Multilevel Predictive Study of Breast Cancer Screening Patterns in Pennsylvania, 2021-2022.

Area of Science:

• Public Health
• Health Disparities
• Machine Learning in Healthcare

Background:

• Social determinants of health (SDOH) are critical factors influencing cancer screening adherence.
• Understanding geosocial risk factors is essential for targeted interventions.
• Machine learning offers novel approaches to analyze complex health data.

Purpose of the Study:

• To identify personal, systemic, and geosocial risk factors associated with breast cancer screening nonadherence.
• To evaluate the efficacy of machine learning techniques in enhancing cancer screening rates.

Main Methods:

• A cohort of 21,543 women aged 50-74 from 15 family medicine clinics in Pennsylvania (2021-2022) was analyzed.
• Geosocial data were geocoded to Census blocks, incorporating neighborhood socioeconomic measures.
• Multiple logistic regression and machine learning-derived Area Deprivation Index (ADI) were used to assess adherence predictors.

Main Results:

• Minority women, excluding Hispanic individuals, showed decreased odds of screening (aOR: 0.659).
• Uninsured (aOR: 0.334) and Medicaid recipients (aOR: 0.793) had higher nonadherence risks.
• Area Deprivation Index (ADI) was inversely proportional to breast cancer screening rates.

Conclusions:

• Social determinants of health significantly affect breast cancer screening likelihood.
• Identifying and addressing compounding barriers is vital for improving screening adherence.
• Machine learning-assisted analysis of geosocial factors can inform public health strategies.