OBJECTIVES: Breast cancer (BC) is a high incident neoplasia worldwide with a high socio-economic impact and a significant pressure in the public healthcare system costs. The aim of this study is to apply Big Data Analytics for the development of an Oncology Structured Information System able to predict incidence, prevalence and costs of HER2+ BC in the Brazilian’s public healthcare system at municipal level.

METHODS: A data lake was created from public databases, provided by Brazilian Public Health System (SUS), National Cancer Institute (INCA) and National Healthcare Agency (ANS), comprising healthcare and demographic information of all Brazilian municipalities. Data records were spliced into 8 different groups at anonymized patient level and based on treatment profile: Chemotherapy (C), Radiotherapy (R), Surgery (S), Chemo+Radio (C+R), Chemo+Surg (C+S), Radio+Surg (R+S), Chemo+Radio+Surg (C+R+S) and No Treatment (N) in order to cover 100% of the patients among the 03 different systems. An algorithm was developed in order to harmonize and prepare the data for statistical analysis. A restrospective period of 10 years (2017-2008) was used to build the linear regression methodology, using additive trend and seasonality, with a confidence interval of 95%.

RESULTS: The model aplication resulted in an automated Oncology Structured Information System (OSIS) which comprises more than 3 billion records, including demographics, epidemics, cancer incidence, prevalence, healthcare procedures and infrastructure for all the 5,570 Brazilian municipalities. The OSIS is automatic updated in a monthly basis, presenting an accuracy of 97,8% for a 3 years prediction.

CONCLUSIONS: The use of big data analytics allow the development of models and tools for estimating BC incidence and prevalence of patients with HER2+ BC, at municipal level in Brazil. Such tool can support health managers in better planning and policies development in a more detailed and granular perspective.