Modelos de regressão Log-Normal e Burr-XII aplicados a dados de sobrevivência de pacientes com COVID-19 no Distrito Federal
Modelos de regressão Log-Normal e Burr-XII aplicados a dados de sobrevivência de pacientes com COVID-19 no Distrito Federal
-
DOI: https://doi.org/10.22533/at.ed.8119112621013
-
Palavras-chave: Distribuição Burr-XII; Distribuição Log-Normal; Modelos de regressão; Dados censurados; COVID-19.
-
Keywords: Burr-XII distribution; Log-Normal distribution; Regression models; Censored data; COVID-19.
-
Abstract: This study aims to identify factors associated with the time of hospitalization until death among patients hospitalized with COVID-19 in the public healthcare system of the Federal District, Brazil, between 2020 and 2023, using parametric survival analysis techniques. Considering the unimodal behavior of the hazard function, the following distributions were evaluated to model the time variable: Log-Logistic, Log-Normal, Reparameterized Inverse Gaussian, Burr-XII, Kumaraswamy-Log-Normal, Kumaraswamy-Log-Logistic, KumaraswamyReparameterized Inverse Gaussian, and Kumaraswamy-Burr-XII. The final models, based on the Log-Normal and Burr-XII distributions, incorporated patients’ clinical and demographic characteristics, indicating that age, sex, hospitalization period, total hospitalization cost, and the interaction between period and cost were statistically significantfactors. The results revealed a lower probability of survival among older patients and male individuals. Furthermore, during the post-pandemic period, higher hospitalization expenditures were associated with a greater probability of survival when compared to the pandemic period. Model adequacy was assessed using Cox–Snell residuals, and a simulation study confirmed the consistency of the estimation method under different sample sizes and censoring levels.
- Carolyne Soares de Brito
- Juliana Betini Fachini Gomes