PATTERNS OF BREAST CANCER SCREENING UTILIZATION IN BRAZIL’S PRIVATE HEALTH SECTOR: AN ARTIFICIAL INTELLIGENCE ANALYSIS, 2014-2023
Author(s)
Mônica V. Andrade, PhD1, Kenya Noronha, PhD2, Leonardo C. Ribeiro, PhD2, Silvana M. Kelles, PhD3, Mariangela Cherchiglia, PhD4, Lucas R. Carvalho, PhD1, Nayara A. Julião, PhD1, Sergio L. Bersan, MSc5, Maria Luisa Rigotti, B.Sc.2, Flávia Colares, M.D., M.S.6, Flávia C. Almeida, M.D.7, Natalia G. Silva, B.Sc.8, Pedro Benner, B.Sc.2, Clara S. Ribeiro, UG Student2, Pedro H. Amorin, UG Student2, Carolina Carvalho, UG Student2, Henrique Bracarense, MSc2, Marcus C. Borin, PhD9, Mariana M. Barbosa, PhD10;
1CEDEPLAR/UFMG, Department of Economics, Belo Horizonte, Brazil, 2CEDEPLAR/UFMG, Belo Horizonte, Brazil, 3Unimed-BH Health Technology Assessment Center (NATS)/Pontifical Catholic University of Minas Gerais (PUC Minas) - Betim, Belo Horizonte, Brazil, 4Universidade Federal de Minas Gerais, Programa de Pós-Graduação em Saúde Pública, Departamento de Medicina Preventiva e Social, Faculdade, Belo Horizonte, Brazil, 5Pontifical Catholic University of Minas Gerais (PUC Minas), Betim, Brazil, 6UNIMED-BH, CASU UFMG, Belo Horizonte, Brazil, 7Oncomed BH - Grupo Orizonti, Belo Horizonte, Brazil, 8Universidade Federal de Minas Gerais, Physics Department, Belo Horizonte, Brazil, 9Unimed-BH Health Technology Assessment Center (NATS)/ Drug Market Regulation Chamber (CMED) – Brazilian Health Regulatory Agency (ANVISA), Brasilia Federal District, Belo Horizonte, Brazil, 10Unimed-BH Health Technology Assessment Center (NATS), Belo Horizonte MG/ UFMG, Belo Horizonte, Brazil
1CEDEPLAR/UFMG, Department of Economics, Belo Horizonte, Brazil, 2CEDEPLAR/UFMG, Belo Horizonte, Brazil, 3Unimed-BH Health Technology Assessment Center (NATS)/Pontifical Catholic University of Minas Gerais (PUC Minas) - Betim, Belo Horizonte, Brazil, 4Universidade Federal de Minas Gerais, Programa de Pós-Graduação em Saúde Pública, Departamento de Medicina Preventiva e Social, Faculdade, Belo Horizonte, Brazil, 5Pontifical Catholic University of Minas Gerais (PUC Minas), Betim, Brazil, 6UNIMED-BH, CASU UFMG, Belo Horizonte, Brazil, 7Oncomed BH - Grupo Orizonti, Belo Horizonte, Brazil, 8Universidade Federal de Minas Gerais, Physics Department, Belo Horizonte, Brazil, 9Unimed-BH Health Technology Assessment Center (NATS)/ Drug Market Regulation Chamber (CMED) – Brazilian Health Regulatory Agency (ANVISA), Brasilia Federal District, Belo Horizonte, Brazil, 10Unimed-BH Health Technology Assessment Center (NATS), Belo Horizonte MG/ UFMG, Belo Horizonte, Brazil
OBJECTIVES: In Brazil, malignant neoplasms are a leading cause of death, and early detection is crucial. However, screening protocols are inconsistently followed, and the widespread availability of services often results in unnecessary overuse, particularly in private sector. This study identifies pathways of breast cancer screening utilization using a unique longitudinal administrative database from a Brazilian health insurer, including beneficiary attributes and healthcare utilization.
METHODS: This observational, longitudinal, study analyzes breast cancer screening using administrative data from a large Brazilian health insurance provider (2014-2023). Artificial intelligence algorithms clustered individual mammography utilization trajectories over ten years, focusing on frequency and timing. The mathematical tools used to identify the patterns were: convolution, Fourier transformation, t-SNE (dimensionality reduction) and DBSCAN (clustering). A neural network was also trained to predict breast cancer occurrence based on the use of 27 types of screening, diagnostic, and treatment procedures. Clusters were characterized by screening frequency, timing, age group, and breast cancer incidence
RESULTS: Using AI-based clustering, 13 macro-groups were identified. Patterns ranged from minimal use (1-3 screenings in the decade) to consistent annual use. Some groups showed intermittent behavior, with clustered use in specific years. Older women (average age 49) show the highest adherence rates, with over 80% of women screened annually, and experience the highest breast cancer incidence, reaching up to 34%. Groups with sporadic screening were younger and had lower incidence rates. These findings align with the epidemiological profile of breast cancer in Brazil. A noticeable drop in utilization occurred around 2020 across most groups, likely due to the COVID-19 pandemic.
CONCLUSIONS: These differentiated screening trajectories offer valuable insights for tailoring public health strategies and improving early detection efforts. Using cost information and the estimated breast cancer incidence for each screening profile, cost-effectiveness indicators can be calculated to inform clinical practice and resource allocation
METHODS: This observational, longitudinal, study analyzes breast cancer screening using administrative data from a large Brazilian health insurance provider (2014-2023). Artificial intelligence algorithms clustered individual mammography utilization trajectories over ten years, focusing on frequency and timing. The mathematical tools used to identify the patterns were: convolution, Fourier transformation, t-SNE (dimensionality reduction) and DBSCAN (clustering). A neural network was also trained to predict breast cancer occurrence based on the use of 27 types of screening, diagnostic, and treatment procedures. Clusters were characterized by screening frequency, timing, age group, and breast cancer incidence
RESULTS: Using AI-based clustering, 13 macro-groups were identified. Patterns ranged from minimal use (1-3 screenings in the decade) to consistent annual use. Some groups showed intermittent behavior, with clustered use in specific years. Older women (average age 49) show the highest adherence rates, with over 80% of women screened annually, and experience the highest breast cancer incidence, reaching up to 34%. Groups with sporadic screening were younger and had lower incidence rates. These findings align with the epidemiological profile of breast cancer in Brazil. A noticeable drop in utilization occurred around 2020 across most groups, likely due to the COVID-19 pandemic.
CONCLUSIONS: These differentiated screening trajectories offer valuable insights for tailoring public health strategies and improving early detection efforts. Using cost information and the estimated breast cancer incidence for each screening profile, cost-effectiveness indicators can be calculated to inform clinical practice and resource allocation
Conference/Value in Health Info
2026-05, ISPOR 2026, Philadelphia, PA, USA
Value in Health, Volume 29, Issue S6
Code
MSR220
Topic
Methodological & Statistical Research
Topic Subcategory
Artificial Intelligence, Machine Learning, Predictive Analytics
Disease
No Additional Disease & Conditions/Specialized Treatment Areas, SDC: Oncology