Sociodemographic questions
A socio-economic index for area (SEIFA) score was calculated based on the residential postcode of participants. A SEIFA score is an Australian Bureau of Statistics (ABS) product used to rank areas in Australia based on relative socioeconomic advantage or disadvantage [15]. Other demographic information collected included age, school grade, gender, sex, and country of birth.
Eating disorder diagnoses
The operationalisation of the eating disorder diagnoses that were assessed in this study has been published previously [5], and an adapted Table of this operationalisation is included in the Supplementary material. Diagnoses assessed included the three major eating disorders (anorexia nervosa, bulimia nervosa, binge eating disorder), the five Other Specified Feeding and Eating Disorders (OSFED; atypical anorexia nervosa, subthreshold bulimia nervosa, subthreshold binge eating disorder, purging disorder, night eating syndrome) and Unspecified Feeding and Eating Disorder (UFED). Most symptoms were captured by items of the Eating Disorder Examination Questionnaire (EDE-Q), which assesses the presence and severity of cognitive and behavioural eating disorder symptoms and features [16]. This questionnaire has previously been validated in Australian adolescent boys and girls and demonstrates sound reliability [17]. Items used in this study included the behavioural frequency items (self-induced vomiting, laxative misuse, driven exercise, and binge eating), and the Likert-type items that comprise the combined weight and shape concern subscales. As the frequency of behaviours were only assessed over the past 1 month (not the 3 months duration required for bulimia nervosa and binge eating disorder), we use the term “probable” for these diagnoses. Cronbach’s alpha for the combined weight and shape concern subscale in the present study was 0.96 for both First-Australian and other-Australian adolescents.
Participant’s self-reported current weight and height, which was converted to age and gender adjusted body mass index (BMI) percentiles for children and adolescents. A BMI percentile < 10 was used for the underweight criterion of anorexia nervosa, as this cut-off has most frequently been used in adolescent epidemiological studies of DSM-5 anorexia nervosa [18,19,20,21]. Three items from the Night Eating Questionnaire (NEQ) [22] were used to assess symptoms of night eating syndrome, including proportion of daily food intake consumed following supper, nocturnal eating (eating after going to bed), and awareness during nocturnal eating. The NEQ has been validated in adolescents and is superior to parent report [23].
Several additional questions were developed by the researchers to capture frequency of additional extreme weight control behaviours (fasting, strict dieting, detoxes, insulin misuse, other drug use for weight loss), distress associated with binge eating, and additional diagnostic binge eating disorder features (e.g., eating faster than usual, eating alone due to embarrassment). Participants were also asked about any recent weight loss in the past 4 weeks to assess atypical anorexia nervosa.
Quality of life impairment
Scores from the Paediatric Quality of Life Scale (PedsQL) SF15 [24, 25] were used to measure quality of life impairment. The 12 items from the physical functioning, emotional functioning, and social functioning subscales were included in the survey. Items ask participants to indicate on a Likert type scale how true a series of statements are of them in the past 4 weeks. Scores are reversed and transformed on a 0–100 scale, such that higher scores indicate higher functioning. Subscale scores are derived as the mean of the items for that scale. For the purposes of this study we combined the emotional and social functioning scales to create a psychosocial subscale. The PedsQL SF15 has evidence of good reliability and validity in previous studies of adolescents [25]. Cronbach’s alphas in the current study sample for the physical functioning subscale was 0.89 and 0.84 for First Australian and other-Australian adolescents respectively, and for the psychosocial functioning subscale was 0.91 and 0.90 for First-Australian and other-Australian adolescents respectively.
Statistical analysis
Analyses used data weighted to the distribution of gender in adolescents in New South Wales in the 2016 Australian Census. A series of Chi-square tests were performed to determine the prevalence of eating disorders among First-Australian adolescents and to compare this to other-Australian adolescents. This was followed by 2 × 2 ANOVAs and logistic regression analyses to examine the moderation effect of First-Australian status on the sociodemographic (dependent variables: age, sex, SEIFA score, BMI percentile) distribution of eating disorders. Additional 2 × 2 ANOVAs and logistic regressions were employed to test the moderation effect of First-Australian status on the association between eating disorders and specific symptoms (dependent variables: weight and shape concerns, binge eating, self-induced vomiting, laxative use, dietary fasting, driven exercise) and impairment (psychosocial quality of life and physical quality of life impairment). Finally, four logistic regressions were performed with current eating disorder status as the outcome. In regression one, First-Australian status was the only predictor in the model; in regression two, sociodemographic variables (age, gender, SEIFA score, and BMI percentile) were added to the model; in regression three, eating disorder symptom variables (weight and shape concerns, binge eating, self-induced vomiting, laxative use, fasting, driven exercise) were added to the model; in regression four, impairment variables (psychosocial and physical quality of life impairment) were added to the model. The intention of these analyses was to examine the extent to which First-Australian status remained independently associated with the likelihood to meet criteria for a current eating disorder in the presence of other known correlates. In these analyses, SEIFA score was divided by 100 (the normed SD), BMI percentile was divided by 10, and weight/shape concerns and psychosocial and physical quality of life impairment scores were standardised, to allow more meaningful interpretation of the adjusted odds ratios.