Análise preditiva baseada em inteligência artificial: um caminho para a transformação do modelo de vigilância das doenças crônicas não transmissíveis

Abstract

This dissertation, presented in the format of articles/chapters, aimed to identify prediction models based on Artificial Intelligence (AI) that assist in the surveillance of Chronic Noncommunicable Diseases (NCDs) in Brazil. The objective of Chapter I is to analyze the quality of life (QL) of the Brazilian population using the Alkire-Foster Method on microdata from the Surveillance Survey of Risk and Protective Factors for NCDs. Eleven indicators distributed across three dimensions were selected. QoL is a concept intensely marked by subjectivity, involving all essential components of the human condition, whether physical, psychological, social, cultural or spiritual. It is known that these diseases compromise much more than just the biological sphere and end up interfering in different ways with the lifestyle of the people affected, their family and social group. The results found indicate a high proportion of QoL deprivation among survey respondents. QoL appears to be strongly determined by socioeconomic factors, since chronic illness interrupts or hinders integration into the production process, reducing the possibilities of access to consumer goods. Based on the results obtained, the importance of considering the multidimensionality of the concept of QOL in the care of people with NCDs is suggested. Based on this argument, constructed in the first article, another theoretical study was proposed, which is found in Chapter II, whose objective is to present a synthesis of worldwide scientific articles published on predictive models for NCDs, and analyze how this topic has been addressed in Information Administration (ADI), on the area of Technology and Health. For this, a systematic literature review (RSL) was carried out, which searched for publications on the topic, on the SCOPUS and WEB of SCIENCE search platforms, between 2012 and 2022. This type of review contributes to mapping and, at the same time, synthesizing, in order to document recent discoveries on the researched topic. The set of 40 articles presents how the topic was treated at ADI, showing the evolution of the topic. The justification and relevance for this research lies in the need to understand how researchers explore the various applications of prediction techniques in the surveillance of these diseases. The contemporary use of Artificial Intelligence (AI) in healthcare is observed, through a research framework on the topic. Among future research, it is suggested to propose a Research Agenda with proposals for topics to be explored within the area of Information Systems. We move on to Chapter III, which presents an empirical study, using prediction models based on Machine Learning (ML) algorithms to predict risk in Obesity, Diabetes or both, based on variables collected in the VIGITEL 2021 survey The aim was to mathematically model a data set of 24,217 thousand survey respondents, based on thirteen input variables. As an output model, three tests were carried out. The models with the best performance were classified in the following order: Obesity, Obesity and Diabetes, Diabetes. It was concluded that the AM technique can improve health preparedness, as it allows digitally identifying subjects at greater risk of developing these diseases, recommending intervention before their worsening.