Study design and setting
This cross-sectional study was conducted in six coeducational schools (three unaided private and three government-aided public schools) in Mumbai, India. For administrative purposes, the district administration of the Greater Mumbai area is divided into seven zones and each of these zones comprises 3 to 5 administrative wards (38). For the study, two wards from a randomly selected zone were selected and twenty private and twenty public schools from the selected wards were invited to participate. The first three private and public schools each that provided permission were selected as the study sites.
All students, attending grades 6 and 7 at the selected schools (n=6) were eligible to participate. Out of 1086 eligible students, 765 (70.4% of those invited to participate) provided written parental consent and 712 adolescents completed the survey. Given that in India, private schools are typically attended by students who belong to upper or upper-middle-class families and public schools cater to low-middle and low socioeconomic backgrounds. Similar to previous studies, the type of school attended by adolescents was used as a proxy indicator of the SES (15,39). The study protocol was approved by Intersystem Biomedica Ethics Committee, Mumbai, India (Approval Version 02/ dated 19 February 2019). Written informed consent from parents and written informed assent were obtained from adolescents.
Sample size estimation
The sample size was estimated to be 323 at 5% precision, 95 % Confidence Interval (CI) with the assumption that 70% of urban adolescents in India practice unhealthy eating behaviors (8,40,41). After taking into account a 20% non-response rate and a proportional representation from adolescents attending private and public schools, the sample size was calculated as 775.
Adolescents reported their age, sex, grade (grade 6 or 7), type of living arrangement (nuclear or joint), and religion (Hindu, Muslim, Christian, and others) and the parents/caregivers provided socio-economic information- parents’ highest educational qualifications and the family’s monthly income (Table S1).
Adolescents provided information about eating habits such as weekly frequencies of consuming breakfast at home, carrying lunch or snacks to school, and watching television while eating at home. Adolescents were asked ‘In the last 7 days, how many days did you perform the following activities….”. The response options, ‘never’ to ‘always, indicating < 1/week to 6-7 times/ week, were scored on a five-point scale from 0 to 4. Higher scores indicated a higher frequency of indulging in a specific eating habit.
Consumption of unhealthy snacks
In line with the previous studies (7,10,42), the snacks that are high in calories, salt, and sugar content and low in nutrients were considered unhealthy. We employed a brief qualitative food frequency questionnaire (FFQ) to assess adolescents’ consumption of unhealthy snacks during the past week. The FFQ was validated as a part of our previous study to evaluate the practices related to healthy eating in 10-12 years old adolescents in Mumbai (43). In summary, the steps involved creating a list of locally available and commonly consumed unhealthy snack items based on the review of previous studies and the results of a pilot study entailing two non-consecutive interviewer-administered 24-hour diet recalls (24h DR) in adolescents (n= 55) in Mumbai. This was followed by the development of a food list comprising 28 snacks that were high in fats, sugar, and salt. This FFQ draft was next pretested among adolescents (n=32) who were asked to report the weekly frequencies of consuming specific foods and indicate the most appropriate portion sizes of the consumed food items in standardized household measures such as katoris (bowl or cup), spoons, and glasses. Food items that were consumed infrequently (less than once a week) and/or had lower mean daily intakes (< 10g/d) were removed from the list. For data sorting and analysis, similar items were grouped into four broad categories- (1) Snacks high in fat/ fast foods - 4 items (2) Snacks with added sugar - 3 items (3) Snacks with added salt - 4 items and (4) carbonated beverages - 1 item. The FFQ responses were scored 0 to 4, from ‘never’ to ‘2- 3 times a day. The scores corresponding to the consumption of all snack items were aggregated to derive total unhealthy snack consumption scores, possible scores 0 to 48, with higher scores reflecting higher consumption of unhealthy snacks. The aggregate unhealthy snack consumption scores were subsequently categorized into tertiles indicating high, moderate, and low consumption. Moreover, the consumption frequencies were calculated as the number of days/ weeks by coding the response options ‘never’ as 0 d/week, ‘sometimes’ as 1.5 d/week, ‘often’ as 3.5 d/week, ‘frequently’ as 5.5 d/week, and ‘always’ as 7 d/week.
School and home food environments of adolescents
The school and home food environment-related questions were based on a 36 item self-reported questionnaire measuring Food-Related Environments at Schools and Homes (FRESH-Q) in adolescents in India (44). The specific items in the questionnaire were developed based on the results of focus group discussions with adolescents, parents, and teachers reported elsewhere (31) and a thorough review of other food environment-related questionnaires administered in adolescents (45–48). In the questionnaire, the school food environment items examined the type, price, and quality of foods and beverages available for sale in the school canteens or at the nearby street vendors for the schools that did not have an on-site canteen. The adolescents were asked ‘Do you eat at the school canteen or buy foods/beverages at school?’ and ‘If yes, then how often…? Additional questions included ‘Which of the following foods and drinks are available at school to buy?’, ‘Which of the following foods and drinks do you usually buy at school?’, and ‘Why do you buy these foods? (Response options included taste, price, availability, and convenience)’.
To explore factors within home food environments, the adolescents were asked to report the weekly frequencies of availability and accessibility of unhealthy snacks at home. The participants were asked ‘In the last 7 days, how many times were the following foods available at home’ and ‘how many times were the following foods kept in easy to reach places such as countertops or kitchen cabinets? The listed food items included the same 11 unhealthy snack items and 1 carbonated beverage item as reported in FFQ. Moreover, three questions assessed the perceived parental control during mealtime. Response options were ‘strongly disagree’ to ‘strongly agree’, scored from 0 to 4, with higher scores indicating higher perceived parental control on adolescents’ consumption patterns. An additional three questions examined the family dietary habits such as weekly frequency of eating out or ordering takeaways, having at least one meal/d with family, and having meals at the dinner table.
The content validity of the questionnaire was examined by a panel of four experts including two public health nutritionists, a food environment researcher, and an academician. The principal axis method of exploratory factor analysis was used to establish the construct validity and the internal consistency was evaluated using Cronbach alpha values > 0.7 (44). To test the reliability of the items in FRESH-Q, a sample of adolescents (n=108) took a retest of the same questionnaire after two weeks and the test-retest correlation coefficient values of each item were calculated. Based on the content and construct validity measures and the internal consistency and test-retest correlation values, the items were retained or excluded in the final instrument. The summary of items included in the questionnaire is provided in Supplementary Material, Table S1.
Perceptions of barriers in food environments
The adolescents’ perceptions related to eating behaviors and food environments were measured using sixteen statements, based on the constructs of a widely used health behavior framework, the Health Belief Model (HBM). The items assessed adolescents’ perceived susceptibility and severity of adverse consequences of unhealthy eating behaviors, perceived barriers and benefits within food environments, and readiness to change and self–efficacy to adopt healthier dietary practices (49). Responses to the statements were assessed on a five-point Likert scale from ‘strongly disagree to strongly agree’, numeric scores 0 to 4.
Pilot testing of the questionnaire
The face validity of the questionnaire including socio-demographic characteristics, eating habits, frequency of consumption of unhealthy snacks, food environments, and adolescents’ perceptions were evaluated in 34 adolescents, ages 10-12 years (20 from a public school and 14 from a private school). These participants were not included in the final data analysis. All the questions were well understood, and no revisions were considered. The field investigators having postgraduate degrees in nutrition received a full day of training to review the protocol and methods of data collection. The questionnaire was administered to adolescents during school hours in the presence of school teachers and the trained field research team. The summary of items in the questionnaire is provided in Supplementary Material Table S1.
Adolescents (n=712; private schools n= 384 and public schools n= 328) who were present in school on the survey day and had provided >70% of complete information and a written, informed assent to participate in the study comprised the final sample for analysis. The data were analyzed using IBM SPSS for Windows version 20.0 (IBM Corporation, Armonk, NY, USA). Assuming a normal distribution, the demographic variables, eating habits, unhealthy snack consumption patterns, food environments, and perceptions were compared between private and public-school adolescents using frequency analysis, chi-square statistics, cross-tabulation, and student t-tests. For the comparative analysis of perceptions between private and public-school adolescents, the response options of ‘strongly agree’ and ‘agree’ were categorized as a single category and ‘strongly disagree’ and ‘disagree’ as another while the response option ‘neither agree nor disagree’ was retained as a separate category. Univariate regression analyses were performed to determine the unadjusted effects of independent variables – demographic and food environment characteristics, eating habits, and adolescents’ perceptions on the dichotomized dependent variables- unhealthy snack consumption (highest tertile of aggregate scores indicating high consumption of unhealthy snacks and moderate/lowest tertiles of consumption) in separate models for private and public-school adolescents. Next, we used a backward stepwise logistic regression method to eliminate the least significant predictor variables and derive a reduced model that best explained the data. The independent variables that remained were entered as the covariates in the final multivariate regression model and the adjusted odds ratio with 95% confidence interval (CI) of odds ratio were calculated, considering p values ≤ 0.05 as a measure of significant association between variables.