Adolescent Mental Health and the Choking Game

Data Sources and Study Design

This search was based on the merger of 2 preexistent databases stemming from 2 cross-sectional studies conducted respectively in 2009 and 2013 (Supplemental Tables 5–7). The data were collected from French middle school students aged 9 to 16 years (mean age: 13.0; SD: 1.3) and distributed within 13 randomly selected public establishments located in different regions in France. After the 2009 (n = 746) and 2013 (n = 1025) data sets were merged by retaining only the youth who responded to the CG question, the final sample consisted of 1771 participants. Their demographic characteristics are displayed in Table 1.

It should be noted that the 2 studies were based exactly on the same research protocol (eg, same variables, same measures, same recruitment method, etc). They were implemented independently to obtain 2 research sites from 2 different regions. The time lapse between the 2 studies is only due to the fact that the ethical agreements took more time to obtain than initially foreseen for the second study. At last, it should be specified that there were no particular trends, changes, or events between the 2 surveys related to awareness or efforts to mitigate the CG.

Ethics approvals were obtained from the French ethical research committee (Comité de Protection des Personnes Ile de France IV) and from the French Commission on Data Privacy and Public Liberties. The directors of each selected middle school also approved the study procedures. The parents and/or guardians of students provided written informed consent before data collection. After the presentation of the objectives of the study, the assessment instruments were collectively administered and collected in the classroom by a doctoral student. All questionnaires were anonymous, so participants were assured that their responses were confidential. No compensation was offered to participate in the study.

Measures

The dependent variable (lifetime participation in the CG) was assessed by the following question: “Have you ever participated in the Choking Game?” Responses were limited to “yes” or “no,” and respondents who answered yes were defined as having a history of CG participation. Thus, by “participation” or “have participated,” we mean having oneself participated at least once in the CG during one’s lifetime, whether alone or with other people. Those who responded negatively formed a control group of adolescents who were not involved in any form of choking activities. Before the question was given, participants were read the following description of the CG (known as “jeu d’asphyxie” or “jeu du foulard” in France): “The Choking Game, also called Blackout, the Dream Game, the Pass-out Game, or The Fainting Game, is an activity in which self-strangulation or strangulation by another person is used to obtain a ‘high’ by restricting blood and oxygen flow to the brain.”

Demographic variables were assessed with questions used to collect information concerning sex (male or female), age, family characteristics (living arrangement and siblings), and schooling (grade level, grade repetition, and geographic location of schools).

Depressive symptoms were investigated with the French version of the Children’s Depression Inventory (CDI).^21,22 The CDI is a 27-item self-report measure originally designed to assess cognitive, behavioral, and affective symptoms of depression in children and adolescents. Each item consists of 3 statements that represent different levels of severity of a specific symptom of depression, and the child is required to choose 1 statement that best describes him or her. Scores range from 0 to 54, with higher scores indicating more depressive symptoms. The CDI is the most widely used measure of childhood depressed affect²³ and has revealed sound psychometric properties, including internal consistency, test-retest reliability, and discriminant validity.^21,24

Risk-taking behaviors and substance use were assessed with a self-report measure composed of 8 yes-or-no items used to assess current practice of high-risk sports activities; current risk-taking in a 2-wheeled vehicle; lifetime history of using alcohol, tobacco, and marijuana; and current consumption of alcohol, tobacco, and marijuana. Response options were dichotomous (eg, 0 = never used, and 1 = ever used). Thus, for each item, participants were classified into 2 categories (eg, no history of tobacco use versus lifetime history of tobacco use; no current consumption of alcohol versus current consumption of alcohol).

To measure the conduct problem symptoms of the teenagers, we constructed a self-report questionnaire on the basis of the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, Text Revision criteria for conduct disorder²⁵ (unchanged in the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition²⁶). In the current study, not all the criteria of the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, Text Revision (N = 15) were held, according to the wishes of the school inspectorate. Three deleted criteria were the following: “has forced someone into sexual activity,” “has deliberately engaged in setting fire with the intention of causing serious damage,” and “has broken into someone else’s house, building, or car.” Finally, our measure of conduct disorder symptoms consisted of a 12-item self-reported scale that included 7 questions referring to aggressive symptoms and 5 questions used to assess the nonaggressive but rule-breaking form of antisocial behavior. The measurement of these 2 distinct but correlated dimensions made it possible to consider heterogeneity within antisocial behaviors.²⁷ The timescale for all violent and antisocial behavior questions referred to the past 12 months, and respondents were asked to report responses limited to yes or no. The total score, ranging from 0 to 12, represented the number of conduct problem symptoms that were present. In the current study, Cronbach’s α coefficient for this self-administered questionnaire was 0.72.

Statistical Analysis

First, we categorized the sample into 2 groups: (1) youth who had reported participating in the CG versus (2) those who had never participated in it. Then, bivariate logistic regressions were conducted to examine the possible association between CG participation and (1) certain demographic characteristics, (2) conduct disorder symptoms, (3) depressive symptoms, (4) risk-taking behaviors (high-risk sports activities and 2-wheeled vehicle risk), and (5) substance use (experimentation and current consumption of tobacco, alcohol, or marijuana). The logistic regression models were fitted to each of the variables independently. Demographic and health risk-factor variables revealing a significant bivariate association with CG participation were then used for the multivariate logistic regression, controlling for key and potentially confounding demographic variables (sex, age, grade repetition, and geographic location [urban or rural]) to better predict CG participation. Odds ratios and 95% confidence intervals (CIs) are presented to reflect association strength. All statistical analyses were conducted with R software (R Core Team, Vienna, Austria), and statistical significance levels were fixed at P < .001.