Uncovering Hidden Risk Profiles of Adolescent Health Behaviors in Lebanon using Latent Class Analysis

Abstract

Background: Adolescence is a critical developmental period during which multiple health risk behaviors often emerge and cluster. Although these behaviors are commonly studied individually, they may co-occur within the same adolescents, reflecting broader behavioral risk patterns that are better captured using person-centered approaches. This study aimed to identify latent classes of co-occurring risk behaviors among Lebanese adolescents and to examine demographic characteristics, protective factors, and emotional well-being indicators as predictors of class membership. The secondary objective was to assess the sensitivity of findings to different missing data handling approaches.

Methods: Secondary data analysis of the 2024 Lebanon Global School-based Student Health Survey was conducted. The survey collected data on 3,745 adolescents attending Grades 7 through 12 in public and private schools in Lebanon. Latent class analysis was conducted using twelve binary behavioral indicators capturing dietary behaviors, physical activity, sedentary lifestyle and screen time, substance use, violence, as well as suicidality. Full information maximum likelihood estimation was used as the primary missing data approach. Models with two to six classes were estimated and compared using various fit statistics (BIC, AIC, Adjusted BIC, and Entropy R-squared). Multinomial logistic regression with separate models adjusted for sex, grade, and school type was used to examine predictors of class membership. A complete-case sensitivity analysis was conducted on the 2,956 participants (78.93%) with complete data.

Results: The best-fitting model was a five-class solution. Based on conditional item probabilities (class characteristics), the classes were labeled: high-risk multi-problem group (6.81%), low-risk (27.01%), lifestyle risk group (46.21%), mixed-risk group (9.98%), and unhealthy diet group (9.99%). Class 3 (lifestyle risk group) was the largest single class, a key difference from international studies where a low-risk class typically predominates. Female sex, parental support, and peer support were significantly protective against Class 1 (high-risk multi-problem group), while frequent loneliness (adjusted RRR=5.294), worry-related sleep disturbance (adjusted RRR=5.237), bullying victimization, and non-Lebanese nationality were associated with higher risk in the same class. School-based drug education was protective against Class 1 (high-risk multi-problem group) membership (adjusted RRR=0.583). Public school attendance was significantly associated with a higher relative risk of belonging to Class 3 (lifestyle risk group) after adjustment (adjusted RRR=1.469, 95% CI: 1.054, 2.046). The complete-case sensitivity analysis converged on the same five-class solution with consistent profiles, though Class 1 (high-risk multi-problem group) was smaller (4.42% vs 6.81%).

Conclusion: Lebanese adolescents surveyed as part of the 2024 GSHS exhibited five distinct behavioral risk profiles shaped by multiple factors. The dominance of a sedentary and screen-based risk profile as the largest class reflects the impact of Lebanon's overlapping crises on adolescent daily lifestyle. Emotional distress, particularly loneliness and worry, emerged as the strongest predictor of high-risk class membership, highlighting the need for integrated behavioral and mental health approaches in school health programs. These findings provide an evidence base for targeting and tailoring adolescent health interventions.