Personalized Breast Cancer Screening

Area of Science:

• Oncology
• Medical Informatics
• Machine Learning

Background:

• Current cancer screening guidelines primarily use patient age, potentially leading to delayed or excessive screenings.
• Existing systems lack interoperability, hindering the integration of patient data across healthcare providers.
• Individual medical characteristics are often overlooked in standardized screening protocols.

Purpose of the Study:

• To develop a clinical support tool using claims data for enhanced physician decision-making in cancer screening.
• To create a machine learning framework for personalized, dynamic, and data-driven cancer screening recommendations.
• To address the limitations of age-centric screening by incorporating individual patient data.

Main Methods:

• Utilized claims data and medical insurance transactions with standardized coding for diagnoses, procedures, and medications.
• Developed a novel machine learning framework to generate personalized screening recommendations.
• Applied the methodology to breast cancer mammogram screening using data from 378,840 female patients.

Main Results:

• Personalized screening demonstrated a statistically significant reduction in average cancer diagnosis delay by 2-3 months across diverse risk populations.
• Individual patient benefits showed even greater improvements, with delays reduced by up to 10 months.
• The study highlights the effectiveness of data-driven approaches in optimizing screening timelines.

Conclusions:

• Integrating personal medical characteristics and machine learning into cancer screening enhances timeliness and adapts to evolving patient risks.
• The proposed methodology offers a valuable support tool for clinicians, improving screening decisions.
• Future implementation in healthcare settings can lead to more effective and personalized cancer care pathways.