Intersectional analysis of inequalities in self-reported breast cancer screening attendance using supervised machine...

Area of Science:

• Public Health
• Epidemiology
• Health Disparities

Background:

• Breast cancer screening (BCS) is crucial for public health in Spain, with programs running since the 1990s.
• Despite bi-annual invitations for women aged 45-69, significant attendance inequalities persist across different population groups.
• This study uses a quantitative intersectional approach to identify at-risk groups for non-attendance.

Purpose of the Study:

• To identify intersectional positions associated with lower breast cancer screening attendance in Spain.
• To understand how social and demographic factors intersect to influence BCS uptake.
• To inform targeted interventions for improving BCS coverage.

Main Methods:

• Utilized data from the 2020 European Health Interview Survey in Spain (N=22,072).
• Applied inequality indicators from the PROGRESS-Plus framework.
• Employed decision tree models (CARTs, CHAID, CITs, C5.0) and an ensemble algorithm (XGBoost) to identify intersectional groups.

Main Results:

• XGBoost identified regional differences as the primary factor for BCS attendance, followed by education, age, and marital status.
• C5.0 model showed that the importance of individual characteristics varies by region and interaction.
• Highest risk of non-attendance was found in illiterate, older, lower social class women born in Spain residing in specific regions, and married/divorced/widowed women.

Conclusions:

• Decision trees and ensemble algorithms are effective tools for identifying at-risk populations for underutilization of public health resources.
• Findings can guide the development of targeted interventions to increase breast cancer screening attendance.
• Addressing intersectional inequalities is key to improving equitable access to cancer screening programs.