EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Exploring GPT-based models for depression detection in adolescents using naturalistic speech: A proof-of-concept study

Anna Viduani, Leonardo Z Ferreira, Claudia Buchweitz, Jader Piccin, Helen L Fisher, Alastair Van Heerden, Brandon A Kohrt, Valeria Mondelli, Christian Kieling, Ricardo Matsumura Araújo

Journal of Affective Disorders · 2026-10 · 410 · 121956

DOI: 10.1016/j.jad.2026.121956 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of Affective Disorders, 2026-10. Verfasst von Anna Viduani und 9 weiteren Autor:innen. Beteiligte Einrichtung: Prodia - Child & Adolescent Depression Program, Hospital de Clínicas de Porto Alegre (HCPA), Porto Alegre, Brazil; Graduate Program in Psychiatry and…. Hauptthemen laut MEDLINE: Large Language Models, Major Depressive Disorder, Speech. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 40 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Conselho Nacional de Desenvolvimento Científico e Tecnológico, Coordenação de Aperfeiçoamento de Pessoal de Nível Superior, MQ Transforming Mental Health.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

This study investigates whether GPT-4o, a Large Language Model (LLM), can identify depression from daily, naturalistic non-symptom-specific speech transcripts; how its performance varies with input volume; and how it compares to other LLMs. Thirty Brazilian adolescents (aged 16-19) completed a 14-day remote data collection using IDEABot, a WhatsApp-based chatbot. Half of the participants (n = 15, 8 girls) had a confirmed Major Depressive Disorder (MDD) diagnosis established by a child and adolescent psychiatrist using gold-standard procedures. GPT-4o was prompted to classify participants as depressed or non-depressed based on daily transcribed audio messages in Brazilian Portuguese. Performance was evaluated across varying input lengths (number of days), and a secondary analysis compared GPT-4o with 12 alternative GPT-based models. GPT-4o achieved an overall accuracy of 86.6% (95% CI [70.3%, 95.6%]), with 100% sensitivity and 73.3% specificity (95% CI [55.6%, 86.2%]). Classification remained stable across days, and increasing input length yielded only modest improvements. No statistically significant differences were observed between models, although GPT-4o showed numerically higher performance than newer and larger models. As a proof-of-concept study using a purposively selected sample representing opposite ends of the symptom spectrum, these findings are preliminary. Nonetheless, they suggest that LLMs may support depression detection from naturalistic conversational text, extending prior work based on structured assessments to more ecologically valid contexts. This study highlights both the promise and current limitations of AI-based approaches for mental health screening.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Large Language ModelsMajor Depressive DisorderSpeechAdolescentFemaleHumansMaleYoung AdultBrazilProof of Concept StudySensitivity and SpecificityGenerative Artificial Intelligence

Stichwörter der Autor:innen

Depression screeningLarge Language ModelsNatural language processing

Autor:innen und Institutionen

  • Prodia - Child & Adolescent Depression Program, Hospital de Clínicas de Porto Alegre (HCPA), Porto Alegre, Brazil; Graduate Program in Psychiatry and Behavioral Sciences, Universidade Federal do Rio Grande do Sul (UFRGS…
  • Prodia - Child & Adolescent Depression Program, Hospital de Clínicas de Porto Alegre (HCPA), Porto Alegre, Brazil. Electronic address: lferreira@equidade.org.br.
  • Social, Genetic & Developmental Psychiatry Centre, Institute of Psychiatry, Psychology & Neuroscience, King's College London, London, United Kingdom; Economic and Social Research Council (ESRC) Centre for Society and Me…
  • Wits Health Consortium, University of Witwatersrand, South Africa. Electronic address: Alastair.vanheerden@wits.ac.za.
  • Division of Global Mental Health, Department of Psychiatry, School of Medicine and Health Sciences, The George Washington University, Washington, DC, United States. Electronic address: bkhort@email.gwu.edu.
  • Department of Psychological Medicine, Institute of Psychiatry, Psychology & Neuroscience, King's College London, London, United Kingdom; National Institute of Health and Care Research (NIHR) Maudsley Biomedical Research…

Förderung

  • Conselho Nacional de Desenvolvimento Científico e Tecnológico
  • Coordenação de Aperfeiçoamento de Pessoal de Nível Superior
  • MQ Transforming Mental Health
  • South London and Maudsley NHS Foundation Trust
  • UK Research and Innovation Medical Research Council
  • Economic and Social Research Council
  • Royal Academy of Engineering
  • National Institute for Health and Care Research

Zitieren

APA

Viduani, A., Ferreira, L. Z., Buchweitz, C., Piccin, J., Fisher, H. L., Van Heerden, A., Kohrt, B. A., Mondelli, V., Kieling, C., & Araújo, R. M. (2026). Exploring GPT-based models for depression detection in adolescents using naturalistic speech: A proof-of-concept study. Journal of Affective Disorders, 410, 121956. https://doi.org/10.1016/j.jad.2026.121956

Vancouver

Viduani A, Ferreira LZ, Buchweitz C, Piccin J, Fisher HL, Van Heerden A, et al. Exploring GPT-based models for depression detection in adolescents using naturalistic speech: A proof-of-concept study. Journal of Affective Disorders. 2026;410:121956. doi:10.1016/j.jad.2026.121956

BibTeX
@article{Viduani2026,
  author = {Viduani, Anna and Ferreira, Leonardo Z and Buchweitz, Claudia and Piccin, Jader and Fisher, Helen L and Van Heerden, Alastair and Kohrt, Brandon A and Mondelli, Valeria and Kieling, Christian and Araújo, Ricardo Matsumura},
  title = {Exploring GPT-based models for depression detection in adolescents using naturalistic speech: A proof-of-concept study},
  journal = {Journal of Affective Disorders},
  year = {2026},
  volume = {410},
  pages = {121956},
  doi = {10.1016/j.jad.2026.121956}
}
Kostenlos starten →