Models for Predicting Depression Risk in Elderly Cancer Patients

Overview

Depression is common and detrimental among elderly cancer patients, impacting quality of life and survival. This study developed and validated risk prediction models (RPMs) using European cohort data to identify individuals at high risk for depression, leveraging clinical and psychosocial predictors.

Background

Depression in cancer patients is linked to poorer treatment adherence, immune dysfunction, and increased mortality. Early identification of high-risk individuals can facilitate timely interventions and personalized prevention. Existing predictive models are limited, often focusing on specific cancers or cross-sectional data. This study builds on prior general population models to create cancer-specific RPMs using machine learning and large-scale longitudinal data.

Data Highlights

The study utilized data from the SHARE cohort, including Europeans aged 55 and older with a history of cancer, collected between 2013 and 2019. Depression was assessed via the EURO-D scale, with a score ≥4 indicating major depression. Sample size exceeded 1,000 events, meeting recommended thresholds for model development. Predictors included demographic, psychosocial, and cancer-related variables available through questionnaires.

Key Findings

• Depression prevalence was high among elderly cancer patients, necessitating effective risk stratification tools.
• Multiple RPMs were developed using advanced machine learning methods, demonstrating feasibility in this population.
• A parsimonious subset of predictors retained strong predictive power, facilitating practical application.
• Psychosocial and demographic factors played a significant role alongside cancer-specific variables in predicting depression risk.
• The final model was implemented as a publicly accessible web-based risk calculator to support clinical use.

Clinical Implications

Clinicians can use these RPMs to identify elderly cancer patients at elevated risk for depression, enabling earlier psychological assessment and intervention. Incorporating both psychosocial and clinical factors improves prediction accuracy. The accessible web-based tool supports personalized care planning and resource allocation in oncology settings.

Conclusion

This study successfully developed and validated machine learning-based RPMs for depression risk in elderly cancer patients, highlighting the importance of multidimensional predictors. The publicly available calculator offers a practical resource to enhance early detection and management of depression in this vulnerable population.

References

1. Manto et al. 2022 -- Development of a Depression Risk Prediction Model in Older Adults
2. SHARE Project 2013-2019 -- Survey of Health, Ageing and Retirement in Europe
3. Prince et al. 1999 -- EURO-D Scale for Depression Assessment
4. TRIPOD-AI 2023 -- Guidelines for Reporting AI Prediction Models