Forecasting Harm with Machine Learning and Causal Inference: The Impact of Anti-LGBTQ+ Legislation on Youth Suicidality

Background and Purpose

Hundreds of anti-LGBTQ+ bills have already been introduced across the United States in 2025 alone, continuing a disturbing trend of increasing hostility toward sexual and gender minority (SGM) communities. While considerable research has examined school-based or interpersonal predictors of suicidality among youth, less attention has been paid to how structural-level factors, such as proposed or enacted legislation, impact population-level health. State-level legislative environments, including protective (e.g., anti-discrimination) and discriminatory policies, may play a critical role in shaping the mental health outcomes of young people, regardless of their sexual orientation or gender identity.

This study aimed to examine the extent to which exposure to anti-LGBTQ+ legislation is associated with self-reported suicidality among U.S. high school students. We hypothesized that an increase in discriminatory legislation would be associated with increased suicidality, particularly among sexual minority youth. We also sought to compare the predictive performance of difference-in-differences (DiD) and machine learning (ML) models in forecasting suicidality based on legislative context and individual-level characteristics.

Methods

We used a multi-method analytic strategy drawing on both traditional and computational approaches. Individual-level data were obtained from the Youth Risk Behavior Survey (YRBS) for the years 2017, 2019, 2021, and 2023, representing a sample of 135,846 students across 28 U.S. states. Legislative data were sourced from the American Civil Liberties Union (ACLU), which documents proposed and enacted LGBTQ+ legislation by year, issue, and state.

We merged these sources to assess student suicidality in relation to the number and type of bills introduced in each state. A logistic regression DiD model was used to compare changes in suicidality over time across states with varying levels of legislative exposure. We then trained logistic regression and random forest ML models using legislative variables and individual-level sociodemographic data (e.g., age, gender, sexual identity, race/ethnicity) as predictors.

Results

As discriminatory bills increased, suicidality also increased by 20% among all students (OR = 1.20, 95% CI: 1.04–1.41, p = 0.010). The effect was especially pronounced among sexual minority students (OR = 1.33, 95% CI: 1.08–1.40, p = 0.043), suggesting that these students are particularly vulnerable to shifts in the legislative climate. The random forest and logistic regression ML models performed comparably in predicting suicidality, with AUC scores of 0.728 and 0.727, respectively.

Conclusions and Implications

This study demonstrates that the proposal of anti-LGBTQ+ bills, regardless of whether they are passed into law, poses a significant threat to the mental health of both heterosexual and sexual minority youth. These findings highlight the need for policymakers, educators, and mental health professionals to consider the structural environments for youth. Legislative threats not only shape public discourse but may also have measurable psychological consequences.

Our results underscore the urgency of integrating legislative monitoring into mental health risk assessments and public health systems. Understanding these pathways is essential to crafting policy environments that promote, not endanger, the well-being of all youth.