 (0.041), moderation item 3 (0.035), moderation item 4 (0.028), moderation item 5 (0.019), moderation item 6 (0.036), moderation item 7 (0.041), moderation item 8 (0.035), moderation item 9 (0.027), practice item 1 (0.122), practice item 2 (0.238), and practice item 3 (0.190). Sums of weights of 58.6 or higher were evaluated as “High,” those between 44.0 and 58.5 were evaluated as “Medium,” and those below 44.0 were evaluated as “Low” [18].

Statistical analysis was conducted using SPSS 29.0 (IBM Co., Armonk, NY, USA), and statistical significance was defined as P < 0.05. To understand the demographic characteristics of the participants, frequency analysis and descriptive statistics were employed. Principal component analysis was used to examine the segmentation of food-related lifestyle factors. Varimax orthogonal rotation was employed as the factor rotation method in the factor analysis. Items with communalities below 0.4 (Item 1, Item 14), factor loadings below 0.5 (no items), inter-factor correlations indicating non-independence (Item 8, Item 13, Item 16), and only one factor (Item 15) were excluded from the analysis [2728]. Ultimately, the 19 items that met the criteria of communalities above 0.4, factor loadings above 0.5, and eigenvalues above 1 were selected for inclusion in the analysis. After assessing internal consistency using Cronbach’s α, each factor was subjected to K-means cluster analysis for clustering, and the between-cluster differences and significance levels were calculated using analysis of variance (ANOVA). To analyze the differences in media use and food consumption behavior characteristics according to the derived clusters, a chi-square test was conducted. To analyze the differences among clusters in terms of food literacy and NQ, ANOVA analysis was conducted, followed by Tukey’s post-hoc test.

Table 1 presents the demographic characteristics of the 392 participants. Among them, 57.1% were male, and 42.9% were female. The mean age of the participants was 15.0 ± 1.36 yrs; 52.0% were middle school students, and 48.0% were high school students. The average BMI was 21.31 ± 3.78 kg/m², which was within the normal weight range. Regarding current living arrangements, 99.1% of the participants reported living with their families.

The food-related lifestyle variable factor analysis (Table 2) extracted four factors. The Kaiser-Meyer-Olkin (KMO) measure was found to be 0.890, indicating high explanatory power. Bartlett's test of sphericity yielded a significant result of 3,036.361 (P < 0.001), supporting the suitability of factor analysis [29]. The total variance explained by the factors was 60.9%. The four extracted factors were named based on previous research [24]. Factor 1 consisted of 5 items, including “I try not to eat instant foods” and “I sort and then eat foods for health” and was named the “Health-seeking.” It had an eigenvalue of 3.708 and a Cronbach’s α of 0.859 and explained 19.5% of the variance. Factor 2 consisted of 6 items, including “I tend to accept food trends easily” and “I know delicious and trendy food” and was named the “Popularity-seeking.” It had an eigenvalue of 3.329 and a Cronbach's α of 0.832 and explained 17.5% of the variance. Factor 3 consisted of 5 items, including “I always check the price of foods when grocery shopping” and “I buy groceries that are discounted first” and was named the “Economic benefit-seeking.” It had an eigenvalue of 2.364 and a Cronbach’s α of 0.756 and explained 12.4% of the variance. Factor 4 consisted of 3 items, including “I buy products of well-known brands that people buy a lot” and “I prefer products with exact uses” and was named the “Safety-seeking.” It had an eigenvalue of 2.173 and a Cronbach's α of 0.717 and explained 11.4% of the variance.

Table 3 presents the results of the K-means cluster analysis on the four food-related lifestyle factor scores of the participants. Adolescence is a crucial developmental period characterized by changes in dietary habits, attitudes, and voluntary food choices and consumption, marking the transition from childhood to adulthood in terms of dietary behaviors. Moreover, the lifestyles of adolescents often exhibit variations in the degree of interest in food, rather than distinct characteristics as seen in adults [30]. Therefore, drawing on previous research [3132], this study classified three clusters based on the distribution of factor scores, reflecting different levels of interest in food. Cluster 1 showed the highest scores in all food-related lifestyle factors. Therefore, it was named the “High-interest in food group.” Cluster 2 demonstrated scores similar to the mean across all food-related lifestyle factors. As a result, this cluster was named the “Moderate-interest in food group.” Cluster 3 had lower scores in all food-related lifestyle factors than the other clusters; therefore, it was named the “Low-interest in food group.”

Table 4 presents the results of the cross-analysis in this study on cluster-based differences in media use. Concerning food-related content on social networking services (SNS), 91.1% of participants reported no uploading experience, while 76.5% reported viewing food-related content. A separate analysis was conducted on the responses from the 300 adolescents who had experience viewing food-related content, focusing on their most-used SNS platforms, the frequency of their viewing food-related content, the impact of SNS viewing on their eating habits, and their reasons for viewing. The most-used SNS platforms were Instagram (54.7%) and YouTube (30.7%). For frequency of viewing food-related content, 56.7% viewed “1–2 times a week,” followed by 17.7% “3–4 times a week,” and 14.6% “More than 5–6 times a week.” Among adolescents who viewed food-related content 1–2 times a week or more, preferences for content types were as follows: Mukbang 51.7%, ASMR (autonomous sensory meridian response—a relaxing sensation some people experience from certain sounds or visuals) 19.5%, restaurant introductions and reviews 16.5%, and Cookbang 12.3%. Regarding the impact of food-related content on their eating habits, 45.7% mentioned “Eating a lot,” and 31.7% mentioned “Following food menus, trying unique dishes or cooking in unconventional ways.” Cross-analysis of these patterns across clusters yielded statistically significant results; in Cluster 3, a significant percentage of participants (56.1%) reported experiences of “Eating a lot,” and a notable percentage (12.3%) reported experiences of content “Stimulated eating,” while a relatively low percentage (10.5%) reported experiences of “Following food menus, trying unique dishes, or cooking in unconventional ways” (P < 0.01). In contrast, Cluster 1 exhibited the highest percentage (40.2%) of experiences related to “Following food menus, trying unique dishes, or cooking in unconventional ways,” compared to Cluster 3 (10.5%). Additionally, Cluster 1 showed a lower percentage (40.2%) of experiences related to “Eating a lot,” compared to Cluster 3 (56.1%), highlighting significant variations among the clusters (P < 0.01). Among the reasons for watching food-related content, the most prevalent were “To experience vicarious satisfaction by watching others eat food I want to eat” (34.7%), “To satisfy my curiosity about trendy foods, popular restaurants, and new products” (17.7%), and “To learn cooking techniques and adopt healthy eating habits, and to experience psychological satisfaction by listening to food-related sounds” (17.0%). This indicates that adolescents primarily watch food-related content for vicarious fulfillment and information seeking.

Table 5 presents the results of the cross-tabulation analysis conducted to examine between-cluster differences in food consumption behavior characteristics. More than half of the subjects (54.8%) responded to the item on late-night eating with “Hardly ever do.” When purchasing food for meals, 46.9% preferred physical retail stores for primary grocery shopping, while 53.1% preferred online retail stores. In contrast, for snack purchases, 81.7% preferred physical retail stores, while 18.3% chose online retail stores. Most participants purchased snacks from physical retail stores, with Clusters 2 (84.1%) and 3 (85.9%) showing significantly higher usage rates of physical retail stores compared to Cluster 1 (73.5%) (P < 0.05). Additionally, 57.1% of participants indicated that nutritional labeling influenced their food choices, with statistically significant differences between clusters (P < 0.001). A total of 78.6% of participants in Cluster 1 and 54.1% in Cluster 2 reported that nutrition labeling influenced their food choices, while only 37.5% of participants in Cluster 3 reported that nutrition labeling influenced their food choices. Higher levels of interest in food were associated with a higher likelihood of considering nutritional labeling when making food choices (P < 0.001). Among the 224 adolescents who responded that nutritional labeling influenced their food choices, the most impactful nutritional information components were energy (41.5%), carbohydrates (18.8%), sugars and protein (each at 12.5%), fat (9.4%), and sodium (5.3%).

Table 6 presents the results from the ANOVA examining the differences in food literacy and NQ scores among the food-related lifestyle clusters. Based on the calculation of weights for each food literacy domain and item, total scores of 62 or higher were classified as “High,” 51 to 61 as “Medium-High,” 41 to 50 as “Medium-Low”, and below 41 as “Low” [142526]. Cluster 1 scored 72.59, which falls into the “High” category. Cluster 2 scored 54.15, a “Medium-High” rating. Cluster 3 scored 38.07, falling into the “Low” category. Consistent with these results, based on the weights assigned for each NQ domain and item, total scores of 58.6 or higher were evaluated as “High,” those from 44.0 to 58.5 were evaluated as “Medium,” and those below 44.0 were evaluated as “Low” [18]. Cluster 1 scored 59.52, which falls into the “High” category. Cluster 2 scored 47.59, a “Medium” rating. Finally, Cluster 3 scored 39.89, falling into the “Low” category. The post-hoc analysis results for the overall scores and each domain indicated a significant difference among the three groups in scores for food literacy and NQ (P < 0.001), but not for “Moderation” (P > 0.05).