Participants
In total 362 participants aged 18 and over who self-identified as engaging in dieting behaviours and/or having experienced mental health symptoms took part in the study.
Measures
Demographic data including age, gender, highest education (high school, vocational training, graduate, postgraduate, other), geographic location (rural, remote, metropolitan), and self-reported height (cm) and weight (Kg) were collected. Body Mass Index was calculated by dividing self-reported weight by height in metres squared.
Orthorexia Nervosa
The ORTO-15 was used to measure ON [15]. It consists of 15 items that assess beliefs about the perceived effects of eating health food, attitudes governing food selection, habits of food consumptions and the extent to which food concerns influence daily life. Responses are recorded on a four-point Likert scale, with lower score indicating more severe symptoms. Various cut-scores have been proposed in the literature, including 40 and 35 [see 15, 16, 17]. Poor to adequate internal consistency has been reported previously [e.g., 18], consistent with this study (a = .402). A shortened version, the ORTO-9, which uses nine of the original items has been proposed as an alternative, and is considered preferable for categorical diagnosis with a cut-score of 26 [19]. Internal consistency in the present study for ORTO-9 was a = .702. Despite these psychometric limitations, the ORTO-15 is the most commonly used tool to measure ON [1, 7].
Anorexia Nervosa
The 26-item Eating Attitudes Test (EAT-26) was used to assess eating behaviours characteristic of AN and other EDs [20]. Responses are rated on a six-point Likert scale with higher scores indicating greater symptom severity (range 0-78). A cut-score of 20 is most commonly used and has high specificity [21]. The EAT-26 has good concurrent, criterion and discriminant validity [22], as well as high internal consistency [a =.900; 20]; consistent with this sample (a = .928).
Obsessive-Compulsive Disorder
The 18-item Revised Obsessive-Compulsive Inventory (OCI-R) was used to assesses OCD [23]. Frequency of experiencing symptoms are reported on a five-point Likert scale with higher scores indicating greater symptom severity. The measure was developed according to the DSM-IV criteria and includes Hoarding symptoms. To reflect the changes in DSM-5 with Hoarding now a distinct disorder [24], a 15-item version (OCI-OCD) that eliminates three items pertaining to Hoarding has been validated for assessing OCD, with a cut-score of 12 proposed [25]. In addition, the measure can produce subscales, including an Obsessional thinking scale (3 items), with a cut-score of 5 for this subscale [23]. The OCI-OCD has shown to have good convergent and discriminant validity [25]. The scale has excellent internal consistency (a = .929), as does the Obsessions subscale (a=.903).
Procedure
Ethics approval was granted by the University of New England Human Research Ethics Committee. Participants were recruited through social media sites of mental health foundations, eating disorder groups, OCD-support groups, dieting groups, and eating and mental health related support groups. Interested participants were directed to an online survey hosted by Qualtrics, which took approximately 20 minutes.
Data Analysis
To determine if ON measures are able to predict the presence/absence of AN and OCD in order to provide an indication of the diagnostic classification of ON, Receiver Operating Characteristics (ROC) curve analyses were conducted using SPSS version 26. The Area Under the Curve (AUC) was used to ascertain the predictive ability of the ORTO-15 for predicting presence of AN and OCD. A diagnosis of AN was determined as those with an EAT-26 score above clinical cut-off and a BMI less than 17.5 [as per DSM-5 criteria for low body weight; 24]. A diagnosis of OCD was determined as those scoring above cut-off on the OCI-OCD. Sensitivity, specificity, PPV and NPV are reported for the ORTO cut-scores established in the literature, in addition to the best cut-scores on the measures for predicting AN and OCD, as determined by the Youden Index [26]. Power for conducting ROC analyses, has been proposed as needing at least 10 participants with a diagnosis, 10 without a diagnosis, 10 false positives and 10 false negatives [27]. Due to a lack of agreed upon diagnostic criteria or validated diagnostic interview for ON, a final determination of false positives and negatives for ON was not possible. However, given that there were over 10 cases for presence and absence of each diagnosis (see Table 1), the large sample, and the study aims, it was considered appropriate to conduct the ROC analyses.
Due to the low internal validity and questionable psychometric properties of the ORTO-15, secondary analyses were conducted using the ORTO-9. Due to the core overlapping feature of ON and OCD being related to obsessions [4], secondary analyses were also conducted examining the predictive ability of the ORTO-15 and ORTO-9 for predicting Obsessional thinking. Not all respondents completed the OCI-R, however, data were missing at random and there were no significant demographic differences between those who did and did not complete each measure. Therefore, all available data were used for each analysis (OCI-OCD, n=269, obsessions subscale, n=274).