2.2 Procedure
A convenience sample was recruited from February 2021 to May 2022. Informed written consent was obtained from the adolescents and from their parent(s) or legal guardian(s). All participants were informed of the study details, the advantages and the potential risks of the study, and were advised that they were free to participate or not. The study was approved by the Ethic and Scientific committees of the CIUSSS de l’Estrie-CHUS and the Université de Sherbrooke in January 2021.
In the hospitalized group, adolescents were recruited on a specialized adolescent psychiatric ward. Adolescents of the community group were recruited after a visit by the research team to the classrooms (in person or virtually) to explain the study. After receiving the consent forms signed by the student and their parents, participants received a link to complete their questionnaire. Participants had to complete the questionnaire on LimeSurvey© V5.3.5, an online survey software. Participants who completed the questionnaire were enlisted in a draw, with prizes values which complied with the ethics guidelines of the Université de Sherbrooke.
2.3 Measures and analysis
Socio-demographic data such as age and gender were collected. Participants were also questioned on their SM use, including the frequency of use, perceived use of SM (positive, negative or neutral), and the changes in use during the COVID-19 pandemic.
To assess PUSM, we used a French translation of Bergen Social Media Addiction Scale (BSMAS). The 6-item scale is adapted from the Bergen Facebook Addiction Scale(24) and encompasses the six domains of the component addiction model described above, rated on a five-point Likert scale ranging from 1 (very rarely) to 5 (very often). It has been translated into multiple languages(5,25–28). It was previously validated in a cohort of adolescents, with a threshold of 19 or more suggested to classify a youth as having PUSM, based on a latent profile analysis(5). A recent paper suggested a cut-off of 24 to retain an “SM disorder”, based on a clinical sample of adolescents who were diagnosed by certified psychiatrists and a set of clinical criteria based on IGD found in the DSM-5(29). We retained the threshold of 24 in this study to suggest an adolescent had PUSM. French version of BSMAS is presented in table 1.
Table 1: French version of the Bergen Social Media Addiction Scale.
Échelle de dépendance aux médias sociaux de Bergen (BSMAS) – Version française
Directive Les questions qui suivent portent sur votre rapport aux médias sociaux (Facebook, Twitter, Instagram et autres plateformes semblables) et l’utilisation que vous en faites. Pour chacune d’elles, veuillez cocher la réponse qui décrit le mieux votre situation.
The back-translation method of Vallerand transcultural translation(30) was performed by two independent translators. Two experts (V.B. and R.D.R.) compared the original scale with the translated one and made the necessary adjustments. The translated scale was administered to a pilot group of ten French-speaking participants to ensure that all items of the scale were understandable. Internal validity was measured using Cronbach-alpha statistic.
Concurrent validity between BSMAS, PIUQ, ADOSPA and PHQ9 scores was assessed using correlation matrix with Spearman’s rho coefficients, with a level of statistical significance set at a p value of < 0,05. The Problematic Internet Use Questionnaire (PIUQ), an 18-item scale(31) measuring Internet addiction, translated into French(32) and validated in an adolescent cohort(33), was used for concurrent validity. Problematic use of Internet can be understood as an umbrella concept encompassing multiple online behavioural addictions, including PUSM(34). Patient Health Questionnaire (PHQ-9) was used to measure depression level in the sample. It is a nine-item self-administered scale rated on a four-point Likert-scale ranging from 0 (never) to 3 (almost every day) that can be used to define the severity of the depression (minimal to severe)(35). Depression is related to PUSM(14) and hence can be viewed as an indicator of concurrent validity. ADOSPA (adolescents et substances psychoactives) is a French translation of the CRAFFT questionnaire(36). A cut-off score of 2 or more indicates a risk of substance use disorder. It is used here as a marker of substance abuse, which has been associated with PUSM(37).
Part of the sample (n = 101) answered questions regarding their perception of their own use of SM and the impact on their functioning, relationships and behaviours. Items were based on the Deba-Internet Scale(38) and used to document the self-perceived consequences on participants from their PUSM.
For construct validity analysis, a confirmatory factor analysis (CFA) was performed. A series of parameters assessing goodness of fit were obtained. Chi-square test with its degree of freedom (χ2/df), with non-significance at p > 0,05 was used, as it is more sensitive for larger samples(39). Other indices used were the Comparative fit index (CFI) and Tucker-Lewis indices (TLI), with cut-off value of more than 0,9 suggesting a good comparative fit(40). Root mean square error of approximation (RMSEA) with 95% confidence interval (CI) and the standardized root mean square residual (SRMR) are also provided, with a satisfactory fit indicated by a value of 0,08 or less(41). Measurement invariance (MI) across gender was examined. MI for age was not included as the age range was limited in a sample of adolescents. We tested for configural invariance, metric invariance and then scalar invariance. There is a debate as which fit statistics should be used to measured fitness of the MI, but the majority of experts recommend comparing the fit of two nested models by computing the difference between fit indices, with change criteria as follow: ∆ χ2 (having the same restriction as stated above concerning larger sample), ∆ CFI < 0,01, ∆ SRMR < 0,030 (for metric invariance) and < 0,015 (for scalar invariance), and ∆ RMSEA < 0,015(42).
Convergent validity was assessed by computing the average variance extracted (AVE), using a cutoff of 0,5, and with Composite reliability (CR), with a threshold of 0,6. Standard error of measurement was computed, with an acceptable cut-off of < SD/2. Statistical analysis were performed using IBM SPSS V28 and R V.4.05, using lavaan, bluegrafir and semTools packages.