Previsão de AVC em Sergipe utilizando modelos de regressão logísticos com dados da Pesquisa Nacional de Saúde - PNS

Abstract

Stroke is one of the main causes of mortality and disability in Brazil, making it essential to create prediction tools that allow early intervention and reduce the impacts of this condition on public health. The general objective of this study is to develop a predictive model using logistic regression to predict the occurrence of Stroke based on data from the National Health Survey (PNS) for the State of Sergipe. Conducted in 2019 by the Ministry of Health in partnership with IBGE, the PNS aims to characterize the health situation and lifestyles of the population. To build the predictive model, the methodology adopted followed a quantitative approach. Given the unbalanced nature of the data, with a greater number of patients without a history of stroke compared to those with a history of the disease, balancing techniques such as over-sampling, under-sampling, and smote were applied. These methods were essential to balance the database before applying logistic regression, minimizing possible biases and maximizing the accuracy of the model. The development and analysis were performed in the R programming environment, which allowed the implementation of balancing techniques and the evaluation of the model’s performance. The results showed that the down-sampling model was able to better identify individuals with the disease, as it had good accuracy and sensitivity. The three most relevant variables for the model chosen to predict stroke were age, white skin color, and brown skin color. The model developed aims to help identify individuals at higher risk, providing valuable support for preventive strategies and informed decision-making in the health area in Sergipe.