In-depth Assessment of Snacking Behavior in Unmarried Adolescent Girls Living in Urban Centers of Java, Indonesia.

Adolescence is a critical period characterized by physical, social and developmental changes that impact on health and eating behavior. Indonesia is experiencing dramatic economic and infrastructural changes, causing greater access to the global food industry and media. This transition is influencing food intake trends, leading to new nutritional challenges in adolescent girls. Qualitative research was conducted between November 2016 and January 2017 in five urban sites in Java, Indonesia to examine individual, social, environmental and macrosystem factors affecting snacking behaviors in unmarried adolescent girls 16-19 years of age. Methods entailed 30 freelisting exercises, nine key informant interviews, and 16 in-depth interviews. Freelisting results identified over 200 snack foods, with the most salient processed convenience foods such as chips and cookies. Respondents typically snacked multiple times daily. Widespread availability of affordable and "tasty" snacks makes snack foods appealing meal substitutes. Snacks provide a distraction to boredom and loneliness and an enhancement to social gatherings. Girls exhibited limited understanding or concern about potential negative effects of snacking. Parents facilitate acquisition of nutrient-poor snacks, while friends exert pressure for routine consumption of snack foods. Social media infiltrated with promotions of eateries and snack foods is likely contributing to the preponderance of snack food consumption. Routine consumption of snack foods high in sugar, salt and fat and skipping meals will likely have long-term consequences on the nutritional status and health of Indonesian adolescent girls. This in-depth study provided context-specific information for a social media campaign designed to motivate girls to follow healthier snacking practices.


| INTRODUCTION
Adolescence is a transitional period in the life cycle characterized by physical, social, and developmental changes that influence health and eating behaviour. Starting at the onset of puberty and continuing to young adulthood, the World Health Organization (WHO) defines adolescence as between 10 and 19 years of age, with the period commonly divided into early (10-14 years) and late (15-19 years) adolescence (Black et al., 2013; G. C. Patton et al., 2018). During this period, rapid growth and dramatic physical changes create increased demand for nutrients and energy as simultaneously, adolescents undergo many psychosocial changes, such as seeking independence and self-identity and growing concerns about body image, all of which can impact on food selection and nutritional intake (Das et al., 2017;Story, Neumark-Sztainer, & French, 2002).
Further, the nutritional status of females before and during pregnancy is critical for the health and survival of the mother and her baby (Black et al., 2013;Han, Mulla, Beyene, Liao, & McDonald, 2011). Research has consistently shown that adolescents frequently have less than optimal eating habits, increasing the risk for nutritional problems, including undernutrition, overweight and obesity, iron deficiency anaemia, and micronutrient deficiencies (Akseer, Al-Gashm, Mehta, Mokdad, & Bhutta, 2017). The combination of these factors makes adolescence a nutritionally vulnerable period (WHO, 2005).
Understanding nutritional problems in adolescents is important in lower income and middle-income countries where there has been a rise in noncommunicable diseases while malnutrition persists (Popkin & Gordon-Larsen, 2004;Rivera, Pedraza, Martorell, & Gil, 2014).
Escalating incomes and urbanization, as well as improved communication and the expansion of the global processed food industry, have increased access to packaged energy-dense, nutrient-poor convenience foods and fast food outlets (Popkin, 2017). Economic and social changes have led to modifications in diet and physical activity and are linked to a nutritional transition involving a rapid rise in the incidence of overweight and obesity (Popkin, 1999;Popkin & Gordon-Larsen, 2004). In many lower income and middle-income countries, marriage and first birth still occur at a young age when girls are maturing, leading to increased nutritional demands (Gibbs, Wendt, Peters, & Hogue, 2012;WHO, 2014). Moreover, adolescent behaviours involving food choices, eating frequency, and the social context of food consumption are found to be predictive of adult eating practices (Craigie, Lake, Kelly, Adamson, & Mathers, 2011;Larson, Neumark-Sztainer, Hannan, & Story, 2007;Merten, Williams, & Shriver, 2009;Pedersen, Holstein, Flachs, & Rasmussen, 2013).
Indonesia, the fourth most populated country in the world, is experiencing rapid socio-economic changes affecting food accessibility and intake, which in turn impacts on nutritional status, including that of adolescent girls (United Nations, 2017). Overweight and obesity has increased over the past two decades across all age groups (Rachmi, Li, & Alison Baur, 2017). Prevalence is highest in female adolescents and adults, with a third of women over 18 years of age overweight or obese, whereas close to a 10th of girls in the 16-18 age group are reported to be overweight or obese (National Institute of Health Research and Development, 2013). At the same time, national survey data show that approximately half of nonpregnant girls aged 15-19 are undernourished (as indicated by mid-upper arm circumference measurements <23.5 cm) and a fourth of females between 15 and 24 years suffer from anaemia (Health Research and Development Agency, 2013). These studies highlight a complex double burden of undernutrition and overweight present in Indonesian adolescent girls (Roemling & Qaim, 2013).
A limited number of sociobehavioural studies have been conducted to understand the causes of overweight in Indonesian adolescents. One study demonstrated that perceptions that fast food is modern and practical, limited physical activities, and high intake of soft drinks and processed foods contributed to obesity, with adolescents from higher income families three times more likely to be obese (Collins, Pakiz, & Rock, 2008). Another study showed that decreased physical activity and high consumption of modern foods were linked to higher body mass index (Roemling & Qaim, 2012). Although overweight and obesity in Indonesia has been associated with higher socio-economic status and urban residency, recent studies showing an increase in prevalence of households with a double burden of malnutrition suggest that a shift may be occurring with overweight affecting a broader socio-economic spectrum (Rachmi, Li, & Alison Baur, 2018). Snacking is reported to be pervasive in Indonesia (Global Alliance for Improved Nutrition, 2014), as well as among adolescent girls (Kadence International, 2014). Consumption of fried foods including oily and fried snacks more than four times per week has been associated with overweight among children and adolescents (Prihatini & Jahari, 2007).
A conceptual model designed to understand and explain adolescent food choices and consumption behaviours identifies broad levels of influences including individual (psychosocial, biological, knowledge,

Key messages
• Adolescent girls snack multiple times daily on foods high in sugar, salt, and fat, leading to meal skipping.
• The motivation to snack is primarily driven by the exotic flavours and alluring tastes offered by cheap, unhealthy snacks saturating Javanese cities.
• Snacking serves as an antidote to boredom and loneliness and is considered central to socializing.
• Adolescent girls have limited understanding of snack food contents or concerns about the negative effects of snacking.
• Unless trends of excessive snacking and meal skipping are addressed, there will likely be a further rise in overweight and obesity among female adolescents and adults. and lifestyle), environmental (family, friends, schools, worksites, fast food outlets, and convenience stores), and macrosystems (socio-political changes, food availability, food production and distribution, and media and advertising) (Story et al., 2002). Although most studies focus on single-factor determinants of food intake and nutritional status at the household and individual levels, this ecological model facilitates an exploration of multiple levels of influence, including often neglected broader community and societal determinants, and enhances an understanding of their relationships (Madjdian, Azupogo, Osendarp, Bras, & Brouwer, 2018;Story et al., 2002). This conceptual framework can be useful when examining the multifaceted, dynamic factors affecting food consumption, and the way they interact in influencing adolescent eating behaviours.
Given the spectrum of nutritional challenges faced by Indonesian adolescent girls, social science research is needed to understand better factors affecting eating behaviours to inform intervention strategies. Using the model, we carried out qualitative research to examine the ways in which individual motivations, social and environmental factors, and media and advertising influence eating practices, with a focus on snacking behaviours of girls aged 16-19 in urban Java, Indonesia. The findings provide context-specific information for the future development of policies and programmes designed to promote healthy eating behaviour in adolescent girls.

| Study setting and population
The research was carried out in five urban centres in Java, namely, Central Jakarta, East Jakarta, Surabaya, Yogyakarta, and Malang, over a 5-week period between November 2016 and January 2017. Selection of research sites was designed to focus on hubs across the island that ranged in geographic locations, population size, economics, and cultural backgrounds. The study concentrated on unmarried adolescents aged 16-19 years, an age group in Indonesia shown to be susceptible to overweight and obesity, representing three categories of girls who typically have contrasting schedules and lifestyles including high school students, college students, and girls working full-time.

| Study design, sampling, and methods of measurement
We employed complementary qualitative data collection methods using structured, semistructured, and open-ended techniques to assess snacking behaviours as described below.
a. Freelisting is a structured exercise used to generate a preliminary inventory of the cultural domain of snack foods and associated terminology (Pelto & Pelto, 1978). We carried out two freelisting exercises, asking respondents to list (a) commonly consumed snack foods and (b) motivations for consumption of snack foods. Exercises were administered to girls living in the five research sites. Using purposive sampling to select unmarried female adolescents aged 16-19 years representing the three study categories of girls, high school and college girls were identified in education institutions and malls known as "hang out" places, and employed girls were approached in shopping and trade centres where they frequently work. The initial aim was to carry out freelisting with 30 adolescent respondents.
b. In-depth interviews with adolescent girls were designed to explore individual motivations and environmental influences guiding eating behaviours of adolescent girls, especially snacking, including how they choose snacks, where and when they obtain and eat snacks, and the social context in which snacks are consumed. Eligible high school and college girls were identified on school campuses, and employed girls were approached for interviews in shopping centres where they worked. Inclusion criteria established prior to the study were as follows: high school students aged 16-17 years enrolled in a private, public, or vocational school and living with their family; college students 18-19 years of age enrolled in a college or university full-time and living either with their family or in a dorm; and workers 18-19 years, employed full-time, and living with their family or on their own. Girls who had participated in the freelisting exercises were not eligible to partake in the in-depth interviews.
A recruitment questionnaire that had been previously tested for validity and reliability and employed by the data collectors was used to identify girls living in families from a range of socioeconomic backgrounds. Potential respondents were asked a series of questions related to the household economic status such as the materials used for construction and size of their home, family assets, type of transport family members use, and source of drinking water. Each response had a precoded score, with scores tallied once the questionnaire was completed. Selection criteria were used to ensure that girls represented a range of socio-economic status with the overall target to include eight high school students, four college students, and four working adolescent girls in the study.
c. Key informant interviews were conducted to examine motivations for and patterns of eating and snacking behaviours and recommendations for behavioural change strategies from a diverse spectrum of informal and formal experts on adolescence. Respondents were selected purposively based on their background and expertise as it relates to the study topic and included mothers, food sellers, journalists, and health experts, with most key informants located in Jakarta.
d. Informal observations of food venues were carried out in the study sites. Researchers were asked to observe, record, and describe the eating environment respondents were exposed to routinely.

| Data collection procedures
Study instruments were administered by seven experienced qualitative researchers assigned to different city sites. Prior to data collection, a 3-day training was held to introduce the study protocol, methods, and ethical procedures and to test and modify the instruments. All interviews were administered in bahasa Indonesia, the national language of Indonesia. Key informant and in-depth interviews were audio recorded; interviewers also took handwritten notes of information that could provide additional insights into the data.
Researchers were in regular communication throughout the data collection period, either through a cloud account or through group calls to share preliminary results. An iterative process involving the review of initial findings and additional questioning continued until data saturation was reached. A debriefing meeting involving study investigators was convened at the completion of data collection.

| Data analysis
Freelisting exercises were analysed on Anthropac 4.983 software.
Salience, which refers to the perceived importance, was derived using a salience index (Smith's S) defined as follows: where L is the length of each list, Rj is the rank of item J in the list, and N is the number of lists in the sample (Borgatti, 1999). Items with a higher salience score were considered to have greater significance in the local framework of snack foods (Bernard, 1988;Romney & d'Andrade, 1964;Weller, 1984). The audio-recorded interviews were translated into English and transcribed in a Microsoft Word document; after reviewing the transcripts, a coding system was developed for the key informant and in-depth interviews. Coding categories were derived from the initial research themes, as well as from key concepts that emerged during data collection. Coding of the interview transcripts was done in ATLAS.ti, a text-organizing software. Content analysis was used to identify trends of concepts in and across individual codes.
The combination of data and methodological triangulation allowed us to analyse data across the different research methods and across and between respondents (Bernard, 2017;M. Q. Patton, 2015).

| Ethics
The study protocol was reviewed and ethical approval granted by the institutional review board at the Atmajaya University, Jakarta, Indonesia. Signed informed consent was obtained from all the freelisting, key informant, and in-depth interview respondents.

| Background information
Freelisting exercises were administered to 30 girls across the five research sites. Nine key informants were interviewed, including two mothers and two food vendors plus one nutritionist, magazine editor, psychologist, beauty blogger, and food blogger. In-depth interviews were carried out with eight high school, four college, and four working respondents located in the five sites. Half of the college and working respondents lived apart from their parents with college respondents living in school dorms inhabited only by college students and working respondents residing in boarding houses with kitchen facilities. Employment of working respondents included administrative assistant, cleaner, cashier, and salesperson. Table 1 presents background information of the in-depth interview respondents.

| Freelisting
Freelisting generated 213 different snack foods. Typical of a cultural domain, the procedures elicited a core set of more significant snack items and a wide range of less important foods, with the 25 most salient items listed in Table 2. Core snacks were predominantly modern foods such as chips, cookies, puffed snacks, French fries, and ice cream, whereas traditional foods such as dumplings and pentol (skewered meatballs) were considered of less importance. The most salient motivation for consumption of snack foods was taste, followed by easy to get, filling, cheap, and meal substitute, with the designation "healthy" mentioned by only one respondent.

Food meanings
Adolescents consistently described makanan (meals) as heavy, filling foods that stave off hunger over extended periods. These foods are eaten at specific times of the day, including breakfast, lunch, and dinner, and often composed of big portions. The central meal ingredient is rice, the staple and ultimate filling food, with many respondents confirming that without rice, it is not a meal. Other meal foods characterized as having filling properties included noodle-based dishes, meatballs, meat or chicken, soups, and "Western" foods such as hamburgers, hot dogs, fried chicken, and pizza. When these Western foods are sold in restoran cepat saji (fast food chains), they are labelled as junk foods or makanan cepat saji, which were described as containing preservatives, processed and stored over a long time, prepared quickly, unhealthy, and delicious. The same foods sold by street vendors or in warungs (family-owned restaurants or food shops) are not considered "junk" because they take time to prepare, highlighting that fast processing signifies junk food. Fast foods were defined as packaged meal items, such as instant noodles, and high in preservatives.
Cemilan (snack foods) are described as light and nonfilling and consumed in small portions. Jajan, a verb that respondents signified involves buying small foods or snacks, was labelled a common activity among adolescent girls. Snacks are eaten to replace or in-between meals, with a few respondents calling them makanan penutup (literally dessert foods) consumed directly after meals.
Many girl respondents differentiated between modern and traditional snacks, with modern snacks described as packaged, containing preservatives, and branded by a company; traditional snacks are typically not packaged, made with local ingredients, and sold to eat immediately.

Lifestyles and eating behaviours
High school students and employed girls had busy, structured schedules, often leaving home at 6 a.m. and returning after 5 p.m., with high school students studying in the evening. College girls have less regimented timetables, with long breaks between classes. Leisure time at home is spent watching television, talking to family members,

| Environmental influences
Family Parents were consistently named the most influential people in the girl respondents' lives, guiding major decisions related to school and career, as well as choices concerning extracurricular activities and relationships. High school and employed girls indicated that parents place restrictions on social activities, whereas parents of college students were reported to give more freedom. Girls share personal issues with their mothers, who they frequently called friends. Even in the few households where parents did not regularly purchase snacks, girls were permitted to buy and store packaged snacks.
All parents of high school and college students provided a relatively large stipend, ranging from 10,000 to 42,857 Rupiah per day (0.74-3.17 USD), with much of the money intended for food purchases while the girls are outside the household. Working respondents used earned money to purchase snacks.

Friends
Our respondents typically had a close set of friends they hang out with during school or work breaks or in eateries, malls, and movie theatres. Girls indicated that friends influence fashion, clothing, and hairstyles. Key informants emphasized that adolescents strive to be accepted by their peers, making alterations in their appearance or behaviours to adhere to social trends.
Eating is a common activity done with friends, often serving as an impetus for frequenting fast food restaurants, cafes, warungs, and trendy specialty food shops. Respondents alluded to the social significance of snacking with friends. One high schooler (R3) living in Our respondents commonly discovered new snack foods through talking to friends or observing what friends eat.

Physical
The research areas offered a plethora of modern foods in a range of settings and venues, entailing convenience stores, minimarkets, and supermarkets; shopping malls that typically house a wide range of restaurants including fast food chains, cafes, small food shops, and food courts; and street carts, food stands, and warungs. Local markets selling traditional delicacies were less prevalent. Canteens are located in high schools and colleges, with multiple stands offering both meal and snack foods such as packaged cookies or candy, fried foods or bread, or traditional foods such as dumplings during school breaks and lunchtime. Food vendors also position themselves outside schools, selling items popular among teens such as fried foods and fruit juices, which students purchase during breaks or after school.

| Macrosystems
The research cities had extensive Internet networks, making Internet both widely available and affordable. Our respondents who all had smartphones consistently cited the Internet as their primary source of information, with social media sites the most frequently visited.
Social media was reported to be infiltrated with food advertisements and culinary information, including well-liked eateries, restaurant promotions, cooking tutorials, and popular snack foods. Information on new and trendy eateries and snacks was also obtained through friends' social media accounts. Many respondents reported watching television routinely, indicating that television commercials frequently promote snack foods. Television was cited as the primary source of information related to food hygiene and safety.
Key informants indicated that social media defines the way adolescents obtain news, providing continually changing and exciting visually dominated information, which sparks their incessant desire to browse popular apps to follow trends and connect with friends. They stressed that the Internet is altering the way adolescents interact and express themselves, facilitating the opportunity to acquire information and communicate rapidly at any time through predominantly visual mediums and making digital connectivity critical to daily activities, including eating.

| DISCUSSION
This qualitative study explored individual, social, physical, and macrosystem factors influencing snacking patterns and behaviours of urban dwelling adolescent girls in Java. The widespread availability of affordable snacks and inclination to consume tasty foods make snacks appealing substitutes for meals. Snacks provide a distraction to boredom, solace when girls are alone, and an enhancement to social gatherings. Girls exhibited limited understanding or concern about the contents or potential negative effects of snacking.  (Pliner & Bell, 2009). Limited understanding of nutrition together with lack of concern of the potential health ramifications appears to sanction routine snacking in adolescent girls. Research in other contexts has shown that adolescents prioritize immediate gratification provided by taste and popularity over health and nutrition (Cross, Babicz, & Cushman, 1994).
Traditionally, Indonesian mothers are responsible for overseeing the daily activities and needs of their children, including those related to food. In addition to providing three meals daily, most mothers routinely purchase a variety of packaged snacks high in fat and sugar.
Adolescent girls' partiality for snacks combined with availability of snack foods at home leads to skipping meals, which are considered heavy and monotonous. Meal skipping was more frequent in respondents living away from home and parental care. The perception that skipping meals is an effective way to lose or maintain weight and that snacks are "light" and low in calories also affects eating patterns, particularly the tendency to replace a rice/noodle dinner-an optional meal not necessarily consumed with family members-with snacks.
The evening is also the time when girls are studying or browsing the Internet, making packaged snack foods a convenient food option.
Other studies have shown that snacking can lead to meal skipping in adolescents (Savige, Macfarlane, Ball, Worsley, & Crawford, 2007). In other contexts, regular consumption of family meals has been linked to better dietary diversity, nutrient and micronutrient intake and food choices, and may have a positive influence on the quality of diets and meal patterns in young adulthood (Cusatis & Shannon, 1996;Lachat et al., 2012;Larson et al., 2007;Neumark-Sztainer, Hannan, Story, Croll, & Perry, 2003;Sjoberg, Hallberg, Hoglund, & Hulthen, 2003 influencing food choices and intake, we omitted to examine certain determinants of eating behaviours, such as household socioeconomics and mother's education, shown to be associated with nutritional status of adolescent girls in Indonesia and elsewhere (Kunto & Bras, 2018;Madjdian et al., 2018). We recognize that a wider range of potential determinants of eating patterns exist, as presented in a recently developed conceptual framework on adolescent malnutrition based on a systematic review of quantitative studies, which can be analysed through quantitative assessments (Madjdian et al., 2018).
Our findings underline the urgent need to develop contextually relevant, targeted behavioural change strategies to modify the potentially harmful eating and activity patterns of adolescent girls identified in this study and to curb the trajectory of overweight in urban Indonesia.

| CONCLUSION
In Indonesia, the coexistence of a high prevalence of overweight and undernutrition signals a double burden of malnutrition among adolescents living in urban centres. The pattern of poor eating behaviours and lack of physical activity illuminated in this study, combined with an escalating reliance on social media, underline the urgent need to develop policies and programmes to combat extreme forms of malnutrition simultaneously through targeted approaches. From a sociobehavioural standpoint, the challenges to address the nutritional problems identified are heightened due to the susceptibility of adolescents to environmental influences and unhealthy lifestyles and a tendency to disregard long-term consequences of potentially harmful behaviours typical of this age group (WHO, 2005). Unless the trend of excessive snacking and meal skipping is addressed, eating behaviours in urban adolescent girls are likely to deteriorate, leading to further rise in overweight and associated chronic diseases, which over the long term, can have devastating effects on the health and economy in Indonesia. The Indonesian Ministry of Health (2015)

ACKNOWLEDGMENTS
We thank the study respondents for participating in the study; PT Kadence International for support in study design, data collection, analysis, and reporting; and Frances Aboud for her valuable comments on the manuscript.

CONFLICTS OF INTEREST
The authors declare that they have no conflicts of interest.

CONTRIBUTIONS
MVL, LSB, EKS, SS, RM, DI, and AT designed the study. INY supervised field operations, and AM conducted data collection. LSB, AM, and INY carried out data analysis and interpretation. This paper was written by LSB with substantial input from all other authors. All authors have reviewed and approved the submitted manuscript.