| Cox proportional hazards model is a classical model in survival analysis,which is used to describe the relationship between survival life and covariates.In practical problems,some covariates may change with time.Therefore,this thesis discusses the calculation of value of Cox model with time-dependent covariates.Moreover,the lifetime may also depend on the history of the covariates,so we explore the Cox model with time-delay covariates.By using traditional methods such as Pvalue to select significant variables,it is found that the interaction effect(combined drugs)has a better impact on patient survival compared to drug alone.So we bring combined drugs as a covariate into models for analysis,and through likelihood ratio test it is found that the Cox with time-delay covariates is better than other models.In terms of model performance,the missing value of CD4 cell count at the time of death is implemented.Under the evaluation of C-index,the Cox model with time-delay covariates has a greater improvement,which has once again proven that considering historical moments is effective.Finally,the best model is used to provide personalized treatment for patients and determine the survival time of patients so as to provide better treatment and heathcare. |
