Aplicações de análise de sobrevivência na área médica e a escolha de modelos paramétricos

Abstract

Survival analysis is data whose dependent variable is the time until the occurrence of an event of interest. In applications in the medical field is common to use non-parametric methods (Kaplan-Meier) and semi-parametric (Cox proportional hazards model) to study patient survival. Parametric models are also attractive in those cases, since the proportional hazards assumption is required. In this approach several types of parametric distributions are available. The choice of a parametric model can be made by statistical tests based on a more general parametric family. The distribution generalized gamma (GG) is a flexible parametric family which includes most commonly used distributions commonly used in the literature for this purpose. However the setting and test for this model based on Likelihood methods require the use of optimization procedures implemented in statistical software. The best-known routines R software for GG setting often have limitations and can not work that in more recent versions of R. In this context, the objective of this paper is to briefly describe procedures for adjustment and choice of parametric survival models based on usage computational routines developed in Silva (2013) in R software via adaptations of flexsurv package (flexible parametric survival models). Are considered three applications in the medical field to illustrate the procedures studied, two sets of data available in the literature (time between occurrences of childhood diarrhea and survival of patients with cancer in the lung) and assigned by Applied Statistics Laboratory (LEA) UFRN which refers to the time after the organ donation diagnosis of brain death.