Modelagem em análise de sobrevivencia com eventos recorrentes aplicada a dados da área médica

• Ano de defesa: 2013
• Autor(a) principal: Mota, Thiago Santos [UNESP]
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Estadual Paulista (Unesp)
• Palavras-chave em Português: Biometria; Dialise; Insuficiencia renal cronica; Dialysis
• Link de acesso: http://hdl.handle.net/11449/108465

The aim is to study the marginal and conditional models used in multivariate survival analysis and to evaluate the effect of the covariates measuared over time for ocurrence of recurrring cardiovascular events in patients under dialysis treatment from clinics hospital of the Medicine Faculty, UNESP - Botucatu, SP, with follow-up time from 2008 to 2011. As the data treat the recurring events of the same type, the marginals models – AG (Andersen & Gill, 1982), PWP (Prentice et al., 1981) and WLW (Wei et al., 1989) – and the conditionals models – frailty models with lognormal and gamma distributions – were used in the modeling. The covariates considered in this study were clinical, nutritional, laboratory and dialytic. The covariates were considered significant at 5% in all models using the covariates selection criteria: backward, forward and stepwise. It also presented the values of the criteria of AKAIKE (AIC), corrected criteria of AKAIKE and criteria of the Bayes (BIC) for these models. Other issues were discussed to select the best model for the data set, as frequency and the influence of previous events to the occurrence of new events. Based on these criteria, the PWP model was selected to model the data. Residual analysis was done by using the Shoenfeld and Martigale residuals, showed that this model fits the data well. All the results were obtained in R software (R Development Core Team, 2012). The results showed association between the quotient extracellular mass by body cell mass (MEC/MCC) covariate and the occurrence of cardiovascular events in dialysis patients such exploration such finding is new for multiple events per individual