A Latent Class Analysis of Unhealthy Lifestyle Behaviors and Suicide Among Adolescents

Background & Purpose: Suicide has emerged as a significant public health concern, ranking as the second leading cause of death among adolescents from 15 to 19 year olds. Previous studies have focused on the identification of psychological, psychiatric, demographic, and environmental risk factors of adolescent suicide, however, relatively little is known about health-related lifestyle behaviors preceding or concurrent with suicide. Although some existing studies examined the independent associations of diet, sleep, exercise, media use, and weight status with adolescent suicidal behavior, very few analyzed a comprehensive range of lifestyle patterns simultaneously. The extent to which different lifestyles can be similarly clustered, and impact suicide attempts has not yet been reported, and can have important implications for practice and prevention. This study aims to identify and characterize the typologies of lifestyle behaviors related to suicidal behaviors, and explore the demographic profiles across the identified classes among a nationally representative sample of high school students.

Methods: Data were obtained from the 2015 National Youth Risk Behavior Survey (YRBS). Thirteen risky lifestyle behaviors are included based on previous literature: reduced sleep, overweight, underweight, sedentary behavior, high use of Internet/TV, no sports team participation, no physical education (PE) class, not eating breakfast, and unhealthy dietary patterns including limited fruit, vegetable, water intake, and excessive soda drinking. Suicide measures included suicidal ideation, plan, and attempt. All variables were dichotomized. Latent class analysis was performed without any a priori assumption to identify latent subgroups with varying combined levels of risky lifestyle behaviors among 15,617 adolescents. A 3-step approach to modeling was utilized so that the measurement model remained fixed when adding the covariates and outcome variables. Final class model was determinate after estimating two to four latent class models. A multivariate multinomial logistic regression model was developed to explore the relationship between latent classes, demographic profiles (age, gender, race/ethnicity), and suicidal behaviors.

Results: Analyses identified a three-class model with best fit: class one (“low risk”, n= 8,474; 54%) with no or very low frequency of risk behaviors; class two (“moderate risk, n= 5,146; 33%) with low probability of risk behaviors but highest in no fruit/vegetable intake and PE class taking; and class three (“high risk”, n= 1,997; 13%) included adolescents who scored high on all risk behaviors except fruit/vegetable and PE class indicators. Compared with the “low risk” class, the “high risk” class contains older (OR=1.059), male (OR=1.644), African American (OR=1.644), and Hispanic/Latinos individuals (OR=1.621). Adolescents in “high risk” class are more likely to have suicide attempt (OR=1.613) compared to the “low risk” class. There is no statistically significant association of suicide and lifestyle between “low” and “moderate” class.

Conclusion & Implications: Suicide is one of the most perplexing and devastating behaviors. Identifying non-psychiatric risks of suicide provides additional information and promising step for early prediction and treatment. This study indicates that youth engaging in unhealthy lifestyle behaviors represent a unique group at risk for suicidality. The findings also highlight the gender and race/ethnicity disparity with regards to unhealthy lifestyles, and their association with suicide attempt.