Suicide prevention in Japanese cancer care hospitals: a topic modeling analysis of manuals and workflows

People with cancer in Japan have an elevated risk of suicide, yet little information is available regarding in-hospital suicide prevention manuals and workflows. This study investigated the strategies and frameworks described in manuals and workflows developed by Designated Cancer Care Hospitals.

Using snowball sampling, suicide prevention manuals and workflows for patients with cancer were collected from 14 Designated Cancer Care Hospitals. The textual data were quantitatively analyzed using topic modeling, specifically latent Dirichlet allocation, a text-mining technique. Topics were extracted and labeled through discussions with suicide prevention experts. Based on these topics, we proposed a set of recommended components for suicide prevention manuals and workflows for people with cancer.

In total, 329 paragraphs from the manuals and workflows of 13 hospitals were analyzed, yielding 20 topics. Among these, six topics were commonly addressed across hospitals, covering prevention (e.g., "Types of Suicide Hotspots and Examples of Countermeasures"), intervention (e.g., "Procedures and Collaboration System for Managing Patients with Suicidal Ideation"), and postvention (e.g., "On-Site Response Procedures Upon Discovering a Suicide"). In addition, six topics were described in greater length by certain hospitals, reflecting their unique institutional characteristics (e.g., "Overview of Suicide Prevention Measures").

Commonly shared topics should be prioritized in suicide prevention manuals and workflows for patients with cancer. Conversely, hospital-specific topics may serve as useful references tailored to the distinct characteristics of each hospital. Factors such as hospital size, geographic location, and clinical practice should be considered when determining the content of these manuals and workflows.