Participants
To be included in the study, participants had to be at least 16 years of age, reporting current or historical ED symptoms, and own at least one SNS account. In total, 163 participants provided informed consent, but only 146 met the inclusion criteria (those who did not, were automatically directed to the end of the survey). One hundred and three complete survey responses were collected from visitors of a large e-community for individuals with eating problems or disorders as well as from three patient organizations supporting people with EDs. The ad text stated that the aim was to investigate what motivates people with (a history of) eating disorder problems to use social media. The rationale behind not having stricter inclusion criteria (e.g. requiring a diagnosis, or scoring above a clinical cut-off) was to obtain a diverse, representative sample to improve the generalizability of the results(see Table 1. for descriptive information regarding the demographic and clinical variables).
Design and Procedure
A cross-sectional research design was applied. An online survey, taking approximately 20 minutes to complete, was created on a secure SurveyMonkey account. Access to the survey was provided via a link on the websites from 27 May 2020 until 1 August 2020. Following study information and informed consent, participants were asked to complete screening questions relating to inclusion/exclusion criteria. If they did not meet the criteria, they were directed to the end of the survey. Those who completed the study, were offered the option to enter their email address to participate in a raffle to win one of ten bol.com vouchers (10 euros in value). Email addresses were the only identifying information collected, and these were removed from the dataset and kept separately and securely from the rest of the data. The email addresses were only used to contact winners once the study had ended. The Psychology Research Ethics Committee of the Leiden University approved this study.
Measures
Demographics.
Participants were asked to indicate their gender (male, female, other); age (in years); living situation (living with parents/family, living alone, living with partner/children, student housing, other/prefer not to say); highest educational level achieved (answering options: primary, vocational, higher professional, university), employment status (student, homemaker, employed fulltime of part-time, unemployed, disability/unable to work, other)), civil status (single, co-habiting/married/divorced), , and country of birth (Netherlands, Morocco, Antilles, Turkey, Suriname, other). Eating disorder psychopathology
Eating disorder history
To further describe the sample, participants were asked about current/past ED treatment, and the duration of their symptoms.
Eating disorder symptom severity
The Eating Disorder Examination Questionnaire (EDE-Q 6.0) (Fairburn et al., 2008) consists of 28 items assessing the frequency and severity of ED symptoms over the past 28 days, using a seven-point Likert-scale ranging from 0 to 6. These include six questions assessing the frequency of core ED behaviors (binge eating, self-induced vomiting, laxative use, excessive exercise, fasting, use of diuretics), and 22 questions relating to psychological features (dietary restraint, eating concern, weight concern, and shape concern). A global score of eating psychopathology was obtained by averaging the 22 items relating to psychological features, with higher scores reflecting more severe ED psychopathology (Aardoom et al., 2012). Internal consistency was high for those 22 items (α = 0.96). The questionnaire also asked about weight and height.
Body mass index (BMI) was computed by dividing the self-reported weight (in kilograms) by the squared self-reported height (in meters)
Body (Dis)satisfaction
Participants were asked to rate their (dis)satisfaction with their appearance, the shape and size of their body, their weight, and their physical attractiveness compared to others, on a scale from 1 (very dissatisfied) to 10 (very satisfied). The total score was obtained by computing the mean of the four responses (Veldhuis et al., 2017). Internal consistency was high (α = 0.92).
Self-Esteem
Participants were asked to complete the Single-Item Self-Esteem Scale (SISE) (Robins, 2001), a one-item measure of global self-esteem (“I have high self-esteem.”). The answer was provided on a 5-point Likert scale, ranging from 1 (not at all true of me) to 5 (very true of me).
Readiness to Change
The Eating Disorder Readiness Ruler (ED-RR) (St-Hilaire et al., 2017) is an 18-items questionnaire examining readiness to change in nine ED symptom domains paralleling those in the EDE-Q. In the present study only the first section was used (items 1 to 9) measuring readiness to change over the past 28 days. Responses ranged from 1 to 10 (1–2 = not at all ready to change, 3–5 = unsure, 6–8 = ready to change, 9–10 = actively changing already). There was also the option of responding “not applicable”, if the behaviour or cognition in question was not relevant to the participant. A mean score for readiness to change was obtained by computing a summed score and then dividing the sum by the number of items completed for each participant so symptom domains that were not applicable to the participant were not included in the calculation(St-Hilaire et al., 2017). A higher score reflected a higher readiness to change. Cronbach’s alpha was only calculated for the items that had more than 50% endorsement: dietary restraint, shape concern, weight concern, and excessive exercise. Internal consistency was good for these four items (α = 0.84).
General internet and social media use
Participants were asked about their amount of average daily internet and social media use (Meier et al., 2014). The options were: never/almost never, less than an hour a day, between one and two hours a day, between two and three hours a day, or more than four hours a day.
Specific SNS use
For the present study SNS was defined as applications and websites that enable users to create and share content with networks (i.e., friends, followers, etc.) they construct for themselves (Pittman et al., 2016). The focus was on general-purpose sites, and therefore specialized instant messaging platforms (e.g. WhatsApp), career services (e.g. LinkedIn), and dating applications (e.g. Tinder) were excluded. Similarly, specialized pro-ED (pro-ana/pro-mia), and mHealth/eHealth and fitness/diet mobile applications were beyond the scope of the present study. The SNSs included were the most popular general-purpose ones in the Netherlands (Statista, 2023): Facebook, Instagram, Pinterest, Twitter, YouTube, Snapchat, and Tumblr. Given the recent popularity of the TikTok platform (Statista, 2023), this was added as one of the options. Participants were asked to indicate the amount of time they spent daily on each of the platforms using two drop-down menus: one for hours per day and another for minutes per day. These were combined into overall minutes per day. Total daily SNS use was computed by adding the responses for each platform together (Yellowlees et al., 2019).
Motives for SNS use
Participants were asked to express their agreement/ disagreement with 75 statements obtained from the uses and gratifications literature (for the origin of the items and specific references: see supplemental Table 1). These statements measured the following 16 motives: enjoyment (three items), social interaction (eight items), passing time (five items), surveillance (seven items), information-seeking (five items), information-sharing (three items), relaxation (four items), avoiding loneliness (three items), escapism (three items), support (four items), self-documentation (four items), connecting with similar others (seven items), self-expression (five items), social pressure (six items), popularity (five items), and self-presentation (three items). These motives were selected because of their relevance in several different studies and applicability to several SNS platforms. The statements were on a five-point Likert-type scale anchored by “Strongly disagree” (1) to “Strongly agree” (5). The scores of the 16 variables were obtained by computing the median of the individual items measuring the variables. For example, a participant’s score for “enjoyment” was determined by obtaining the median of the responses to the three items measuring enjoyment: “Because it is fun”, “Because it is enjoyable”, “Because it is entertaining”.
Data Analysis
The Statistical Package for the Social Sciences (SPSS) version 26.0 was used for all analyses. Since none of the variables were normally distributed Spearman rank-order correlations were computed with bootstrapping (1000 samples) to investigate the association between the variables.
Regarding the motives for SNS use, an exploratory factor analysis was conducted to reveal any latent variables influencing the covariance of the motives. Initially, the factorability of the motives was examined. The Kaiser-Meyer-Olkin measure of sampling adequacy was .761, well above the recommended value of .6, and Bartlett’s test of sphericity was significant, χ2 (120) = 590.49, p < .001. As such, factor analysis was deemed to be suitable with all 16 motives. Given the non-normality of the motive variables, the robust method of principal axis factor analysis was selected (Costello et al., 2005). Costello and Osborne (2005) recommend that oblique rotations should be selected over orthogonal rotations if the variables are even somewhat correlated (see correlation matrix in Supplemental Table 3). As such, a promax rotation was applied. The criteria for a factor to be retained were an eigenvalue of at least one, and at least three items meeting primary loadings of at least .5 with no secondary loadings above .4 (as per (Park et al., 2016).
Following factor analysis and identification of the factors, factor scores were created for the purposes of using these in multiple regression analyses. Factor scores were computed by calculating the mean of all the items that measured the motives that loaded above .4 on a specific factor[1]. It was decided that it was not necessary to standardize these values, given that all the items were on the same Likert scale.
Factors were treated as outcome variables in a series of hierarchical multiple regression analyses, to investigate how ED severity, readiness to change, self-esteem, and body satisfaction, and total amount of daily SNS use (in minutes) predict the motives driving SNS use. Given previous evidence that younger people are heavier SNS users than older people (Keles et al., 2020), the decision was made to control for age. The assumption of collinearity was met if VIF values were below ten and tolerance values were greater than .1. The assumption of independence of errors was met if Durbin-Watson values were greater than one and less than three.
[1] For example, the factor score for the factor Positive Use was computed by obtaining the mean from all the items that were used to measure enjoyment, relaxation, self-expression, and self-documentation.