Developing a model for predicting suicide risk among prostate cancer survivors

Area of Science:

• Oncology
• Psychiatry
• Biostatistics

Background:

• Cancer survivors face elevated suicide risks compared to the general population.
• Prostate cancer survivors represent the largest group of cancer survivors, necessitating targeted risk assessment.
• Predicting suicide risk in this vulnerable population is crucial for effective intervention.

Purpose of the Study:

• To develop and validate a predictive model for suicide risk in prostate cancer survivors.
• To identify key clinical predictors associated with increased suicide risk.
• To provide a tool for early identification of high-risk individuals.

Main Methods:

• Utilized the Surveillance, Epidemiology, and End Results (SEER) database, including 238,534 prostate cancer patients.
• Employed Cox regression, Best Subset Regression (BSR), and LASSO for initial variable selection.
• Selected final model variables using backward stepwise Cox regression and evaluated performance with C-indices and ROC curves.

Main Results:

• A model incorporating age, race, marital status, income, PSA levels, M stage, and surgical status was developed.
• The model showed good discriminative ability (C-indices 0.702 training, 0.688 validation).
• High-risk survivors identified by the model had a 3.5 times higher suicide risk.

Conclusions:

• A reliable, seven-predictor model for prostate cancer survivor suicide risk was successfully established.
• The model aids healthcare professionals in identifying high-risk survivors for prompt preventive measures.
• This predictive tool supports timely interventions to mitigate suicide risk in this population.