Icon Legend

This session is not in your schedule.

This session is in your schedule. Click again to remove it.

Presentation Icons

Recorded Session

Ticketed Session

383 Views

ViewAttendees 7

Symposia

Suicide and Self-Injury

Advancing Suicide EMA Research: Diverse Research Questions Call For Diverse Analytic Approaches

4 - (SYM 7) Linguistic Signals of Acute Suicidal Ideation in Major Depressive Disorder: NLP Analysis Reveals Contextual Sentiment Patterns in Ecological Momentary Assessments

Friday, November 21, 2025

4:09 PM - 4:30 PM CST

Location: Celestin B, Level 3

Keywords: Suicide, Technology / Mobile Health, Methods
Recommended Readings: Sedano-Capdevila, A., Porras-Segovia, A., Bello, H. J., Baca-García, E., & Barrigon, M. L. (2021). Use of ecological momentary assessment to study suicidal thoughts and behavior: a systematic review. Current Psychiatry Reports, 23(7), 41. https://doi.org/10.1007/s11920-021-01255-7, Nock, M. K., Kleiman, E. M., Abraham, M., Bentley, K. H., Brent, D. A., Buonopane, R. J., ... & Pearson, J. L. (2021). Consensus statement on ethical & safety practices for conducting digital monitoring studies with people at risk of suicide and related behaviors. Psychiatric Research and Clinical Practice, 3(2), 57-66. https://doi.org/10.1176/appi.prcp.20200029, Kleiman, E. M., Glenn, C. R., & Liu, R. T. (2019). Real-Time Monitoring of Suicide Risk among Adolescents: Potential Barriers, Possible Solutions, and Future Directions. Journal of Clinical Child & Adolescent Psychology, 48(6), 934–946. https://doi.org/10.1080/15374416.2019.1666400, Shiffman, S., Stone, A. A., & Hufford, M. R. (2008). Ecological momentary assessment. Annual Review of Clinical Psychology, 4, 1-32. https://doi.org/10.1146/annurev.clinpsy.3.022806.091415,

Speaker(s)

Nicholas C. Jacobson, Ph.D.

Associate Professor
Dartmouth College
Lebanon, NH, United States

Co-author(s)

DL

Damien Lekkas, PhD (he/him/his)

Research Assistant Professor
Northwestern University
Lebanon, NH, United States
Amanda C. Collins, Ph.D. (she/her/hers)

Post-doctoral Fellow
Center for Technology and Behavioral Health, Geisel School of Medicine, Dartmouth College, Lebanon, NH, United States; Department of Biomedical Data Science, Geisel School of Medicine, Dartmouth College, Lebanon, NH, United States
Hanover, NH, United States
Michael V. Heinz, M.D.

Assistant Professor of Psychiatry
Center for Technology and Behavioral Health, Geisel School of Medicine, Dartmouth College, Lebanon, NH, United States; Department of Psychiatry, Geisel School of Medicine, Dartmouth College, Hanover, NH, United States
Hanover, NH, United States
TG

Tess Z. Griffin, Ph.D.

Research Coordinator
Center for Technology and Behavioral Health, Geisel School of Medicine, Dartmouth College, Lebanon, NH, United States
Hanover, NH, United States
AP

Arvind Pillai, M.S.

Graduate Student
Department of Computer Science, Dartmouth College, Hanover, NH, United States
Hanover, NH, United States
SN

Subigya Nepal, Ph.D.

Post-doctoral Fellow
Stanford Institute for Human-Centered-AI, Stanford University, Palo Alto, CA, United States
Stanford, CA, United States
DM

Daniel M. Mackin, Ph.D.

Assistant Professor
Dartmouth-Hitchcock Medical Center and Dartmouth College
Lebanon, NH, United States
AC

Andrew Campbell, PhD (he/him/his)

Professor
Dartmouth College
Lebanon, NH, United States

Abstract Body

Background: Suicidal ideation (SI) fluctuates significantly over hours, yet most research examines it cross-sectionally or over long intervals. While linguistic markers of SI have been identified, how these markers vary across different acute SI trajectories remains unexplored. Understanding how dispositional characteristics (hostility, verbosity, emotional expression) manifest differently across SI patterns is critical for developing sensitive, temporally-aware risk detection models.

Methods: We analyzed diary entries from 268 participants with major depressive disorder who completed EMAs three times daily over 90 days. Each EMA included SI severity ratings (0-100) and an optional diary entry. Using difference scores between consecutive EMAs, we defined a probability of acute change threshold (≥12 units) to classify SI into eleven distinct trajectory patterns (e.g., Apex, Trough, Escalation, Stability-Low/Medium/High) based on changes across three consecutive timepoints. The SEANCE tool quantified linguistic features across eight lexica for 5,938 diary entries. Kruskal-Wallis tests with Benjamini-Hochberg corrections controlled for Type I error.

Results: We identified 131 statistically significant pairwise differences between SI trajectory groups, implicating 31 NLP features. Results confirmed established markers of heightened SI (personal pronouns, passivity, negative valence) while revealing contextual patterns in verbosity, hostility, anger, and pleasantness. For example, increased hostility signaled transient peaks in SI (Apex), while consistently severe SI states showed lower hostility. Similarly, emotional expression patterns differed based on whether heightened SI was short-lived or prolonged.

Discussion: This work demonstrates that linguistic signals of SI vary based on immediate past and future trajectories. Our findings support the importance of dense temporal sampling when developing NLP-based SI detection tools. Future risk assessment approaches should consider not just current SI severity but its trajectory, which may be detectable through linguistic patterns.