Semantic network analysis of cognitive-affective patterns related to suicide risk and end-of-life attitudes in cancer

Suicide, euthanasia, and physician-assisted suicide (PAS) represent significant clinical challenges in oncology and palliative care. Although these topics are conceptually related, they are often studied separately. Previous studies have relied on structured, close-ended measures, but none have applied semantic network analysis (SNA) to patient-generated language in this context.

The study seeks to explore patterns of lexical association related to suicide risk and attitudes toward end-of-life interventions, applying SNA to Sentence Completion Test (SCT) responses. Networks were compared by suicide risk status, approval or disapproval of active euthanasia, and PAS.

A total of 298 patients with cancer completed a seven-item SCT covering five domains: self, relationships, future, distress, and cancer appraisal. Group-specific undirected semantic networks were constructed. Global network metrics and node-level centrality were computed. Group differences were tested via permutation procedures and conceptual similarity was assessed using Jaccard similarity coefficients of top-ranking central terms.

The suicide risk networks were more narrowly illness-related terms, with words such as cancer, suffering, and death occupying more central positions. Similarly, networks of participants approving euthanasia and PAS were organized around illness and mortality-related terms, whereas disapproving groups showed more diverse and distributed patterns, including positively valenced terms such as appreciation and hope. Although there was moderate-to-high overlap in key terms across groups, each group showed distinct patterns in how these terms were organized.

Attending to emotionally salient language may provide insights into patients' experiences and could help inform psychosocial support and suicide prevention efforts.