The Use of Artificial Intelligence Predictive Models in the Prevention of Noncommunicable Chronic Diseases

• Grant number: 24/16653-8
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Start date: October 01, 2024
• End date: September 30, 2025
• Field of knowledge: Interdisciplinary Subjects
• Principal Investigator: Jussara Marques de Almeida
• Grantee: Samuel Norberto Alves
• Host Institution: Instituto de Ciências Exatas (ICEx). Universidade Federal de Minas Gerais (UFMG). Ministério da Educação (Brasil). Belo Horizonte , SP, Brazil
• Company: Universidade Federal de Minas Gerais (UFMG). Instituto de Ciências Exatas (ICEx)
• Associated research grant: 20/09866-4 - Artificial Intelligence Innovation Center for Health (CIIA-Health), AP.PCPE

Abstract

Disease prevention, at the population and individual levels, aims to minimize the burden of disease and associated risk factors. In public health, the goal is to focus on disease prevention (primary, secondary and tertiary) and quality of life. In this context, the importance of Chronic Noncommunicable Diseases (NCDs) stands out, which account for 75% of annual deaths and high prevalence of risk factors in the Brazilian adult population.Currently, predictive models of artificial intelligence (AI) have been used to estimate the risk of a given outcome occurring (disease, death), based on the identification of socioeconomic, demographic, environmental variables and variables related to lifestyle habits and health conditions. In the context of NCDs, pioneering experiences such as the estimation of health indicators in small geographic areas stand out.This project focuses on the use of state-of-the-art Artificial Intelligence models to support the estimation (or prediction) of the prevalence of NCDs and associated outcomes in small areas. The development and validation of these predictive models pose important research challenges that are still being explored in the literature, notably challenges associated with the imputation of missing data and the development of accurate models in the face of low data availability as well as the biases often present in these data. Several data processing techniques are necessary to deal with the impact, on the prediction effectiveness, of biases, imbalances, inconsistencies, errors or absence in the data and of possible differences in the available aggregation levels.Furthermore, it is essential that techniques and models are aligned with ethical principles associated with transparency, privacy, impartiality, accountability, reliability and inclusion, and with aspects of related legislation. However, most state-of-the-art AI solutions focus on maximizing effectiveness, disregarding the impact that ethical principles have on them. For example, privacy guarantees are often achieved by hiding sensitive information. However, the resulting reduction of information can impact the learning of models and, consequently, their effectiveness. Thus, the development of solutions with a good compromise between prediction effectiveness and alignment with these principles remains an important open research problem.This project includes the estimation of NCD indicators related to morbidity, mortality and hospitalizations and their risk factors using data from the National Health Survey (PNS), the NCD Surveillance System by Telephone Survey (VIGITEL) and the Mortality Information System (SIM), census data and other databases.