Journal Article

Using Bayesian networks to explore differences in the relationships of factors related to mental health for LGBTQ+ and non-LGBTQ+ adolescents in a UK cohort study

Brennan, Alan, Weich, Scott, Saidi, Yekta

2024

Overview

LGBTQ+ adolescents have higher rates of mental health problems and reduced wellbeing compared to non-LGBTQ+ adolescents. We aimed to understand the possible differences in causal mechanisms behind these disparities using a representative UK cohort. The Millennium Cohort Study (MCS) is a UK cohort study following children born in 2000–02 in England, Scotland, Wales and Northern Ireland. We analysed data from this study split into LGBTQ+ and non-LGBTQ+ groups to see if and how the relationship between determinants of mental health outcomes differ for the two groups. Outcome variables were self-harm and suicide attempts from responses provided at age 17 years. The determinants included demographics, and variables describing experiences, behaviours, and social environments provided both at ages 14 and 17 years, to explore the interaction of variables between timepoints. The analysis utilised a Bayesian network (BN) approach with Directed Acyclic Graphs (DAGs) to estimate the likely direction of conditional dependencies between the variables and assess possible causal mechanisms. Hypothesis testing was performed to determine if the DAGs for the two groups were statistically significantly different. At age 17 years, 9612 of 10 757 cohort members provided responses about their sexuality and gender identity. 2645 of 9612 were identified as LGBTQ+ and 6967 of 9612 were identified as non-LGBTQ+. Compared to the non-LGBTQ+ group, the LGBTQ+ group had higher rates for the outcome variables (42·50% vs 17·31% for self-harm, 16·07% vs 4·37% for suicide attempts), and higher rates of victimisation (68·50% vs 52·64% at age 17 years) and lower rates of high self-esteem (52·27 % vs 75·07% at age 17 years), both identified as key variables for both groups. The structures for the two groups were found to be significantly different (p<0·001), and inspection of the DAGs showed some key differences such as suicide attempts being directly dependent on substance use and peer problems for the LGBTQ+ group only. Relationships between determinants of self-harm and suicide differ for LGBTQ+ and non-LGBTQ+ adolescents, this must be accounted for in the design of interventions and policies to reduce disparities. Wellcome Trust [Grant number 108903/B/15/Z].

• Publisher: Elsevier Ltd,Elsevier Limited
• Subject: Adolescents / Age / Bayesian analysis / Child & adolescent mental health / Cohort analysis / Demographic variables / Demographics / Gender aspects / Graph theory / Health problems / Hypothesis testing / Mental disorders / Mental health / Self destructive behavior / Self-injury / Sexuality / Social behavior / Social environment / Substance use / Suicide / Suicides & suicide attempts / Teenagers