Predicting Fear of Breast Cancer Recurrence from Healthcare Reimbursement Data using Deep Learning

Breast cancer is the prevalent and leading cause of mortality among women in France, with profound impacts on physical, emotional, and psychological well-being. Despite advances in treatment, the tear of cancer recurrence (FCR) persists among survivors and may lead to increased healthcare utilization and diminished overall well-being. However, accurately predicting FCR probability using traditional statistical models and machine learning (ML) algorithms remains a challenge due to the complex interplay of healthcare data over time. In this study, we propose an approach utilizing neural network-based predictive models and healthcare reimbursement data to predict FCR likelihood in women five years post-diagnosis. Our proposed method integrates temporal information and leverages both labeled and non-labeled medicoadministrative data through Neural Network (NN) architectures and semi-supervised learning techniques, offering a promising avenue for personalized risk assessments and tailored intervention strategies for BC survivors. The ongoing experimental results using ML and NN demonstrate promising outcomes, suggesting that neural network models may have the potential to further improve the prediction performance.