GAMLSS Models: Application in cases of Severe Acute Respiratory Syndrome with emphasis on Influenza and other causes

Abstract

One of the major causes of Severe Acute Respiratory Syndrome (SARS) is Influenza, a respiratory disease caused by several strains of the virus. The global spread of Influenza has represented a significant challenge since 2009. In this context, this research proposes to analyze SARS with an emphasis on Influenza and other causes (viruses and other etiological agents), in order to investigate and understand the relationship between predictor variables (comorbidities) and the specific response variable (evolution) in the cases registered in Brazil. From the analysis of the results found, referring to the period from 2020 to August 2022, it was noted that the data indicated an unsatisfactory fit to the Multiple Linear Regression model and the Generalized Linear Models, which motivated the use of Generalized Additive Models for Location, Scale and Shape. To analyze these data, we used classical probability distributions in the context of this modeling: binomial, negative binomial, geometric and Poisson. Of these distributions worked, the binomial distribution proved to be effective, allowing us to map the adequacy of this modeling. Our training and test sets were performed in order to verify possible overfitting in the obtained model. This approach provided a deeper understanding of the data set and the relationships between the studied variations. Asthmatic patients demonstrated a strong and significant association with the mortality rate, while those with chronic liver conditions had a lower risk of death. These results provide important strategies for intervention and care of populations affected by these health conditions.