Back

A National Evaluation of GoGuardian Beacon and Youth Suicide Trends

Zhu, X.; Wood, K.; Ahrens, R.; Damani, R.; Rawat, M.; Batorsky, B.; Wilcox, H. C.

2025-04-28 public and global health
10.1101/2025.04.22.25326214 medRxiv
Show abstract

Over the time of the COVID-19 pandemic, many school systems started to utilize educational software to identify students actively planning suicide and other acts of violence. This study examines associations between county-level youth suicide rates and the implementation of GoGuardian Beacon, a school-based software using machine learning methods for identifying students at risk for suicide. Using difference-in-differences and event study methods, we analyzed 2018-2022 suicide data comparing 70 counties with sustained Beacon implementation to 1,215 matched comparison counties that never implemented Beacon. In our primary analysis, counties that maintained consistent Beacon use had 24.4% lower youth suicide rates during 2021-2022 (p < 0.05). In sensitivity analyses defining implementation based on initial adoption regardless of subsequent use, the association was attenuated and not statistically significant. Taken together, these findings indicate that counties with sustained use of Beacon had lower youth suicide rates in our primary analyses, while also highlighting the possibility that broader contextual factors (e.g., local mental health infrastructure and school system characteristics) contribute to the observed differences. Randomized trials with prospective follow-up, more information on school and community resources, and quality of Beacon response pathways after identification are needed to understand the effect of Beacon and clarify the independent contribution of digital monitoring tools within comprehensive youth suicide prevention strategies.

Matching journals

The top 5 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.