*Article* **The Prevalence of Overweight Status among Early Adolescents from Private Schools in Indonesia: Sex-Specific Patterns Determined by School Urbanization Level**

**Eveline Sarintohe 1,2,\*, Junilla K. Larsen 1, William J. Burk <sup>1</sup> and Jacqueline M. Vink <sup>1</sup>**


**Abstract:** (1) Background: Few studies have investigated (demographic) correlates of (prevalent) overweight rates among early adolescents, especially from higher socioeconomic positions (SEP) in developing countries, such as Indonesia. The current study aims to fill this gap. (2) Methods: Participants included 411 adolescents from five private schools in Indonesia. Adolescents' weight and height were measured, and adolescents completed questionnaires on demographic factors (i.e., sex, school area, ethnicity, pocket money) and previous year dieting. (3) Results: Results showed that more than one-third of the sample was overweight, with higher rates among adolescent males (47%) than females (24%). Moreover, adolescents attending schools in urban areas (compared with suburban areas), and those reporting past dieting (compared with those reporting no dieting) had higher overweight rates. Ethnicity and the amount of pocket money were not related to overweight status. Finally, a clear sex-specific interaction was found involving school area, showing that males in urban areas had a significantly higher risk to be overweight, whereas this did not apply to females. (4) Conclusions: males from urban area private schools in Indonesia may be an important target group for future preventive overweight interventions.

**Keywords:** obesity; overweight; developing countries; Indonesia; adolescents; sex differences; demographic; high SEP

#### **1. Introduction**

The prevalence of obese and overweight individuals has continued to increase over the past years, particularly in developing countries, such as Indonesia [1–3]. In contrast to Western countries, obesity is positively related to socioeconomic position (SEP) in many developing countries [4–6], meaning that being overweight is more prevalent among adults and adolescents with higher SEP. So far, studies have suggested that in developing nations, people with a higher SEP, compared with lower SEP, have easier access to junk food or calorie-dense foods, which may explain the higher overweight rates, particularly in these groups [5,6].

Adolescence is a particularly vulnerable period for the development of overweight, not only in Western countries but also among developing countries such as Indonesia [7,8]. Moreover, overweight prevalence seems to show sex-specific differences in Indonesia. Among adolescents, the prevalence of overweight was higher in females than in males [4,9]. The same has been found for adult populations in Indonesia (i.e., higher prevalence rates among women compared with men) [2,4,10]. However, among children, the prevalence of overweight was higher in boys compared with girls [4]. These (review) results suggest some shifting sex-specific patterns during early adolescence regarding the prevalence of overweight status. The current study has a specific focus on early adolescence, a critical

**Citation:** Sarintohe, E.; Larsen, J.K.; Burk, W.J.; Vink, J.M. The Prevalence of Overweight Status among Early Adolescents from Private Schools in Indonesia: Sex-Specific Patterns Determined by School Urbanization Level. *Nutrients* **2022**, *14*, 1001. https://doi.org/10.3390/ nu14051001

Academic Editor: Roberto Iacone

Received: 11 February 2022 Accepted: 25 February 2022 Published: 27 February 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

stage where (gender-specific) lifestyle choices change, for example, because of the availability of energy-dense (junk) food and increasing peer influences in (changing) school environments [9,11,12], with the highest possible impact among adolescents at private schools [4].

In general, school is an environment in which adolescents spend much of their time with peers, and where junk food is available. As such, school area (i.e., urban versus non-urban areas or less urban) might play an important role in obesogenic behaviors, including junk food consumption [9,13]. In line with this, some studies in Indonesia have shown that overweight patterns differ according to school area [11,14]. Adolescents living in urban (school) areas have higher prevalence rates of overweight status compared with those living in less urban or rural areas [11,14]. This might be explained by their greater access to more types of junk food or fast food compared with people living in less urban or rural areas [9,13–15]. Moreover, previous research has also found that overweight patterns differ according to ethnicity. To date, overweight status is more prevalent among people with Orang Asli Malaysian compared with Chinese Malaysians backgrounds [13]. Further research is needed to examine whether and how overweight status among early adolescents at private schools, with generally higher overweight risk, might differ according to school area and ethnic background in Indonesia.

Furthermore, given that the amount of pocket money may be an indication of the possibility to buy fast food, pocket money may also be related to overweight status among early adolescents from private schools and mostly higher SEP backgrounds. Finally, studies in industrialized countries (in Europe and America) have shown that dieting behavior is associated with greater weight gain over time among adolescents [16]. Due to the increasing impact of Western society on developing countries such as Indonesia, it is important to identify whether dieting behavior is similarly linked to overweight status among early adolescents at relatively high risk for being overweight from private schools in Indonesia.

Moreover, some previous studies among Indonesian populations have also shown sexspecific links regarding demographic factors explaining overweight status. Specifically, two previous Indonesian studies in adult samples showed that females from urban areas were at higher risk to be overweight compared with males [2,5]. In contrast, one recent large-scale study among children and adolescents (10–18 years old) has shown that, specifically, males living in urban areas were more likely to be overweight and obese than females living in urban areas [12]. Given these contrasting findings, it is important to further examine sex-specific links between school area and (over)weight status among early adolescents, particularly among those from private schools with higher overweight prevalence [4]. Another study suggested that overweight status may be more strongly linked to ethnicity among males compared with females [17]. Finally, some studies among Western countries also show sex-specific links between dieting and (over)weight status [14,16]. As such, we will also explore sex-specific demographic or dieting correlates of overweight status among early adolescents from private schools (higher SEP background).

To conclude, recent research on demographic correlates of overweight prevalence rates among early adolescents from private schools in Indonesia is limited. However, society, particularly around private schools (higher SEP backgrounds), might have changed rapidly with regard to eating behavior in Indonesia (more fast-food restaurants, larger influence from Western society [2,9]) making it urgent to explore correlates of overweight in early adolescents at private schools in Indonesia nowadays. As such, the aim of the current study is to examine (sex-specific) correlates of overweight status in a relatively large sample of early adolescents from private schools.

#### **2. Materials and Methods**

#### *2.1. Participants and Procedure*

The participants in this study were part of the baseline measurement (Wave 1) from an ongoing longitudinal study on adolescents' weight-related behavior in Indonesia. Wave 1 took place in October until December 2019. Adolescents were recruited through five private junior high schools in four cities (Jakarta, Surabaya, Bandung, and Manado) in Indonesia. A total of 411 students (47.7% females) participated. All adolescents (Mage = 12.02 years; SD age = 0.45; range = 11.02 to 14.11 years) were in 7th grade or in their first year of junior high school.

A letter describing the longitudinal project was initially sent to officials of school foundations (some private schools are organized by private foundations) or directly to school officials. If the school foundations provided approval, the agreement letter was then sent to the principal of the schools. School officials informed both the parents and students about the goals of the project. Parents were asked to return a signed consent form indicating they agreed to their child's participation. Students were also asked to return a signed consent form indicating whether they agreed to participate in the study. Of the five schools that agreed to participate, three schools obtained consent forms from parents and students. The remaining two schools informed the parents about this project (passive consent) based on the school policy and collected the signed consent forms from students only. The original and amended (passive consent) procedures were approved by the Ethics Committee Social Science of Radboud University, Nijmegen, The Netherlands (ECSS-2019-115).

Researcher informed students that their participation was voluntary, that answers would be processed confidentially and would be stored separately from personal data (with a key file to link the data), and that they could withdraw from the study at any time. Adolescents completed a paper self-report survey at school during one classroom hour (approximately 60 min). In addition, adolescents' weight and height measures were taken by the researcher with the assistance of the school nurse. Weight and height of participants were assessed using school equipment (stadiometer). Students were rewarded with a small gift when they completed the questionnaires.

#### *2.2. Measurements*

#### 2.2.1. Anthropometric Measurements

Adolescents' height was measured to the nearest 0.1 cm with a validated stadiometer (Seca around 217), and their weight was measured to the nearest 0.1 kg with a weighing scale (Seca around 840). Based on the Center for Disease Control and Prevention (CDC) 2000 Body Mass Index for age growth charts for males and females, the cut-off for defining overweight was based on the sex and age in months and BMI (weight (kg)/height (m2)).

#### 2.2.2. Demographic Characteristics

Adolescent's sex was coded, with 0 = female and 1 = male. School area was coded as 0 = suburban (Bandung and Manado) and 1 = urban area (Jakarta and Surabaya). We divided the area based on modernization and levels of Westernization [8]. There were no exclusion criteria involved. All students from urban and suburban areas participated in this study. Moreover, ethnicity was coded as 1 = Javanese, 2 = Sundanese, 3 = Sulawesi, 4 = Tionghoa (Chinese Indonesian), 5 = other ethnic (Papua, Kalimantan, Sumatra, and Bali), and 6 = mixed ethnicity. The percentage of Chinese-Indonesian ethnic students was almost half of the sample (49.8%), so we decided to dichotomize ethnicity as 0 = Indonesian ethnic (Javanese, Sundanese, Sulawesi, etc.) and 1 = Chinese Indonesian. The amount of pocket money was coded as 1 = < IDR 500,000, 2 = IDR 500,000–IDR 1,000,000, 3 = IDR 1,100,000–IDR 1,500,000, 4 = IDR 1,600,000–IDR 2,000,000, 5 = IDR 2,100,000–IDR 2,500,000, and 6 = > IDR 2,500,000. The percentage of students with pocket money less than IDR 500,000 was more than half of the sample (63%), so we decided to dichotomize pocket money as 0 = < IDR 500,000 and 1 = ≥ IDR 500,000.

#### 2.2.3. Dieting Behavior

To measure previous dieting behavior, participants were asked, "In the past year, how often did you diet in an attempt to have the same weight or lose weight?" The response categories for this item were: 1 = never, 2 = 1–2 times, 3 = 3–4 times, 4 = 5–6 times, and 5 = 7 times or more often. Initial inspection of the distribution of this item indicated a substantial group of adolescents who reported no past year dieting (50.68%), so this item was also dichotomized as 0 = no past year diet and 1 = did past year diet.

#### *2.3. Statistical Analyses*

Chi-square analyses were performed to examine univariate demographic and dieting differences between overweight and non-overweight groups. Moreover, a logistic regression analysis was performed to explain overweight status group membership (0 = not overweight; 1 = overweight) from several predictors. The independent variables included in this analysis were student´s sex, school area (suburban vs. urban), ethnicity (Indonesian vs. Chinese Indonesian), pocket money (<500,000 vs. ≥500,000), and previous dieting behavior (never vs. did diet). Moreover, sex-specific interactions (i.e., sex by school area, sex by ethnicity, sex by pocket money, and sex by previous diet) were tested in four separate analyses (one interaction per analyses added to the main effects model). Statistically significant interactions were further probed using the PROCESS module in SPSS [18].

#### **3. Results**

#### *3.1. Descriptive Statistics*

Data from a total of 411 students were examined in this study. The sample was equally divided according to sex (53.3% boys). In total, 59.1% of the adolescents attended a school located in an urban area. The sample was also equally divided according to ethnic background (51.2% Indonesians and 48.8% Chinese Indonesians). Most of the respondents had less than IDR 500,000 per month (63%) and had not dieted (50.6%).

In the total sample, 36.3% of the adolescents were characterized as being overweight. Chi-square independence tests indicated that overweight status was more prevalent in males compared with females (see Table 1). Moreover, adolescents living in urban school areas had a higher overweight prevalence compared with those living in suburban areas. Finally, the adolescents reporting previous dieting were more likely to be overweight compared with those who did not report dieting. Overweight status did not differ according to ethnic background or amount of pocket money.



*3.2. Unique Contributions of Demographics and Dieting in Explaining Overweight Status*

A binary logistic regression was performed to examine the unique contributions of the five predictors in explaining overweight status. Males were 3.28 times more likely to be overweight compared with females (CI 95% (2.08, 5.18)). Adolescents from urban school areas were 1.84 times more likely to be overweight compared with those from suburban school areas (CI 95% (1.07, 3.14)). Adolescents reporting dieting were 3.84 times more likely to be overweight compared with their non-dieting counterparts (CI 95% (2.45, 6.03)). The main effects of ethnicity and pocket money were not statistically significant (see Table 2). All variables together explained 20.3% of the variance in overweight status.

**Table 2.** Logistic regression predicting overweight status by demographic and dieting correlates in the total group.


Note: \* *p* ≤ 0.05, \*\* *p* ≤ 0.01. Sex: 0 = females, 1 = males; school area: 0 = suburban, 1 = urban; ethnicity: 0 = Indonesian, 1 = Chinese Indonesian; pocket money: 0 = <500,000 Rp, 1 = >500,000 Rp; and past year dieting: 0 = no dieting, 1 = dieting. B: Beta; SE: Standard Error; OR: Odds Ratio; CI: Confidence Interval.

#### *3.3. Sex-Specific Interactions*

Four separate sex-specific interaction analyses were performed, in which one interaction was added to the main regression model. Of these four interactions, only the interaction between sex and school area was statistically significant (see Table 3). The explained variance for the total model including the interaction was 21.5% (b = 0.99, SE = 0.47, CI 95% (0.07, 1.89)). We further probed this interaction using Model 1 PROCESS module for SPSS. The results showed that males living in urban areas were more likely to be overweight compared with males living in suburban areas (b = 0.99, SE = 0.48, and CI 95% (0.34, 1.65)), whereas this did not apply to females (b = 0.01, SE = 0.39, and CI 95% (−0.75, 0.78)). The other sex-specific interactions were not statistically significant.

**Table 3.** Logistic regression predicting overweight status by demographic and dieting correlates in the total group including interaction effects with sex.


Note: \* *p* ≤ 0.05, \*\* *p* ≤ 0.01. Two-way interactions were tested separately (one interaction per analyses added to the main model). Total model explained variance were reported per separately tested interaction. Sex: 0 = females, 1 = males; school area: 0 = suburban, 1 = urban; ethnicity: 0 = Indonesian ethnic, 1 = Chinese-Indonesian ethnic; pocket money: 0 = <500,000, 1 = >500,000; and past year dieting: 0 = never did diet, 1 = did diet. Sex-specific interactions (Sex X school area): girls=b= 0.01 (CI95% (−0.75, −0.78)); boys = b = 0.99 (CI95% (0.34–1.65)) using Model 1 PROCESS module for SPSS.

#### **4. Discussion**

The current study aimed to examine the (sex-specific) demographic and dieting factors that potentially explain overweight status among a relatively large group of Indonesian early adolescents attending private schools (i.e., higher SEP background). Children and adolescents from private schools are more likely to be obese [4] and early adolescents' weight is predictive of their weight status in adolescence and adulthood [11]. Finding correlates of overweight in this specific period might give insights for future prevention or intervention

and may have both direct and longer-term health benefits. Our findings showed that the general prevalence rate of overweight in this early adolescent sample at private schools was relatively high (i.e., 36.3%) compared with previous national prevalence rates (i.e., 16%) [11]. The seemingly higher percentage of overweight status among adolescents from higher SEP backgrounds may be environmentally driven [4]. As mentioned, higher SEP private schools, particularly those in urban environments, are often located in areas with more junk food outlets [3,9]. Food outlets usually sell fried products, that are highly preferred, and these kinds of products are highly energy dense. People from higher SEP backgrounds often opt to eat out rather than at home, and food served in restaurants or food outlets usually contains more calories [3,15,19]. In addition, most Indonesian parents from higher SEP are proud when their children look big or fat, reflecting a higher socioeconomic status [3]. Together, these factors may probably explain the relatively high overweight prevalence rate in our study sample.

The relatively high overweight prevalence in our sample makes further insights into (sex-specific) demographic correlates even more interesting, given the increased statistical power to detect effects. We found that males were almost four times more likely to be overweight than females. This result is consistent with most previous studies among Indonesian children [4,9]. However, these findings are in contrast with previous studies among adolescents and adults, where prevalence rates are mostly reported to be higher among females compared with males [2,4,10]. Our findings indicate that early adolescent males are (still) more likely to be overweight compared with early adolescent females, at least among adolescents attending private schools. Future longer-term studies following early adolescents to emerging adulthood are needed to further shed light on a potential sex-specific switch in terms of overweight vulnerability. Specifically, sex-specific parental perceptions of ideal body weight among children and early adolescents may explain the higher prevalence rate of overweight status among (early adolescent) males. Parents seem more supportive of higher body weights of males compared with females, possibly because of the male body ideal (big is more ideal for males than females, [20,21]). As such, these explanations may thus explain our sex-specific findings involving overweight status, given that parents may still have a considerable influence on what their children eat (potentially impacting their weight development) during early adolescence [22].

We also found a significantly higher prevalence of overweight status among Indonesian adolescents who attended schools in urbanized areas compared with those in suburban areas. This has similarly been reported before among children and adolescents [4,11,23]. However, this finding should be interpreted carefully in our case because we also found a clear sex-specific interaction with school area. We found that specifically male adolescents in urban school areas had higher overweight rates. This finding is in line with another recent study among children and adolescents in Indonesia [11], but in contrast with previous studies among adults showing that females from urban areas were the ones at highest risk [2,5]. We speculate that (early adolescent) males may be more vulnerable to these unhealthy urban environments with junk food cues from fast-food outlets, as they often show higher efforts to get food as a reward compared with females [22,24]. As such, males might be more likely to actively search for food rewards, which are more often satisfied in high junk food environments. This, in combination with parents possibly more often encouraging adolescent males to gain weight [21], might explain our sex-specific interaction among early adolescents.

In our study, ethnicity was not related to overweight status, which is in contrast with some previous studies [13,25]. However, our findings involving ethnic background are consistent with the results of a previous study investigating adolescents from other Indonesia regions (i.e., Surakarta). This study also found no significant differences between Javanese and Chinese Indonesian adolescents [17]. So it might be that ethnicity findings regarding overweight status are dependent upon the specific Indonesian region (and ethnicities) being examined.

A final result of our study is that adolescents who dieted in the previous year were more likely to be overweight. This finding is in line with well-known findings from Western countries, with recent dieting considered to be a potential proxy of the susceptibility to weight gain [26]. It might be that dieting is unsuccessful and interspersed with binge eating episodes, thus leading to weight gain. Dieting may also be the consequence of being overweight [16]. Further longitudinal research is needed to unravel the directionality of these associations. Importantly, the fact that the dieting findings in this study were rather similar to the ones reported in previous European and American studies, suggests overlap in terms of overweight correlates between higher SEP Indonesian adolescents and adolescents from Western countries.

One notable strength of our study is the inclusion of a relatively large sample of early adolescents from specifically private schools, who are at higher risk for obesity, as also supported by our study findings. Another strength is that we used objectively measured weight and height to determine overweight status. Nevertheless, a couple of limitations should also be mentioned. First, we did not include clear markers for determining "socioeconomic" differences (except pocket money) within our higher SEP group of adolescents from private schools. The amount of pocket money that adolescents received might not reflect socioeconomic position differences. The income of the family per year might have been a better indicator (i.e., [5,12]). Nevertheless, as our total sample was recruited from private schools only, we are rather confident that most adolescents were from mid-to-high SEP backgrounds. Second, as our data are limited by a cross sectional design, we, therefore, do not know the underlying mechanism explaining the observed associations. Future longitudinal studies could shed more light on (predictors of) weight development in specific subgroups, such as males from urban areas compared with suburban areas.

Despite these limitations, our study examining (sex-specific) demographic correlates of overweight status among early adolescents from private schools in Indonesia filled an important gap in the current literature. We have speculated about the most prominent (mostly nutrient-related) mechanisms explaining our findings. Nevertheless, future research should further unravel the underlying (energy intake and expenditure) mechanisms explaining why particularly early adolescent males from urban school areas are more likely to be overweight. This will provide further tools for future tailored preventive interventions. We suggest that this early adolescent phase is a promising period for timely preventive interventions, given that adolescent overweight and obese status in Indonesia is more rapidly increasing in older compared with younger adolescents [8]. To conclude, our findings suggest that males from urban area private schools in Indonesia may be an important target group for future preventive overweight interventions.

**Author Contributions:** J.M.V. and J.K.L. were responsible for the study design; J.M.V. and J.K.L. supervised the data collection; E.S. was responsible for the data collection, the statistical analyses and interpretation of the data in agreement with J.M.V., J.K.L. and W.J.B.; and E.S. wrote the first version of the manuscript, edited by J.M.V., J.K.L. and W.J.B. All authors participated in the revisions of the manuscript. All authors have read and agreed to the published version of the manuscript.

**Funding:** We received no specific grant from any funding agency in public, commercial, or non-profit sectors. This study was funded by Maranatha Christian University in Bandung, Indonesia. The study received no external funding. The analysis and interpretation of the data and the writing of this manuscript were funded by the Behavioural Science Institute of Radboud University in Nijmegen, The Netherlands.

**Institutional Review Board Statement:** All procedures performed were in accordance with the ethical standards of Ethics Committee Social Science of Radboud University, Nijmegen, The Netherlands. Of the five schools that agreed to participate, three schools obtained consent forms from parents and students. The remaining two schools informed the parents about this project (passive consent) based on the school policy and collected the signed consent forms from students only. The original (reference ECSS\_2019\_150) and amended (passive consent) procedures were approved by the Ethics Committee Social Science of Radboud University, Nijmegen, The Netherlands (ECSS-2019-115).

**Informed Consent Statement:** Informed consent was obtained from all subjects involved in the study.

**Data Availability Statement:** The datasets generated and analyzed during the current study are not publicly available due to agreements we have made concerning the exchange and use of our data but are available from the corresponding author (E.S.) on reasonable request.

**Acknowledgments:** We would like to thank all the participating schools and students for their contribution to this research project. Moreover, we would like to thank all the student assistants (head of the schools and class teachers) for their help during the data collection of this project.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


## *Article* **Obesity Risk-Factor Variation Based on Island Clusters: A Secondary Analysis of Indonesian Basic Health Research 2018**

**Sri Astuti Thamrin 1,\*, Dian Sidik Arsyad 2,3, Hedi Kuswanto 1, Armin Lawi <sup>4</sup> and Andi Imam Arundhana 5,6**


**Abstract:** Obesity has become a rising global health problem affecting quality of life for adults. The objective of this study is to describe the prevalence of obesity in Indonesian adults based on the cluster of islands. The study also aims to identify the risk factors of obesity in each island cluster. This study analyzes the secondary data of Indonesian Basic Health Research 2018. Data for this analysis comprised 618,910 adults (≥18 years) randomly selected, proportionate to the population size throughout Indonesia. We included 20 variables for the socio-demographic and obesity-related risk factors for analysis. The obesity status was defined using Body Mass Index (BMI) <sup>≥</sup> 25 kg/m2. Our current study defines 7 major island clusters as the unit analysis consisting of 34 provinces in Indonesia. Descriptive analysis was conducted to determine the characteristics of the population and to calculate the prevalence of obesity within the provinces in each of the island clusters. Multivariate logistic regression analyses to calculate the odds ratios (ORs) was performed using SPSS version 27. The study results show that all the island clusters have at least one province with an obesity prevalence above the national prevalence (35.4%). Six out of twenty variables, comprising four dietary factors (the consumption of sweet food, high-salt food, meat, and carbonated drinks) and one psychological factor (mental health disorders), varied across the island clusters. In conclusion, there was a variation of obesity prevalence of the provinces within and between island clusters. The variation of risk factors found in each island cluster suggests that a government rethink of the current intervention strategies to address obesity is recommended.

**Keywords:** body weight; Indonesia; islands cluster; multiple logistic regression; obesity; risk factor

#### **1. Introduction**

Obesity is a major public health issue causing multiple burdens of co-morbidities and mortalities among adults. The World Health Organization (WHO) reported that globally 39% of adults were overweight and 13% were obese, and this number has nearly tripled within the last three decades [1]. In Indonesia, the obesity prevalence has increased significantly from 18.8% in 2007 to 26.6% in 2013, with a slight decrease in 2018 (21.8%) [2–4].

There are many statistical methods for analyzing large-scale study data. The machine learning method is a powerful statistical analysis approach that can be used for predictive model development of health outcomes. A recent systematic review reported various machine learning techniques that were performed to predict adult obesity from nationwide

**Citation:** Thamrin, S.A.; Arsyad, D.S.; Kuswanto, H.; Lawi, A.; Arundhana, A.I. Obesity Risk-Factor Variation Based on Island Clusters: A

Secondary Analysis of Indonesian Basic Health Research 2018. *Nutrients* **2022**, *14*, 971. https://doi.org/ 10.3390/nu14050971

Academic Editors: Odysseas Androutsos and Evangelia Charmandari

Received: 17 January 2022 Accepted: 22 February 2022 Published: 24 February 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

and large cross-sectional data, finding that logistic regression analysis had the highest accuracy in predicting obesity [5,6]. This finding is in line with our previous study [7], which found that logistic regression had the highest performance in predicting and measuring obesity. Predicting obesity risk factors by considering determinant variables can be advantageous to design and modify local existing nutrition programs and policies better for controlling the obesity problem.

To the best of our knowledge, this is the first study re-analyzing cross-sectional Indonesian Basic Health Research (RISKESDAS in Indonesian) data based on the main islands in Indonesia (we use the term "island clusters"). A previous study in Indonesia investigating the determinants of obesity among adults using the 2007 and 2013 RISKESDAS data concluded that the prevalence of obesity and risk factors varied among the areas [8]. However, this study only grouped the areas based on Indonesia's three different time regions, which might cause bias within the three groups. Therefore, further analysis for obesity determinants in regions with similar population characteristics is essential to minimize the variation bias. We clustered the provinces located on the same island into one cluster as the population characteristics within the same island cluster, assuming that within-island populations share more characteristics than clusters determined only by the time zone.

The main aim of this study is to examine the factors contributing to obesity in adults and investigate how these varied across the island clusters. This study also describes the prevalence of obesity in seven island clusters in Indonesia and reveals what factors increase or decrease the risk of obesity.

#### **2. Materials and Methods**

#### *2.1. Data Source*

Secondary data analysis performed in the current study was based on the data from the RISKESDAS study, a nationally representative cross-sectional study in Indonesia conducted by the Ministry of Health in 2018. Detailed information regarding methods, ethical considerations, and other related aspects of the RISKESDAS study is published elsewhere [9]. Briefly, the RISKESDAS sample was selected based on 2010 population census blocks using multi-stage cluster random sampling. Our data for analysis comprised 618,910 adults (≥18 years) from approximately 300,000 households randomly selected using proportionate to population sub-samples throughout Indonesia [2].

The data can be obtained from the National Institute of Health Research and Development (NIHRD), Ministry of Health, Republic of Indonesia upon request (https://www.litbang. kemkes.go.id/layanan-permintaan-data-riset/, accessed on 3 May 2021).

#### *2.2. Study Variables*

Socio-demographic variables, obesity status, and selected risk factors were identified from RISKESDAS 2018 questionnaires prior to the data request. We included 20 variables for socio-demographic and obesity-related risk factors for analysis. The socio-demographic variables of sex, education, employment, marital status, and urban-rural status were included.

Obesity status was calculated based on the Body Mass Index (BMI) using weight and height. We classify an individual as obese with a BMI ≥ 25 kg/m<sup>2</sup> following WHO BMI cut-offs for Asian populations [10]. Mental and Emotional Disorders (MEDs) were based on 20 items from a Self-Reporting Questionnaire (SRQ) developed by the WHO [11]; we determined a MED with the cut-off point ≥6 (positive predictive value = 70%, and negative predictive value = 92%) [12]. The frequency of consumption of risky food items (sweet foods and beverages, high-in-salt foods, high-in-fat foods, meats, carbonated drinks, energy drinks, and instant foods) was measured. The eating or drinking of risky food items more than 1 time per day was considered as high frequency consumption. Vegetable and fruit consumption was calculated based on the WHO standard [13]; ≥5 portion per day was determined as adequate. Smoking status was classified as "currently smoking", "quit smoking", and "never smoked", based on participant self-reporting. Physical activity in the current analysis was based on the WHO Global Physical Activity Questionnaire (GPAQ) used in the RISKESDAS study; sufficient physical activity was defined according to the WHO recommendations [14]. Drinking alcoholic beverages within 1 month prior to the study was defined as alcohol consumption. Blood pressure was measured using systolic and diastolic blood pressure during data collection; the 8th Joint National Committee guideline was used to classify blood pressures [15].

#### *2.3. Island Clusters*

Indonesia is the largest archipelagic country consisting of clusters of islands divided into 34 provinces. Our current study defined 7 major island clusters: Sumatra (provinces included: Aceh, North Sumatra, West Sumatra, Riau, Jambi, South Sumatra, Bengkulu, Lampung, Kepulauan Bangka Belitung, and Kepulauan Riau); Java (provinces included: DKI Jakarta, West Java, Central Java, Yogyakarta, East Java, and Banten); Bali-Nusa Tenggara (provinces included: Bali, West Nusa Tenggara, and East Nusa Tenggara); Kalimantan (provinces included: West Kalimantan, Central Kalimantan, South Kalimantan, East Kalimantan, and North Kalimantan); Sulawesi (provinces included: South Sulawesi, Central Sulawesi, Southeast Sulawesi, North Sulawesi, West Sulawesi, and Gorontalo); Maluku (provinces included: Maluku and North Maluku), and Papua (provinces included: Papua and West Papua).

#### *2.4. Statistical Analysis*

The sample weights for the complex survey design were considered in the analysis. Descriptive analysis was conducted to determine the characteristics of the population and to calculate the prevalence of obesity within the provinces in each island's cluster.

In order to calculate the adjusted odds ratios (ORs), multivariate logistic regression analyses, which includes other variables associated with obesity, were performed. The selection of multivariate logistic regression to develop a predictive model was based on our prior study that showed a high-performance, including accuracy, specificity, precision, Kappa, and *F<sup>β</sup>* [7]. Multivariate logistic regression was performed using SPSS version 27 (IBM Corp, Armonk, NY, USA). Cohen's and Cliff's Delta analyses were performed using R version 4.0.1 ('effsize' package version 0.7.6, Marco Torchiano, 2019) to validate each factor's variation (effect size) of odds ratios (ORs) by island cluster. The effect sizes were presented in four distinct categories: negligible, small, medium, and large. The effect size was considered negligible if the score was below 0.2, a score of 0.2 to <0.5 was considered small, 0.5 to <0.8 for medium, and 0.8 and above for a large effect size [16,17].

#### *2.5. Ethical Considerations*

Ethical approval of the RISKESDAS survey was obtained from the Ethical Committee of Health Research, the Indonesian Ministry of Health (LB.02.01/2/KE.267/2017) [2].

#### **3. Results**

#### *3.1. Prevalence of Obesity across Island Clusters*

The purpose of this study was to describe the prevalence of obese adults by island cluster in Indonesia and assess the risk of obesity caused by determinant factors using the secondary data of RISKESDAS 2018. Figure 1 illustrates the estimated obesity prevalence distribution across Indonesia. The figure shows that all island clusters have at least 1 province in which the prevalence of obesity is more than 20%, and only 3 clusters with at least 1 province in which more than 25% are classified as obese: Java, Kalimantan, and Sulawesi Island.

In Sumatra, North Sumatra had the highest number of obese adults, while Lampung had the lowest (40.6% vs. 29.6%). The sequence of the regency for the highest and the lowest obesity prevalence in other island clusters were: DKI Jakarta (45.6%) versus Central Java (33.3%) in Java Island; Bali (38.8%) versus East Nusa Tenggara (19.1%) in Bali and Nusa Tenggara Island; East Kalimantan (44.1%) versus West Kalimantan (30.3%) in Kalimantan Island; North Sulawesi (46.6%) versus West Sulawesi (31.4%) in Sulawesi Island; Maluku Utara (37.9%) versus Maluku (33.0%) in Maluku Island; and West Papua (39.6%) versus Papua (35.0%) in Papua Island. DKI Jakarta had the highest proportion of an obese individuals, which outnumbered the national figure (45.6% vs. 35.4%), followed by Sulawesi Utara and East Kalimantan (data displayed in detail in the Supplementary Table S1A).

Table 1 shows the breakdown of obesity prevalence in all the island clusters in Indonesia, according to categories of the determinants associated with obesity. Notably, this table indicates that obesity prevalence varied according to MED status. The distribution of obesity was higher among those without MED in the Sumatra, Java, and Papua Island clusters. In addition, a few variables show the variation of obesity prevalence by categories, including food high in salt, meat, carbonated beverage consumption, and smoking status. Individuals with a high level of education and having a permanent job (e.g., government/police/military officer) show a higher obesity prevalence than their counterparts.

**Table 1.** Prevalence of obese adults by variables associated with obesity in all island clusters in Indonesia.


**Table 1.** *Cont.*



**Table 1.** *Cont.*

#### *3.2. Cluster Variation of Obesity Risk Factors*

The results of the logistic regression analysis are displayed in Figure 2. The variables associated with increased odds (OR > 1) of being obese in all island clusters were the location (X01), gender (X02), marital status (X03), occupational status (X06), high-fat food (X11), and blood pressure (X20). Meanwhile, age group (X04), educational level (X05), sugar-sweetened beverage consumption (X09), FV consumption (X16), and smoking status (X17) factors statistically did not appear to increase the obesity risk in Indonesian adults.

**Figure 2.** Variation of obesity risk factors by island clusters in Indonesia.

These results suggest that six out of twenty variables show a variation among island clusters. For example, adults working as a government/military/police officer are over represented in the data, suggesting that employment in these sectors is one of the highest contributing factors to obesity in all clusters. However, working as an entrepreneur is the highest contributor to obesity levels in Bali and Nusa Tenggara Islands (OR = 1.775; 95% CI: 1.764–1.786). People with mental–emotional disorders are likely to be at risk of obesity only in Sumatra and Papua. Regarding smoking status, those who are still smoking tend to have a normal weight status compared to their counterparts. The OR number is displayed in more detail in Supplementary Table S2A. In order to measure the effect size of each variable across the island clusters, we performed Cohen's and Cliff's Delta analyses (Supplementary Table S3A–N).

#### **4. Discussion**

The present study found that some island clusters share common obesity risk factors, including individual and socio-economic factors. Individual factors (e.g., sex, high-fat food consumption, and blood pressure) and socio-economic factors (e.g., marital status, occupational status, and location) contributed to the risk of obesity in seven island clusters, indicating that these variables are probably strong predictors of obesity in Indonesia.

The present study showed that women have greater odds of being obese than men. Women typically have a body fat percentage around 10% higher than men [18,19]. Although other biological factors, such as age and ethnicity, also contribute to the adiposity distribution and percentage, women still have more considerable body fat in almost all life spans [18]. This result aligned with a previous study in a developing country that reported that the prevalence of obesity in adult females was higher than in males. The study also revealed that obesity was directly proportional to age, but only for females [20]. We also performed additional analyses to measure the risk of obesity among married women. The risk of obesity among women who were already married was likely to increase, but not for those living in the Ambon and Papua Island clusters. This may be due to socio-economic factors forcing women in these clusters to earn money [21], thus increasing metabolic energy expenditure.

In this study, high-fat food consumption (X11) was shown to be the only dietary factor that significantly contributed to the risk of obesity in Indonesian adults in all clusters. This finding was in line with a previous study, which found that the consumption of food containing a high-fat content was the risk factor of obesity in all regions in Indonesia and was consistently found in 2007 and 2013 [8]. These results are likely to be related to people's eating habits differing by region. Eastern Indonesia tends to consume high-fat foods. It can be seen that the clusters of Sulawesi and Bali and Nusa Tenggara have a 1.4 times higher risk factor of being obese due to high-fat food consumption. Fat contributes significantly to the total energy intake, and thus reducing the high-fat food consumption might balance energy expenditure and intake [22]. Additionally, it should be noted that some fat types have beneficial effects on obesity. For example, replacing protein and short fatty acid (SFA) with polyunsaturated fatty acid (PUFA) has been shown to be significantly associated with a lower obesity risk [23].

Level of education (X5), age group (X4), sweet-sweetened beverages (SSBs) consumption (X9), energy drinks (X14), fruit and vegetable (FV) consumption (X16), and smoking status (X17), on the other hand, are statistically considered a lower obesity risk. Adults with low education levels have a lower risk of being obese than those with high education levels. This result is inconsistent with a larger cross-sectional study that found that the lower the years of education, the higher the odds of obesity [24]. What is surprising is that individuals that consumed sugar-sweetened and energy beverages had lower odds of being obese. This finding is contrary to many previous studies, which suggested that a greater intake of SSBs was associated with being overweight and obesity in children and adults [25,26]. This inconsistency might be due to the type of sugar contained in the beverages. Studies have shown that fructose-sweetened beverages increase adiposity levels more significantly than sucrose-sweetened beverages [25,27]. Unfortunately, we did not identify the dominant type of sugar contained in the beverages. Moreover, although low

FV consumption reduces the risk of obesity, it shows a positive effect on the interaction between high-fat food and obesity. People with inadequate FV consumption are likely to consume more high-fat foods than those with adequate FV in-takes.

Adults who already quit smoking had increased odds of obesity in all the island clusters. This may be due to the effect nicotine has on the central nervous system and metabolism with two possible mechanisms. First, people who quit smoking tobacco tend to replace the hand-to-mouth smoking activity with eating, leading to an increase in calorie intake. Second, taste preference is also changed among those who quit smoking to obtain a pleasure "sensation" replacing the effect of tobacco [28]. However, this finding must be interpreted with caution because we have no baseline data to describe how much weight gain occurred after smoking cessation. Additionally, the use of tobacco can promote other diseases, such as cardiovascular diseases (CVDs), hypertension, and even mortality, the same risk posed to obese people [29]. Obesity and tobacco use do not actually show the opposite outcomes. Combining weight management and smoking cessation treatment might be promising in order to improve health quality and prevent the risk of metabolic diseases related to obesity and smoking behavior.

Interestingly, mental health disorders (X7), sweet-food consumption (X8), high-salt food (X10), meat consumption (X12), and carbonated drinks (X13) varied across the island clusters. Health-related behaviors might be different between the island clusters due to several factors, such as health inequalities, socio-economic status, or household deprivation [30,31]. The most obvious variation was carbonated drink consumption, which is a high-risk factor in Papua Island, but is not excessively consumed in Bali and Nusa Tenggara Island (OR = 1.81 vs. 0.84). It seems that this variation occurs because the adult diet in Bali and Nusa Tenggara is different from Papua. There is a great difference in OR values for other diet factors, such as high-salt food, high-fat food, and energy drink consumption. Similar to our findings, a study on children and adolescents found that the consumption of SSB (including carbonated drinks) varied by race/ethnicity, sex, and age [32].

Variation among clusters was also found for the mental health disorder factor (X7). This study shows that Papua Island had the highest risk of obesity caused by mental health disorders. Many conditions could trigger mental health issues, including the limitation of food choices, poor access to health services, high-risk behaviors, and poor education [33], all of which are challenges faced on Papua Island. However, there is no information about the source of mental health disturbance of the population. More extensive evidence investigating the roles of mental health in mediating obesity occurrence is needed.

We also noted that physical activity was a predictor of obesity in all island clusters. This result was consistent with many prior studies conducted in Indonesia and elsewhere that reported that a lack of physical activity is strongly associated with obesity [34–39]. Performing sufficient physical activity might be beneficial to maintain people's energy expenditure and subsequent energy balance. Therefore, health promotion and education to improve physical activity are required, especially for busy adults in urban areas.

Another important finding of our study is that the prevalence of adult obesity varied across the regencies within the island clusters. The regency with the highest obesity prevalence was Jakarta (28.6%), while the highest among the island clusters was in Java (21.2%). The breakdown of the data reveals that the obese adults in urban areas outnumbered those in the rural areas. This difference can probably be explained by the socio-economic characteristics of each cluster. The Java Island cluster, including Jakarta province, are dominated by people living in urban areas or at least adjacent to urban areas. Meanwhile, urban communities are more likely to have unhealthy lifestyles, such as sedentary behavior and consuming more "unhealthy foods" [40,41]. Meanwhile, in the Sulawesi Island cluster and other island clusters outside the Java cluster, the number and distribution of urban communities is relatively fewer and uneven. A similar finding was reported by a study in Ethiopia that found that men living in metropolitan cities were 1.8 times more likely to become obese than those living in rural areas [36].

The major strength of the present study is that it includes a large sample size. In addition, using weighted factors in the analysis might generate results that more closely represent the Indonesian population. We also acknowledge some limitations in this study. First, the study data were collected cross-sectionally. Therefore, the causality of risk factors and obesity should be cautiously interpreted. Second, we did not disqualify outlier BMI measurement results in the dataset. However, this was only < 1% of the samples, which probably caused a small effect in the analysis. Lastly, the data for high-risk foods were collected using a non-validated questionnaire, which raised response biases of participants' answers regarding the consumption in the past 30 days. However, since the questionnaire was developed using neutrally worded questions, the options were not led from one to another and did not overlap each other, so that the participants might understand and respond more easily.

#### **5. Conclusions**

The study implies that there was a variation of obesity prevalence among the provinces and between island clusters. This study provides evidence that obesity risk factors varied across the island clusters, which may have implications in rethinking and redesigning policies and interventions to address the obesity problem in Indonesia. Multiple interventions that address specifically greater risk factors considering cluster characteristics are more likely to be effective in preventing obesity and its negative implications.

**Supplementary Materials:** The following supporting information can be downloaded at: https:// www.mdpi.com/article/10.3390/nu14050971/s1, Table S1A: Risk factors of obesity by island clusters in Indonesia from RISKESDAS 2018, Table S2A: Odds ratios of obesity based on the risk factors in Indonesia from RISKESDAS 2018. Table S3(A–N). Effect size of odds ratios for contribution variables.

**Author Contributions:** Conceptualization and methodology, S.A.T. and D.S.A.; interpretation of data, S.A.T., D.S.A. and A.I.A.; statistical analysis, S.A.T., H.K. and A.L.; visualization, D.S.A., H.K. and A.L.; drafting of the manuscript, S.A.T., D.S.A. and A.I.A. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Ministry of Education, Culture, Research and Technology through the PDUPT Grant contract No. 752/UN4.22/PT.01.03/2021.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Written informed consent was obtained from participants prior to collecting biospecimen samples.

**Data Availability Statement:** The datasets generated and analyzed for this study can be found via https://www.litbang.kemkes.go.id/ (accessed on 3 May 2021) through a request process at the Institute of Health Research and Development of the Indonesian Ministry of Health.

**Acknowledgments:** The first author and the corresponding author would like to thank the Ministry of Education, Culture, Research and Technology for providing funding to conduct this study. In addition, thanks also to the Ministry of Health through the Research and Community Development Agency for providing access to Indonesia's RISKESDAS data.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**

