{
  "abstract": "Suicide risk assessment in clinical conversations relies on the sequential structure of therapist-patient dialogue, yet formal models of how suicide-related content emerges, transitions, and reciprocates across conversational turns are lacking. We applied network analysis to C-SSRS-coded utterances from psychotherapy sessions to characterize the topology and dynamics of suicide-related clinical communication. Transcripts of 110 psychotherapy sessions from the SPEAK-SAFE study were classified for suicidality relevance using the Qwen3-32B large language model with an adapted Columbia-Suicide Severity Rating Scale (C-SSRS) schema. From the suicidality-relevant utterances, we constructed a weighted directed category transition network. Centrality, reciprocity, clustering, assortativity, and path length metrics were computed. Independent human expert ratings for a session-level subsample of 30 sessions from this cohort showed substantial agreement with the LLM classifications (Spearman {rho} = 0.885, ICC[2,1] = 0.708). The network was dominated by therapist inquiry. Reciprocity was high, and closeness centrality decreased monotonically with C-SSRS severity. Active suicidal ideation functioned as a bridge node from which escalation pathways to method, plan, and attempt categories originated. The network exhibited a hub-and-spoke topology with small-world characteristics. Suicide-related clinical dialogue is structured as a therapist-centered hub-and-spoke network with high reciprocity and a consistent severity-peripherality gradient. The extension of network theory from intra-individual symptom structures to inter-individual communication dynamics represents a methodological innovation demonstrated here as a proof of concept, with potential applications to clinical training, automated decision support, and multimodal risk assessment. clinical interviews clinical interviews\n\nAuthor Summaryclinical interviews clinical interviewsWhen clinicians talk with patients about suicide risk, the conversation itself follows a pattern: the therapist asks, the patient answers, and the topic shifts from one aspect of risk to another. We wanted to know whether this back-and-forth has a describable structure, in the same way that networks of interacting symptoms have been used to describe mental disorders. Using a large language model, we automatically labeled which parts of 110 real psychotherapy conversations touched on suicidal thoughts, plans, or behaviors, and then built a network showing how the conversation moved between these topics. We found that the therapists questions sat at the center of this network, that the conversation frequently looped back to earlier topics, and that mentions of active suicidal thoughts acted as a gateway to discussion of more severe risk, such as specific plans or past attempts. This proof-of-concept approach offers a new way to visualize and quantify how suicide risk assessments unfold in real clinical dialogue, with potential future applications in clinician training and automated support tools.",
  "abstract_shown": true,
  "affiliations": [
    "Goethe-Universität Frankfurt am Main: Goethe-Universitat Frankfurt am Main"
  ],
  "article_type": "PUBLISHAHEADOFPRINT",
  "arxiv": null,
  "author_total": 8,
  "authors": [
    {
      "affs": [
        0
      ],
      "name": "O. Grimm"
    },
    {
      "name": "C. Landau"
    },
    {
      "name": "S. Arampatzi"
    },
    {
      "name": "A. Goel"
    },
    {
      "name": "E. Owtscharow"
    },
    {
      "name": "H. Arnaout"
    },
    {
      "name": "I. Gurevych"
    },
    {
      "name": "A. Reif"
    }
  ],
  "bibtex": "https://research.draconitic.com/paper/bzs2qqm53rrm1.bib",
  "cited_by": 0,
  "doi": "10.64898/2026.09.27.26364136",
  "fulltext_url": "https://www.medrxiv.org/content/10.64898/2026.09.27.26364136v1.full",
  "id": "bzs2qqm53rrm1",
  "issns": [],
  "landing_url": "https://www.medrxiv.org/content/10.64898/2026.09.27.26364136v1",
  "language": null,
  "licence": "CC-BY",
  "licence_label": "CC BY",
  "licence_source": "article",
  "licence_url": "https://creativecommons.org/licenses/by/4.0/",
  "metadata_licence": null,
  "native_subjects": [
    [
      "medrxiv",
      "psychiatry and clinical psychology"
    ],
    [
      "openalex",
      "3203|Suicide and Self-Harm Studies"
    ]
  ],
  "openalex": "W7219061947",
  "pdf_url": "https://www.medrxiv.org/content/10.64898/2026.09.27.26364136v1.full.pdf",
  "pmcid": null,
  "pmid": null,
  "published": "2026-09-27",
  "publisher": null,
  "ris": "https://research.draconitic.com/paper/bzs2qqm53rrm1.ris",
  "sources": [
    {
      "first_seen": "2026-10-30T12:19:39Z",
      "native_id": "10.64898/2026.09.27.26364136",
      "source": "medrxiv",
      "url": "https://www.medrxiv.org/content/10.64898/2026.09.27.26364136v1"
    }
  ],
  "subjects": [
    {
      "field": "medicine",
      "origin": "medrxiv:psychiatry and clinical psychology",
      "subfield": "psychiatry"
    }
  ],
  "title": "From Symptom Networks to Conversation Networks: A Cross-Sectional Study Mapping the Topology of Suicide-Related Clinical Dialogue",
  "title_text": "From Symptom Networks to Conversation Networks: A Cross-Sectional Study Mapping the Topology of Suicide-Related Clinical Dialogue",
  "type": "preprint",
  "updated_at": "2026-10-30T12:19:39Z",
  "url": "https://research.draconitic.com/paper/bzs2qqm53rrm1",
  "venue": "medRxiv",
  "versions": []
}