557: Predicting Suicidal Behavior in People with Fibromyalgia: A Machine Learning Approach

Topic: Education, Societal Impact, and Others
Subtopic: Artificial Intelligence/Machine Learning

Background and Aims: Chronic pain is a well-established independent risk factor for suicidality, and psychological comorbidities are highly prevalent in this population and also linked to an increased risk of suicidal behavior. Individuals with fibromyalgia have a risk of death by suicide up to 10.5 times higher than the general population and 3.3 times higher than those with other chronic pain types. This study aimed to develop and validate machine learning (ML) models to predict suicidal behavior (ideation, planning and attempts) among individuals with fibromyalgia.

Methods: We conducted a national web-based survey in Brazil including adults with physician-confirmed fibromyalgia. Suicidal behavior was assessed by the SUPRE-MISS questionnaire. Sociodemographic, clinical, psychosocial, and functional variables were measured using validated instruments. LASSO, Random Forest (RF), and XGBoost models were trained with repeated cross-validation and evaluated on an independent 30% test set. We assessed model performance using Area Under the Curve (AUC), calibration, and standard classification metrics, and evaluated feature importance through permutation-based methods.

Results: Among 1,848 participants, 47.9% reported lifetime suicidal ideation, 26.5% suicide planning, and 18.5% suicide attempts. Model performance varied across outcomes, with RF performing better for ideation (AUC=0.783, Positive Predictive Value [PPV]=0.687) and attempts (AUC=0.759, PPV=0.675), while LASSO performed best for planning (AUC=0.759, PPV=0.675). Key predictors across models included depression severity, psychosocial factors, sleep disturbances, cognitive and functional difficulties, substance use, and contextual variables including age and relationship duration.

Conclusions: Suicidal behavior in fibromyalgia is highly prevalent and multidimensional, with predictors spanning psychiatric, psychosocial, behavioral, and physical domains. Although our models showed modest predictive accuracy, they reflect challenges in predicting suicide behavior in the literature. The results highlight the need for external validation, longitudinal designs, and active, targeted strategies to prevent premature deaths and better address the suffering experienced by people living with fibromyalgia.

References: Ordóñez-Carrasco JL, Sánchez-Castelló M, Calandre EP et al. Suicidal Ideation Profiles in Patients with Fibromyalgia Using Transdiagnostic Psychological and Fibromyalgia-Associated Variables. Int J Environ Res Public Health. 2020;30;18(1):209.

Presenting Author: Fernanda Bono Fukushima
Submitting Author: Fernanda Fukushima

Senior Author: Fernanda Bono Fukushima

IASP 2026 Global Year theme of Neuropathic Pain: No

• Lívia Agostinho Teixeira: BPharm PhD - UNESP
• Edison Iglesias de Oliveira Vidal: MD PhD - UNESP
• Mariana Suéte Guimarães Ruy: MD, PhdD student - UNESP
• Thais Alessandra Grecco: PsyD - UNESP
• Lucas Brito Barbieri: Ft. Master student - UNESP
• Fernanda Bono Fukushima: MD PhD - UNESP