ANN ARBOR – UM-ATLAS, a team of University of Michigan faculty and students studying U.S. suicide risk over the lifespan, has won third prize in a recent Driven Data challenge to improve the National Violent Death Reporting System (NVDRS), the nation’s most comprehensive registry of suicide mortality.

The NVDRS is a key resource for researchers and policymakers to study the circumstances of suicides to better understand them and reduce their occurrence. The Centers for Disease Control (CDC) reporting system captures information about violent deaths across the United States that has been abstracted from narratives contained in sources that include death certificates and reports from law enforcement, coroners, medical examiners, and toxicologists, which can shed light on why suicides happen. The NVDRS uses standard variables that are useful to researchers drawing insights from these data to understand suicide, a leading cause of death in the United States for 5-24 year-olds. 

The Youth Mental Health Narratives Challenge

The goal of the Youth Mental Health Narratives Challenge was to improve the quality and coverage of the standard variables in the NVDRS. In the Automated Abstraction track, participants submitted code to ingest NVDRS narratives and predict 23 standard NVDRS variables. In the Novel Variables track, solvers explored narratives in the NVDRS data and identified new standard variables that could advance youth mental health research. Winners were selected by a panel of expert judges on the basis of technical novelty, subject matter insight, methodological rigor, and communication, according to a blog release from Driven Data.

Over 750 participants joined the challenge. 

U-M ATLAS

The UM team winners, placing third in the Novel Variables track with a project focusing on the role of social media use, included Research Center for Group Dynamics (RCGD) affiliates Viktoryia Kalesnikava and Elyse J. Thulin; Lily Johns, Research Coordinator at the University of Michigan School of Public Health, and students Aparna Ananthasubramaniam, Silas Falde, Lily Johns, Jonathan Kertawidjaja, Alejandro A. Rodríguez-Putnam, and Emma Spring. Thulin, a Research Assistant Professor at the U-M Institute for Firearm Injury Prevention, and Ananthasubramaniam, a joint PhD student in the Schools of Information and Social Work, co-led the team. The Atlas Study, a project at the University of Michigan School of Public Health, is headed by RCGD faculty affiliate Briana Mezuk.

“Research links higher use of social media, longer screen times, and harmful online activities (i.e., cyberbullying, coercive sexting) with increased mental health challenges and youth suicide ideation,” the team wrote in their final submission that aimed to identify specific mechanisms, pathways, and online behaviors linked to youth suicide deaths. “Recent rise in youth suicide highlights the urgent need to understand how online experiences contribute to this public health issue.”

While NVDRS has an existing variable related to suicide disclosure on social media, the team found that there were multiple other themes related to social media in law enforcement (LE) and coronor/medical examiner (CME) narratives that can provide critical contextualizing information as to the role social media played relative to an adolescent decedent’s death.

UM-ATLAS used mixed qualitative and computational methods in the project. They used thematic content analysis to develop a codebook based on information present in CME and LE narratives on social media, and then leveraged the benefits of natural language processing algorithms (e.g., large language models) to apply this codebook to thousands of cases, enabling statistical analysis of trends related to social media experiences and adolescent suicide to generate novel insights. The solution code is available on Github.

Winning Solutions

The team won a mid-point bonus in the contest as well as a $5,000 prize for placing third in the contest’s Novel Variables track.

All winning solutions in the contest relied on general-text large language models; in the Novel Variables track, many suggested variables related to modern factors like internet use, social media, and video games. Participants also focused on influences of gender and sexuality; sleep; military experience; religion; belonging, and others.

The “verto” team from Toronto placed first in the track by extracting temporal information to determine a time series of events leading up to a suicide. Digital health researchers on the “HealthHackers” team placed second by identifying common concepts not yet tracked in the registry’s standard variables.

If you or someone you know is struggling with mental health, you can call or text the national 988 Lifeline for advice and resources.

This post was written by Tevah Platt of the Research Center for Group Dynamics, with content from Hannah Moshontz and Katie Wetstone of Data Driven Labs. The Research Center for Group Dynamics at the University of Michigan Institute for Social Research advances the study of human behavior in social contexts.