Population-Level Profiling of DSM-5 Depressive Symptoms Among Self-Reported ADHD and ASD Users on Twitter: An Exploratory Study Using Advanced NLP and Statistical Analysis

2026-07-05 · Source: cs.CL updates on arXiv.org · Field: Science & Research — Health & Medical Research, Social Sciences & Behavioral Studies · Depth: Advanced, extended

Summary

A study analyzed 1,282,437 tweets from 792 self-reported users (622 ADHD; 170 ASD) to profile population-level differences in DSM-5 depressive symptom expression. Researchers developed a two-stage natural language processing pipeline, combining a zero-shot NLI pre-filter with MentalRoBERTa fine-tuned on ReDSM5 for multi-label symptom classification. This classifier achieved a macro-F1 of 0.901, significantly outperforming previous benchmarks. Using L1-penalised logistic regression and robustness testing across five filtering thresholds (0.45–0.65), the study found stable, modest discrimination (ROC-AUC 0.645–0.653). Six symptoms robustly distinguished groups: cognitive issues, sleep issues, appetite change, and fatigue consistently leaned toward ADHD, while suicidal ideation and anhedonia consistently leaned toward ASD. A largely shared symptom co-occurrence structure was observed, with no robust disorder-specific differences identified.

Key takeaway

For NLP Engineers or Research Scientists developing digital phenotyping tools, this study highlights that population-level depressive symptom expression on social media differs between ADHD and ASD. You should consider these distinct linguistic profiles when designing models for mental health assessment. While not for individual diagnosis, this approach offers a reproducible signal for corroborating clinical hypotheses.

Key insights

Twitter data reveals distinct population-level depressive symptom expression patterns between self-reported ADHD and ASD users, stable across analysis thresholds.

Principles

• Population-level depressive language differs between ADHD and ASD.
• Multi-stage NLP pipelines enhance social media mental health analysis.
• Robust findings require cross-threshold and bootstrap stability.

Method

A two-stage NLP pipeline: zero-shot NLI for depressive relevance pre-filtering, then MentalRoBERTa fine-tuned on ReDSM5 for multi-label DSM-5 symptom classification, followed by L1-penalised logistic regression with bootstrap stability and cross-threshold consistency.

In practice

• Fine-tune MentalRoBERTa for multi-label symptom classification.
• Employ zero-shot NLI for high-recall content pre-filtering.
• Validate NLP findings with cross-threshold and bootstrap stability.

Topics

• MentalRoBERTa
• Digital Phenotyping
• ADHD
• Autism Spectrum Disorder
• DSM-5 Depression
• Twitter Data Analysis

Code references

• rizwan2phd/dsm5-adhd-asd-twitter

Best for: AI Scientist, Research Scientist, NLP Engineer

Related on AIssential

• Speech as a biomarker for supported diagnosis of major depressive disorder using self-supervised representations — Self-supervised speech representations offer robust, objective biomarkers for MDD diagnosis.
• RSPC: A Benchmark for Modeling Stress and Psychiatric Conditions in Digitally Mediated Relationships using Psychiatrist Annotations — Modeling mental health in NLP benefits from relational context, moving beyond individual-centric approaches.
• Before the Labels: How Dataset Construction Shapes Suicidality Detection in Clinical Text — EHR-based suicidality datasets encode specific operationalizations of suicidality, shaped by construction choices, not objective ground truth.
• When PCOS Meets Eating Disorders: An Explainable AI Approach to Detecting the Hidden Triple Burden — Explainable AI can detect co-occurring health conditions like PCOS and eating disorders from social media data.
• Transition-Based Digital Twin Modelling for Alzheimer's Disease under Sparse Longitudinal Data — Local transition modeling offers a data-efficient and robust strategy for personalized disease forecasting in neurodegenerative disorders.
• Reading between the Lines: Leveraging Large Language Models for Global Dementia and Depression Assessment from Clinical Interviews — LLMs can effectively assess neuropsychiatric disorder severity from speech, enabling automated screening.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.CL updates on arXiv.org.